Das also pointed out that coordination between the central government, particularly finance ministry, and RBI in the last six years has been at its best

Finance Ministry, RBI's Perspective May Differ At Times, Says Shaktikanta Das

From the historical issues of data annotation to the technological innovations brought by Web3, Sahara AI, Alaya AI, and Public AI demonstrate the ability of emerging technologies to reshape traditional industries.

Blockchain-Driven AI Data Annotation: Breakthroughs in the Web3 Era from CZ's Perspective - Detailed Analysis of Projects such as Sahara AI, Alaya AI, Public AI, etc - ChainCatcher

Personal financial planning isn't just about spreadsheets and savings accounts. It’s also about creating a life where we can find joy, security, and growth opportunities.

Reasons To Be Thankful This Season: A Personal Finance Perspective

Early learning specialists explain the journey of shifting schools to embrace science-based reading instruction.

Perspective | Solving the literacy puzzle: How the science of reading is transforming our schools

Attosecond science is a versatile discipline for studying ultrafast dynamics in matter on the microscopic scale. This Perspective explores the theoretical and experimental developments in this field focusing on distinguishing genuinely quantum observations from classical phenomena.

Quantum phenomena in attosecond science | Nature Reviews Physics

Perspective

perspective

The solutions that ARPA-H seeks have the potential to vastly improve healthcare, but could also contribute to a technocratic, biomedical security state: perspective The Advanced Research Projects Agency for Health (ARPA-H) announces its first four focus areas for funding opportunities. Today, ARPA-H launched its first (BAA), identifying four initial focus areas for investment that touch upon elements of the (IoB), genetic engineering, vaccine development, tissue regeneration, and AI-enabled health surveillance, data collection, and analyses. Open Broad Agency Announcement Internet of Bodies https://twitter.com/TimHinchliffe/status/1636080829381107712?embedable=true The solutions that ARPA-H seeks have the potential to vastly improve the overall health and wellbeing of citizens, but at the same time they could also contribute to a technocratic, biomedical security state in which citizens’ physical and mental health statuses are constantly monitored and surveilled to enforce compliance with whatever mandates public and private entities decide to issue. As investigative journalist and author back in May, 2021, ARPA-H could “ . Whitney Webb warned in Unlimited Hangout merge ‘national security’ with ‘health security’ in such as way as to use both physical and mental health ‘warning signs’ to prevent outbreaks of disease or violence before they occur “ ”. Such a system is a recipe for a technocratic ‘pre-crime’ organization with the potential to criminalize both mental and physical illness as well as ‘wrongthink Now, let’s take a look at the first four focus areas that ARPA-H is looking to fund. “Topics of interest include […] New approaches to accelerate and routinize mammalian and microbial cellular engineering to enable next generation therapeutic applications” — ARPA-H, Open BAA, 2023 https://twitter.com/ARPA_H/status/1635993598222282753?embedable=true The four initial focus areas in the ARPA-H Open BAA include: , . Health Science Futures which seeks to develop the innovative tools, technologies, and platforms that can be applied to a broad range of diseases , . Scalable Solutions which seeks to improve access and affordability and address health ecosystem challenges that impede effective and timely development and distribution of healthcare and disease response at a scale , Proactive Health which seeks to improve personal health and wellness to reduce the likelihood that people require medical intervention or minimize the time that they remain in acute care through accelerated recovery and regeneration capabilities. , Resilient Systems which seeks to create capabilities, develop mechanisms, and accelerate system integrations to enhance stability in the face of disruptive events. Out of the four, appears to have the most red flags in terms of overall digital surveillance and privacy concerns for citizens. Proactive Health For example, for , ARPA-H is looking to: Proactive Health . Characterize brain and other deep tissue and organ health with quantitative and accurate outcomes equivalent to invasive health monitoring . Develop robotics, wearables, and other devices to enhance independence for aging populations and people with cognitive or movement disorders . Develop novel approaches to continuously measure, analyze, and enhance health promoting activities Again, these solutions could represent breakthroughs in improving human health, but in order to be successful, they require constant surveillance and data collecting. With ARPA-H looking to “wearables and other devices,” the government agency is looking to tap into the Internet of Bodies ecosystem of that measures and surveils everything that is going on inside your body. wearables, implantables, and consumables According to a World Economic Forum (WEF) , the IoB is an ecosystem consisting of “an unprecedented number of sensors attached to, implanted within, or ingested into human bodies .” briefing paper to monitor, analyze, and even modify human bodies and behavior In 2020, the stating that the IoB “ .” RAND corporation published a report might trigger breakthroughs in medical knowledge […] Or it might enable a surveillance state of unprecedented intrusion and consequence Another observation that Webb made in her 2021 report was that “If ARPA-H/HARPA is approved by Congress and ultimately established [ ], it will be used to resurrect dangerous and long-standing agendas of the national-security state and its Silicon Valley contractors, .” Unlimited Hangout which it now is creating a ‘digital dictatorship’ that threatens human freedom, human society, and potentially the very definition of what it means to be human For years, historian Yuval Noah Harari has warned about potential , and at the 2020 WEF annual meeting in Davos he declared that “ .” digital dictatorships We are no longer mysterious souls; we are now hackable animals , “To hack human beings you need a lot of biological knowledge, a lot of computing power, and especially a lot of data. According to Harari “ .” If you have enough data about me and enough computing power and biological knowledge, you can hack my body, my brain, my life. You can reach a point where you know me better than I know myself https://youtu.be/eOsKFOrW5h8?embedable=true The historian even came up with a “danger formula” for hacking human beings, which he believes “might be the defining equation of life in the 21st Century.” That equation is — which means iological knowledge multiplied by omputing power multiplied by ata equals the bility to ack umans. B x C x D = AHH B C D A H H “The power to hack human beings can of course be used for good purposes like provided much better healthcare,” said Harari, adding, “but if this power falls into the hands of a 21st Century Stalin, the result will be the worst totalitarian regime in human history, and we already have a number of applicants for the job of 21st Century Stalin.” https://youtu.be/-0bXUxRqy8g?embedable=true With today’s Open BAA, ARPA-H is looking to invest in research that truly has the potential to by merging humans with machines and through genetic manipulation. fundamentally alter what it means to be human Apart from funding research into the IoB, ARPA-H is looking to invest in: New approaches to accelerate and routinize mammalian and microbial cellular engineering to enable next generation therapeutic applications, develop multiscale interventions, and automate hypothesis generation and discovery to expand those applications to disease states in which cellular therapies have not traditionally been employed. Novel methods to engineer resilient tissues, microbiomes, and biophysical systems to combat disease or maintain health. Foundational advances in genetic, cellular, tissue, and organ replacement therapies that enable personalized medical interventions at scale. Novel techniques to reduce the spread of disease or eliminate risk factors, including new vaccine or therapeutic modalities that block pathogen transmission, induce mucosal immunity, or boost or sustain native immunity without triggering auto-immune dysfunction. https://youtu.be/d5QZf9-joO4?embedable=true Other areas of interest in the ARPA-H Open BAA include: Strategies and technologies to leverage homes, community centers, pharmacies, and other accessible locations as distributed clinical trial sites to diversify participation in clinical trials and integrate end-user feedback to rapidly iterate prototype designs. Novel ways to protect, secure, integrate, analyze, communicate, and present health data, including but not limited to advances in privacy, cyber security, artificial intelligence with enhanced patient safety properties, low-code or no-code technologies, semantic approaches, and rapid integration techniques. AI-enabled, and empirically validated physiological models that accurately reflect the biological basis of complex diseases, the interface between biological and physical systems, and mimic human response to potential therapeutic or multiscale interventions from the atomic/molecular to systemic/whole human scales. Approaches to build trust in the healthcare system and distribute high-quality health guidance in an understandable manner that improves patient outcomes. The Open BAA comes exactly one year since ARPA-H was established within the US Department of Health and Human Services (HHS) on March 15, 2022. This article was originally published by Tim Hinchliffe on The Sociable.

ARPA-H Announces 4 Focus Areas for Funding

ARPA-H anuncia 4 áreas de enfoque para la financiación

L'ARPA-H annonce 4 domaines prioritaires pour le financement

एआरपीए-एच ने फंडिंग के लिए 4 फोकस क्षेत्रों की घोषणा की

ARPA-H が資金調達のための 4 つの重点分野を発表

ARPA-H anuncia 4 áreas de foco para financiamento

ARPA-H công bố 4 lĩnh vực tài trợ trọng tâm

ARPA-H 宣布 4 个重点资助领域

Preface*: I spent 3 days with MBA students from 2 of the top 10 universities in the world. Our interactions were within a school environment and comprised of academic and social activities. This includes almost 8 hours of lecture time, 5 hours in parties, and multiple formal and informal interactions over 3 days. Below is a summary of what I uncovered.* “I have my suit packed” I texted my friend half joking, half serious. I was going to be spending the next 3 days with a friend at one of the top 10 universities in the world, and I was going to be participating in an MBA class. I was excited to see my friend, but I didn’t know what to expect from the MBA class and his colleagues. You see, my friend and I have always been different in our approach to career growth. Although we both understood the value of hard work, I relied a lot on my hard skills while his approach placed a great deal of emphasis on soft skills. I always supported his decision to pursue an MBA, but I never understood why he wanted to be a “suit”. To understand my thought process, you have to understand that I started my career initially as a freelance software engineer almost 13 years ago. You know, when tech was about being ‘hardcore.’ My first full-time job was at an organization that treated business team members & MBAs better than members of the tech team. It always felt like a situation of ‘us versus them.’ It was common for members of the tech team to do an online MBA and make the switch to the business team just for the chance at better pay & recognition. The movies and podcasts I consumed also backed this up. MBAs (typically called “suits”) were often portrayed as trust fund know-it-alls who were far removed from the realities of everyday people, didn’t understand how things worked, and somehow felt superior to others. I unconsciously carried this around with me. First Impression My first day was fairly quiet. I spent part of it walking around the campus and understanding where the key areas were. I had started meeting some people from the MBA program, and the first thing I noticed was how ethnically diverse they were. They were also quite friendly and happy to strike up a conversation with a stranger. I got some food, watched a movie, and had an early night in preparation for the class tomorrow. Observations in an Academic Setting The next day, I was sitting in a Negotiation class from 9 am to 4:45 pm. The class included a lecture, a case study based on a variation of real events, a debrief session, and a take-home assignment. People were assigned different roles in the case study. They were given clear but conflicting deliverables that they needed to get out of the negotiation in order for their character to be successful. I sat quietly in the corner, observing; not sure what to expect. The negotiations kicked off and it all felt so real! Everyone embodied their character and was determined to get what they wanted. There was also something different about it. This didn’t feel like a gathering of people who didn’t know how the real world worked. It felt like quite the opposite. I saw people use elite tactics including teamwork and betrayal to get what they wanted from the deal. After negotiating intensely for hours, the group I was part of finally came to a conclusion that was suitable for all parties. When the class was over, I had a chat with a few of the students to understand their motivation to get an MBA and how the journey had been so far. They candidly shared their motivations, which I have summarized in the end notes of this post. Observation in Socials I kicked off socials at a barbecue party with about 50 MBA students moving in and out. As I moved around and got introduced to various groups, our conversations ranged from who made the best soup to Man Utd’s performance this season. It was not what I expected. To be honest, the conversations were a lot more down-to-earth than conversations I’ve had at tech/start-up social events. Nobody was trying to portray themselves as a ‘genius’ or the person who is changing the world. Even though a lot of the people there had done great things before they joined the program. The next stop was a party at one of the popular spots on campus. I was attending with a smaller group as most people were tired from the barbecue or had work to do. We had pre-drinks in one of the dorms and headed out for the night. This time, I spent more time observing the group than actually participating. I could see clearly that they had started building strong friendships that would last a lifetime. Bonding over shared suffering, happiness, and shared destinations. Perhaps this is why it could sometimes seem to outsiders like there’s some sort of ‘clique’. They were spending all this time together and were starting to build friendships that they would gladly defend if necessary. Conclusion/End Notes I woke up on the final day tired but filled with a feeling that I couldn’t quite explain. That was when I noticed it for the first time in 3 days. It was a sticker on the only window in the dormitory that I was in. “Build something Massive.” was boldly written across it. It summarized my thoughts about the MBA program and the students I had met. The program was a chance for ambitious people to learn and to build their lives and careers into something massive, and most of them sacrificed a lot for this opportunity. Over 90% of them are immigrants.
A lot of them are on scholarships that they had to fight for, while some took loans to attend the program.
A lot of them are anxious and working hard to find internships/jobs months before they graduate because they feel the need to start providing value that can offset the cost of their program.
Some of them left their families and cushy jobs/businesses to move into student accommodation for this program. I believe that the summary of these factors is why some of them seem to have a chip on their shoulders when they graduate. They believe that they have sacrificed a lot and deserve to be where they are. If it was you, you may feel the same way too. It’s humbling for me to write this as it exposes my past ignorance. But it’s important to put this out because there are many people who may be even more ignorant. So what did I really uncover about these MBAs? I learned that They’re parents. They’re lovers. They’re dreamers. They reflect everything that’s great about the world and everything that is wrong about it too. (Just like the rest of us.) The world has promised them in this unwritten contract that if they do [x] (i.e., get a prestigious MBA), then doors will open for them. That their hard work and sacrifice would be rewarded. For a lot of them, the MBA is their path to a better life, recognition for their hard work, and a better future. During my time there, I felt very welcome in every circle, though I was meeting a lot of them for the first time. In fact, I may have annoyed some people anytime I was visibly shocked by something they did that went against the stereotype of an MBA. Would I get an MBA? I don’t know yet. But I have even more respect and appreciation for everyone who has gotten an MBA and everyone who plans to get one. You don’t know me, but I’m proud of you. I want to thank Professor Owen for letting me sit in on the class. I enjoyed your approach to teaching the subject, and I appreciate how attentive you were to everyone. I learned a lot about negotiating and about people. Preface *: I spent 3 days with MBA students from 2 of the top 10 universities in the world. Our interactions were within a school environment and comprised of academic and social activities. This includes almost 8 hours of lecture time, 5 hours in parties, and multiple formal and informal interactions over 3 days. Below is a summary of what I uncovered.* Preface Preface “I have my suit packed” I texted my friend half joking, half serious. I was going to be spending the next 3 days with a friend at one of the top 10 universities in the world, and I was going to be participating in an MBA class. I was excited to see my friend, but I didn’t know what to expect from the MBA class and his colleagues. You see, my friend and I have always been different in our approach to career growth. Although we both understood the value of hard work, I relied a lot on my hard skills while his approach placed a great deal of emphasis on soft skills. I always supported his decision to pursue an MBA, but I never understood why he wanted to be a “suit”. To understand my thought process, you have to understand that I started my career initially as a freelance software engineer almost 13 years ago. You know, when tech was about being ‘hardcore.’ My first full-time job was at an organization that treated business team members & MBAs better than members of the tech team. It always felt like a situation of ‘ us versus them. ’ us versus them. It was common for members of the tech team to do an online MBA and make the switch to the business team just for the chance at better pay & recognition. The movies and podcasts I consumed also backed this up. MBAs (typically called “suits”) were often portrayed as trust fund know-it-alls who were far removed from the realities of everyday people, didn’t understand how things worked, and somehow felt superior to others. I unconsciously carried this around with me. First Impression First Impression My first day was fairly quiet. I spent part of it walking around the campus and understanding where the key areas were. I had started meeting some people from the MBA program, and the first thing I noticed was how ethnically diverse they were. They were also quite friendly and happy to strike up a conversation with a stranger. I got some food, watched a movie, and had an early night in preparation for the class tomorrow. Observations in an Academic Setting Observations in an Academic Setting The next day, I was sitting in a Negotiation class from 9 am to 4:45 pm. The class included a lecture, a case study based on a variation of real events , a debrief session, and a take-home assignment. real events People were assigned different roles in the case study. They were given clear but conflicting deliverables that they needed to get out of the negotiation in order for their character to be successful. I sat quietly in the corner, observing; not sure what to expect. The negotiations kicked off and it all felt so real! Everyone embodied their character and was determined to get what they wanted. There was also something different about it. This didn’t feel like a gathering of people who didn’t know how the real world worked. It felt like quite the opposite. I saw people use elite tactics including teamwork and betrayal to get what they wanted from the deal. After negotiating intensely for hours, the group I was part of finally came to a conclusion that was suitable for all parties. When the class was over, I had a chat with a few of the students to understand their motivation to get an MBA and how the journey had been so far. They candidly shared their motivations, which I have summarized in the end notes of this post. Observation in Socials Observation in Socials I kicked off socials at a barbecue party with about 50 MBA students moving in and out. As I moved around and got introduced to various groups, our conversations ranged from who made the best soup to Man Utd’s performance this season. It was not what I expected. To be honest, the conversations were a lot more down-to-earth than conversations I’ve had at tech/start-up social events. Nobody was trying to portray themselves as a ‘genius’ or the person who is changing the world. Even though a lot of the people there had done great things before they joined the program. The next stop was a party at one of the popular spots on campus. I was attending with a smaller group as most people were tired from the barbecue or had work to do. We had pre-drinks in one of the dorms and headed out for the night. This time, I spent more time observing the group than actually participating. I could see clearly that they had started building strong friendships that would last a lifetime. Bonding over shared suffering, happiness, and shared destinations. Perhaps this is why it could sometimes seem to outsiders like there’s some sort of ‘clique’. They were spending all this time together and were starting to build friendships that they would gladly defend if necessary. Conclusion/End Notes Conclusion/End Notes I woke up on the final day tired but filled with a feeling that I couldn’t quite explain. That was when I noticed it for the first time in 3 days. It was a sticker on the only window in the dormitory that I was in. “ Build something Massive. ” was boldly written across it. Build something Massive. It summarized my thoughts about the MBA program and the students I had met. The program was a chance for ambitious people to learn and to build their lives and careers into something massive, and most of them sacrificed a lot for this opportunity. Over 90% of them are immigrants. A lot of them are on scholarships that they had to fight for, while some took loans to attend the program. A lot of them are anxious and working hard to find internships/jobs months before they graduate because they feel the need to start providing value that can offset the cost of their program. Some of them left their families and cushy jobs/businesses to move into student accommodation for this program. Over 90% of them are immigrants. A lot of them are on scholarships that they had to fight for, while some took loans to attend the program. A lot of them are anxious and working hard to find internships/jobs months before they graduate because they feel the need to start providing value that can offset the cost of their program. Some of them left their families and cushy jobs/businesses to move into student accommodation for this program. I believe that the summary of these factors is why some of them seem to have a chip on their shoulders when they graduate. They believe that they have sacrificed a lot and deserve to be where they are. If it was you, you may feel the same way too. It’s humbling for me to write this as it exposes my past ignorance. But it’s important to put this out because there are many people who may be even more ignorant. So what did I really uncover about these MBAs? I learned that They’re parents. They’re lovers. They’re dreamers. They reflect everything that’s great about the world and everything that is wrong about it too. (Just like the rest of us.) The world has promised them in this unwritten contract that if they do [x ] (i.e., get a prestigious MBA), then doors will open for them. That their hard work and sacrifice would be rewarded. [x For a lot of them, the MBA is their path to a better life, recognition for their hard work, and a better future. During my time there, I felt very welcome in every circle, though I was meeting a lot of them for the first time. In fact, I may have annoyed some people anytime I was visibly shocked by something they did that went against the stereotype of an MBA. Would I get an MBA? I don’t know yet. But I have even more respect and appreciation for everyone who has gotten an MBA and everyone who plans to get one. You don’t know me, but I’m proud of you . You don’t know me, but I’m proud of you I want to thank Professor Owen for letting me sit in on the class. I enjoyed your approach to teaching the subject, and I appreciate how attentive you were to everyone. I learned a lot about negotiating and about people.

The Uncovered Secret Lives of MBAs From a Tech Person's Perspective

Before diving in, I want to clarify that I’m not a project or product manager. As a senior engineer, I do engage in planning and prioritization, but I’m not a SCRUM master or a certified PMP. My insights are drawn purely from my experiences as a developer, not formal training in project management. Over the past 15 years, I’ve worked with various teams and managers, but one consistent truth stands out: time is our most valuable asset. To maintain a proper work-life balance, managing and protecting that time effectively is essential. While project management has long sought to move away from time-based estimates, I believe that, as an engineer, these estimates are crucial for shielding me from the burden of long hours. The real issues we need to address are feature creep and poor estimation practices. Let’s explore why the industry is moving away from time-based estimates, and then I’ll explain why I think they can be effective when used correctly. Why Not Time-Based Estimates? Developers and project managers often have reservations about time-based estimates for several reasons: Uncertainty: Software development is unpredictable. Changing requirements, technical challenges, and team dynamics can all impact task durations.


Pressure: Time estimates can create pressure to meet deadlines, potentially leading to rushed work and compromised quality.


Miscommunication: Estimates may be misunderstood or misrepresented, resulting in unrealistic expectations from stakeholders.


Focus on Speed: Time estimates can shift the emphasis from quality and creativity to merely getting the job done.


Variability: Team members work at different paces and have varying levels of experience, complicating the creation of consistent estimates.


Retrospective Accountability: Consistently inaccurate estimates can lead to blame games, hindering learning and improvement.


Overemphasis on Metrics: A strong focus on time estimates can cultivate a culture more concerned with metrics than with outcomes and customer satisfaction. Pro-Time-Based Arguments Upon examining these concerns, I find flaws in many of them. It seems that the common objections to time-based estimates stem from poor estimation practices or poorly planned deadlines. Issues like uncertainty, pressure, and retrospective accountability often arise from inadequate processes. Miscommunication is a separate issue altogether, and variability and an overemphasis on metrics relate more to planning and culture than the estimation process itself. The real challenge appears to be that many development managers or teams lack the skills or maturity for effective time-based estimation, rather than time-based estimates being inherently flawed. Why Are Time-Based Estimates Important? Time is a precious resource. We have twenty-four hours in a day, typically allocating around eight for sleep and another eight for work. This leaves us with limited time for commuting, family, and personal interests. Many of us find that maintaining a work-life balance requires strict adherence to an eight-hour workday. However, employers often have their own agendas and expect a consistent flow of completed tasks that can often force employees beyond their expected eight hours. So, how do we demonstrate productivity while protecting our time? For me, time-based estimation is the answer. At Wayfair, we refined our sprint process into something remarkable. Every sprint was well-sized, our team was engaged and productive, and we rarely had to work extra hours to meet deadlines. Our project manager established an efficient and informed approach that allowed us to function smoothly, grounded in time-based estimation. Here’s how we did it: Technically, we used “Story Points” based on “days,” allowing for splitting into two-hour increments (0.25 pts). Days were assigned Fibonacci numbers (1, 2, 3, 5), and anything beyond five needed to be broken into smaller tickets. The estimation process included a weekly meeting (or biweekly if necessary) where an engineer familiar with a ticket presented it to the team. We discussed potential solutions, identified unknowns, and noted any external influences impacting our estimates. After the discussion, we would vote on how many “days” the ticket should take, considering developers’ varying levels of experience. If estimates differed, we opted for the largest or most common estimate based on the spread. If we couldn’t reach a consensus, we would discuss until the team reached a satisfactory conclusion. Crucially, everyone had a voice in sizing the tickets, fostering a collaborative atmosphere. The next step involved assessing developer capacity and ensuring our sprints aligned with that capacity. With our tickets focused on expected completion times, we established capacity as follows: Junior Dev: 7 days of work per sprint
Mid/Senior Dev: 8 days of work per sprint
Leads: 6 days of work per sprint These figures are adjusted for workload and vacations. Once we determined each developer's capacity, we held a grooming session to refine the upcoming sprint. By this point, all tickets were estimated, and developer capacity was clear. As an example, suppose we had 27 tickets totaling 56 points, and our team of six developers (two junior, three senior, one lead) had 47 points available. This discrepancy meant we needed to remove nine days of work. Our project manager would then prioritize which less critical tickets to drop, ensuring that developers championed any tickets relevant to their work to prevent important tasks from being sidelined. By the end of this session, we could confidently move forward with a manageable sprint. Conclusion Time is paramount. As a developer who values a healthy work-life balance, I aim to complete my tasks within a 40-hour week, and I need my manager to recognize my efforts. Moving away from time-based estimates can create a situation where developers are expected to work extra hours to complete their tasks, as there’s no reliable method to size the workload effectively. By committing to a robust process that amplifies every developer’s voice and incorporates time as a core element, we can ensure that everyone’s time is respected while still delivering results. This approach helps avoid issues like sprints bleeding into one another or failing to meet deadlines, ensuring work is completed on schedule. Of course, no system is perfect. There will inevitably be tickets that need to carry over, and adjustments may be necessary over time. Nevertheless, this approach provided me with the best working environment and estimation/planning process I have ever experienced. Before diving in, I want to clarify that I’m not a project or product manager. As a senior engineer, I do engage in planning and prioritization, but I’m not a SCRUM master or a certified PMP. My insights are drawn purely from my experiences as a developer, not formal training in project management. Over the past 15 years, I’ve worked with various teams and managers, but one consistent truth stands out: time is our most valuable asset. To maintain a proper work-life balance, managing and protecting that time effectively is essential. While project management has long sought to move away from time-based estimates, I believe that, as an engineer, these estimates are crucial for shielding me from the burden of long hours. The real issues we need to address are feature creep and poor estimation practices. Let’s explore why the industry is moving away from time-based estimates, and then I’ll explain why I think they can be effective when used correctly. Why Not Time-Based Estimates? Developers and project managers often have reservations about time-based estimates for several reasons: Uncertainty: Software development is unpredictable. Changing requirements, technical challenges, and team dynamics can all impact task durations. Pressure: Time estimates can create pressure to meet deadlines, potentially leading to rushed work and compromised quality. Miscommunication: Estimates may be misunderstood or misrepresented, resulting in unrealistic expectations from stakeholders. Focus on Speed: Time estimates can shift the emphasis from quality and creativity to merely getting the job done. Variability: Team members work at different paces and have varying levels of experience, complicating the creation of consistent estimates. Retrospective Accountability: Consistently inaccurate estimates can lead to blame games, hindering learning and improvement. Overemphasis on Metrics: A strong focus on time estimates can cultivate a culture more concerned with metrics than with outcomes and customer satisfaction. Uncertainty: Software development is unpredictable. Changing requirements, technical challenges, and team dynamics can all impact task durations. Uncertainty : Software development is unpredictable. Changing requirements, technical challenges, and team dynamics can all impact task durations. Uncertainty Pressure: Time estimates can create pressure to meet deadlines, potentially leading to rushed work and compromised quality. Pressure : Time estimates can create pressure to meet deadlines, potentially leading to rushed work and compromised quality. Pressure Miscommunication: Estimates may be misunderstood or misrepresented, resulting in unrealistic expectations from stakeholders. Miscommunication : Estimates may be misunderstood or misrepresented, resulting in unrealistic expectations from stakeholders. Miscommunication Focus on Speed: Time estimates can shift the emphasis from quality and creativity to merely getting the job done. Focus on Speed : Time estimates can shift the emphasis from quality and creativity to merely getting the job done. Focus on Speed Variability: Team members work at different paces and have varying levels of experience, complicating the creation of consistent estimates. Variability : Team members work at different paces and have varying levels of experience, complicating the creation of consistent estimates. Variability Retrospective Accountability: Consistently inaccurate estimates can lead to blame games, hindering learning and improvement. Retrospective Accountability : Consistently inaccurate estimates can lead to blame games, hindering learning and improvement. Retrospective Accountability Overemphasis on Metrics: A strong focus on time estimates can cultivate a culture more concerned with metrics than with outcomes and customer satisfaction. Overemphasis on Metrics : A strong focus on time estimates can cultivate a culture more concerned with metrics than with outcomes and customer satisfaction. Overemphasis on Metrics Pro-Time-Based Arguments Upon examining these concerns, I find flaws in many of them. It seems that the common objections to time-based estimates stem from poor estimation practices or poorly planned deadlines. Issues like uncertainty, pressure, and retrospective accountability often arise from inadequate processes. Miscommunication is a separate issue altogether, and variability and an overemphasis on metrics relate more to planning and culture than the estimation process itself. The real challenge appears to be that many development managers or teams lack the skills or maturity for effective time-based estimation, rather than time-based estimates being inherently flawed. Why Are Time-Based Estimates Important? Time is a precious resource. We have twenty-four hours in a day, typically allocating around eight for sleep and another eight for work. This leaves us with limited time for commuting, family, and personal interests. Many of us find that maintaining a work-life balance requires strict adherence to an eight-hour workday. However, employers often have their own agendas and expect a consistent flow of completed tasks that can often force employees beyond their expected eight hours. So, how do we demonstrate productivity while protecting our time? For me, time-based estimation is the answer. At Wayfair, we refined our sprint process into something remarkable. Every sprint was well-sized, our team was engaged and productive, and we rarely had to work extra hours to meet deadlines. Our project manager established an efficient and informed approach that allowed us to function smoothly, grounded in time-based estimation. Here’s how we did it: Technically, we used “Story Points” based on “days,” allowing for splitting into two-hour increments (0.25 pts). Days were assigned Fibonacci numbers (1, 2, 3, 5), and anything beyond five needed to be broken into smaller tickets. The estimation process included a weekly meeting (or biweekly if necessary) where an engineer familiar with a ticket presented it to the team. We discussed potential solutions, identified unknowns, and noted any external influences impacting our estimates. After the discussion, we would vote on how many “days” the ticket should take, considering developers’ varying levels of experience. If estimates differed, we opted for the largest or most common estimate based on the spread. If we couldn’t reach a consensus, we would discuss until the team reached a satisfactory conclusion. Crucially, everyone had a voice in sizing the tickets, fostering a collaborative atmosphere. The next step involved assessing developer capacity and ensuring our sprints aligned with that capacity. With our tickets focused on expected completion times, we established capacity as follows: Junior Dev: 7 days of work per sprint Mid/Senior Dev: 8 days of work per sprint Leads: 6 days of work per sprint Junior Dev: 7 days of work per sprint Mid/Senior Dev: 8 days of work per sprint Leads: 6 days of work per sprint These figures are adjusted for workload and vacations. Once we determined each developer's capacity, we held a grooming session to refine the upcoming sprint. By this point, all tickets were estimated, and developer capacity was clear. As an example, suppose we had 27 tickets totaling 56 points, and our team of six developers (two junior, three senior, one lead) had 47 points available. This discrepancy meant we needed to remove nine days of work. Our project manager would then prioritize which less critical tickets to drop, ensuring that developers championed any tickets relevant to their work to prevent important tasks from being sidelined. By the end of this session, we could confidently move forward with a manageable sprint. Conclusion Time is paramount. As a developer who values a healthy work-life balance, I aim to complete my tasks within a 40-hour week, and I need my manager to recognize my efforts. Moving away from time-based estimates can create a situation where developers are expected to work extra hours to complete their tasks, as there’s no reliable method to size the workload effectively. By committing to a robust process that amplifies every developer’s voice and incorporates time as a core element, we can ensure that everyone’s time is respected while still delivering results. This approach helps avoid issues like sprints bleeding into one another or failing to meet deadlines, ensuring work is completed on schedule. Of course, no system is perfect. There will inevitably be tickets that need to carry over, and adjustments may be necessary over time. Nevertheless, this approach provided me with the best working environment and estimation/planning process I have ever experienced.

A Developer's Perspective on Time-Based Estimates

Machine learning experiments require extensive parametrization, including optimizer parameters, network architecture, and data augmentation. However, we strive for concise, readable code instead of a cumbersome 200 lines dedicated to argparse. Our goal is to focus on programming logic rather than threading new parameters through function signatures. Additionally, we seek a structure that is easily expandable without burdening the project, ensuring the reproducibility of experiments. Hydra offers a solution to these challenges. Below, you will find a basic guide on how to use it. What is Hydra? Hydra is a library with rich capabilities for managing configurations. The main site describes the name like this: “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line.The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” But I have my own interpretation of its name. It is just a combination of some tools that work together impressively well! Hydra offers a seamless solution to the common headaches faced by ML engineers when attempting to replicate experiments. It elegantly replaces the need for argparse or YAML configurations, allowing access to parameters both from the command line and YAML files. Consider the pain points: Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Hydra addresses these issues by enabling dynamic configuration adjustments without the need for multiple bulky files or rigid command-line arguments. Furthermore, it simplifies the process of passing complex configurations, such as model architectures or functions, directly from the config file to the model. This capability eliminates the tedious task of manually feeding parameters into the model, streamlining the workflow and reducing the margin for error. Basic Setup Let's imagine the simplest setup: multiclass classification on MNIST using an MLP. We have a configuration and a training script. .
├── configs
│   └── config.yaml
└── main.py The main script can look like this. import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() The configuration has the following structure. It is the default, and parameters from it will be used by default if not specified otherwise. # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 To start training you can use: python main.py And for changing parameters, you can do it not only from YAML but also from cli: python main.py lr=0.03 If you use it, parameters from cli override the first parameters. As you can see, there is no argparse or additional middleware. It's easy to change parameters from the command line. Pass Class Object With Hydra We see that in class Net, the architecture can be highly customizable. Even more, we can use also another net for this, e.g. CNN. But we do not want to change our pipeline. Hydra can construct almost any Python object with specified parameters. Let’s describe our net in another way, using YAML. model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 and for CNN _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST Pass Function Through Config Sometimes, our parameters can also function. For example, How to Access Config Without Decorating Main - e.g., in ipython Notebook. So, you can use Compose API. from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) Multi Runs For example, you want to find some good hyperparam and Hydra allows a lot for that. Basically, you can launch several experiments sequentially like. python train_model.py -m "batch_size=16,32,64" It launches sequentially, but it is easy to extend to parallel if turned on some launcher like joblib. python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" Run All Experiments From the Folder for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done Resources hydra - https://hydra.cc/
good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main Machine learning experiments require extensive parametrization, including optimizer parameters, network architecture, and data augmentation. However, we strive for concise, readable code instead of a cumbersome 200 lines dedicated to argparse. Our goal is to focus on programming logic rather than threading new parameters through function signatures. Additionally, we seek a structure that is easily expandable without burdening the project, ensuring the reproducibility of experiments. Hydra offers a solution to these challenges. Below, you will find a basic guide on how to use it. What is Hydra? Hydra is a library with rich capabilities for managing configurations. The main site describes the name like this: “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line.The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line. The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” But I have my own interpretation of its name. It is just a combination of some tools that work together impressively well! Hydra offers a seamless solution to the common headaches faced by ML engineers when attempting to replicate experiments. It elegantly replaces the need for argparse or YAML configurations, allowing access to parameters both from the command line and YAML files. Consider the pain points: Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Hydra addresses these issues by enabling dynamic configuration adjustments without the need for multiple bulky files or rigid command-line arguments. Furthermore, it simplifies the process of passing complex configurations, such as model architectures or functions, directly from the config file to the model. This capability eliminates the tedious task of manually feeding parameters into the model, streamlining the workflow and reducing the margin for error. Basic Setup Let's imagine the simplest setup: multiclass classification on MNIST using an MLP. We have a configuration and a training script. .
├── configs
│   └── config.yaml
└── main.py .
├── configs
│   └── config.yaml
└── main.py The main script can look like this. import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() The configuration has the following structure. It is the default, and parameters from it will be used by default if not specified otherwise. # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 To start training you can use: python main.py python main.py And for changing parameters, you can do it not only from YAML but also from cli: python main.py lr=0.03 python main.py lr=0.03 If you use it, parameters from cli override the first parameters. As you can see, there is no argparse or additional middleware. It's easy to change parameters from the command line. Pass Class Object With Hydra We see that in class Net, the architecture can be highly customizable. Even more, we can use also another net for this, e.g. CNN. But we do not want to change our pipeline. Hydra can construct almost any Python object with specified parameters. Let’s describe our net in another way, using YAML. model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 and for CNN _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST Pass Function Through Config Sometimes, our parameters can also function. For example, How to Access Config Without Decorating Main - e.g., in ipython Notebook. So, you can use Compose API . Compose API from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) Multi Runs For example, you want to find some good hyperparam and Hydra allows a lot for that. Basically, you can launch several experiments sequentially like. python train_model.py -m "batch_size=16,32,64" python train_model.py -m "batch_size=16,32,64" It launches sequentially, but it is easy to extend to parallel if turned on some launcher like joblib. python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" Run All Experiments From the Folder for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done Resources hydra - https://hydra.cc/ good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main hydra - https://hydra.cc/ https://hydra.cc/ good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main https://github.com/ashleve/lightning-hydra-template/tree/main

How to Configure Experiments With Hydra - From an ML Enginner Perspective

Have you ever wandered into the realm of GitHub API authentication? In case you did, you have arrived in the right spot, because I have an experience that I want to share with you. When trying to authenticate to the GitHub REST API you need to use either a personal access token , the workflow GITHUB_TOKEN variable or a GitHub installation access token. The first option is not really an option for automation purpose, because a personal access token as the name suggests it is usually attributed to an user and expires after a given time frame. The GITHUB_TOKEN variable can be an interesting option for a stand-alone repo, but it is scoped to the respective repository. Hence the correct option from a GitHub organization perspective is the use of GitHub installation access token. The process of generating an installation access token requires several inputs that we obtain when we register a new GitHub application in an organization: Enter name, description, provide an URL (we’ll only use the app for authentication purposes, so you can use a dummy one) and uncheck the “Active” tick for the webhook. The other optional fields are to be left blank.
While on the same page you need to configure the permissions required by the app.
Once the registration is done, you will receive additional information on the next screen, such as app id.
While on the same screen, press “Generate a new client secret” button.
Import the App's private key in your AWS Account KMS service, under customer-managed keys of type asymmetric, sign-verify, RSA 2048, following steps 1 to 4 from the link. The private key is the client secret from previous step. The previous step follows the security best practices of storing the app private key. Storing the private key in a GitHub secret is not as secure, because if an attacker gains access to the GitHub runner environment, the attacker will consequently have access to the GitHub secret. The solution proposed is similar to Azure Key Vault, but uses AWS KMS instead. Once imported in KMS, the private key can no longer be retrieved. The authorized users can ask the AWS KMS API to sign/verify messages on their behalf, but will not perform such actions directly on the messages using the key. This way, sign/verify actions, key storage as well as user authentication and authorization are delegated to AWS. Please, make sure your AWS account access is properly secured! As you might have realized already, AWS authentication becomes the weakest link in the chain in this scenario. While AWS takes care of the security of the cloud, you are responsible for the security in the cloud. This is a broader topic which is outside the scope of this article, but as a guideline, It is highly advised to use temporary AWS credentials scoped to the least privilege when accessing AWS API in order to sign the JWT token. More on how to set up a GitHub as an OIDC provider for AWS here. Using AWS KMS will generate additional charges in your AWS bill. In the following steps, we’ll prepare the repository to be able to get the GitHub app installation access token: Store the App's ID generated previously in your repository environment variable or secret (example: APP_ID).
Store the previously generated KMS Key ID as a repository secret (example KMS_KEY_ID).
Store the AWS role to be assumed by the action as a repository secret (example ROLE_TO_ASSUME). The setup assumes you are assuming an AWS role, hence using temporary AWS credentials.
Store the AWS session name as an environment variable (example ROLE_SESSION_NAME). This is needed in case you are assuming and AWS role using OIDC.
Store the AWS region name as an environment variable (example AWS_REGION). Once you’re set up with all the above, you have all the prerequisite necessary in order to use this GitHub action for generating the GitHub installation access token for your GitHub app. The action which I have developed unlike the vanilla one uses AWS KMS in order to sign/verify the JWT token required for generating the GitHub installation access token. Terraform - A Case Study Recently I have started to investigate ways to authenticate terraform provider to the GitHub API in order to be able to use it in deploying GitHub Infrastructure (create repos, etc) inside a GitHub organization. The terraform documentation gives 2 options,  GitHub app and OAuth/personal access token. I have taken the inputs required by the GitHub app authentication (i.e. app id and app private key) and developed a GitHub action capable of using AWS KMS for storing the private key and generating the installation access token. Once the token is generated, you can override the GITHUB_TOKEN workflow environment variable with the token produced by the action. Here is a terraform example on how to use the action as well as the respective workflow file. Enjoy! Have you ever wandered into the realm of GitHub API authentication? In case you did, you have arrived in the right spot, because I have an experience that I want to share with you. When trying to authenticate to the GitHub REST API you need to use either a personal access token , the workflow GITHUB_TOKEN variable or a GitHub installation access token. The first option is not really an option for automation purpose, because a personal access token as the name suggests it is usually attributed to an user and expires after a given time frame. The GITHUB_TOKEN variable can be an interesting option for a stand-alone repo, but it is scoped to the respective repository. Hence the correct option from a GitHub organization perspective is the use of GitHub installation access token . authenticate to the GitHub REST API GITHUB_TOKEN GitHub installation access token The process of generating an installation access token requires several inputs that we obtain when we register a new GitHub application in an organization: generating an installation access token new GitHub application Enter name, description, provide an URL (we’ll only use the app for authentication purposes, so you can use a dummy one) and uncheck the “Active” tick for the webhook. The other optional fields are to be left blank. While on the same page you need to configure the permissions required by the app. Once the registration is done, you will receive additional information on the next screen, such as app id. While on the same screen, press “Generate a new client secret” button. Import the App's private key in your AWS Account KMS service, under customer-managed keys of type asymmetric, sign-verify, RSA 2048, following steps 1 to 4 from the link. The private key is the client secret from previous step. Enter name, description, provide an URL (we’ll only use the app for authentication purposes, so you can use a dummy one) and uncheck the “Active” tick for the webhook. The other optional fields are to be left blank. While on the same page you need to configure the permissions required by the app. Once the registration is done, you will receive additional information on the next screen, such as app id. While on the same screen, press “Generate a new client secret” button. Import the App's private key in your AWS Account KMS service, under customer-managed keys of type asymmetric, sign-verify, RSA 2048 , following steps 1 to 4 from the link. The private key is the client secret from previous step. Import the App's private key in your AWS Account KMS service, under customer-managed keys of type asymmetric, sign-verify, RSA 2048 Import the App's private key in your AWS Account KMS service, under customer-managed keys of type asymmetric, sign-verify, RSA 2048 The previous step follows the security best practices of storing the app private key. Storing the private key in a GitHub secret is not as secure, because if an attacker gains access to the GitHub runner environment, the attacker will consequently have access to the GitHub secret. The solution proposed is similar to Azure Key Vault, but uses AWS KMS instead. Once imported in KMS, the private key can no longer be retrieved. The authorized users can ask the AWS KMS API to sign/verify messages on their behalf, but will not perform such actions directly on the messages using the key. This way, sign/verify actions, key storage as well as user authentication and authorization are delegated to AWS. the security best practices Please, make sure your AWS account access is properly secured! Please, make sure your AWS account access is properly secured! As you might have realized already, AWS authentication becomes the weakest link in the chain in this scenario. While AWS takes care of the security of the cloud, you are responsible for the security in the cloud . This is a broader topic which is outside the scope of this article, but as a guideline, It is highly advised to use temporary AWS credentials scoped to the least privilege when accessing AWS API in order to sign the JWT token. More on how to set up a GitHub as an OIDC provider for AWS here . While AWS takes care of the security of the cloud, you are responsible for the security in the cloud here Using AWS KMS will generate additional charges in your AWS bill. Using AWS KMS will generate additional charges in your AWS bill. In the following steps, we’ll prepare the repository to be able to get the GitHub app installation access token: Store the App's ID generated previously in your repository environment variable or secret (example: APP_ID). Store the previously generated KMS Key ID as a repository secret (example KMS_KEY_ID). Store the AWS role to be assumed by the action as a repository secret (example ROLE_TO_ASSUME). The setup assumes you are assuming an AWS role, hence using temporary AWS credentials. Store the AWS session name as an environment variable (example ROLE_SESSION_NAME). This is needed in case you are assuming and AWS role using OIDC. Store the AWS region name as an environment variable (example AWS_REGION). Store the App's ID generated previously in your repository environment variable or secret (example: APP_ID ). Store the App's ID generated previously in your repository environment variable Store the App's ID generated previously in your repository environment variable secret secret APP_ID Store the previously generated KMS Key ID as a repository secret (example KMS_KEY_ID ). Store the previously generated KMS Key ID as a repository secret Store the previously generated KMS Key ID as a repository secret KMS_KEY_ID Store the AWS role to be assumed by the action as a repository secret (example ROLE_TO_ASSUME ). The setup assumes you are assuming an AWS role, hence using temporary AWS credentials . Store the AWS role to be assumed by the action as a repository secret Store the AWS role to be assumed by the action as a repository secret ROLE_TO_ASSUME temporary AWS credentials Store the AWS session name as an environment variable (example ROLE_SESSION_NAME ). This is needed in case you are assuming and AWS role using OIDC. Store the AWS session name as an environment variable Store the AWS session name as an environment variable ROLE_SESSION_NAME Store the AWS region name as an environment variable (example AWS_REGION ). Store the AWS region name as an environment variable Store the AWS region name as an environment variable AWS_REGION Once you’re set up with all the above, you have all the prerequisite necessary in order to use this GitHub action for generating the GitHub installation access token for your GitHub app. The action which I have developed unlike the vanilla one uses AWS KMS in order to sign/verify the JWT token required for generating the GitHub installation access token. this Terraform - A Case Study Recently I have started to investigate ways to authenticate terraform provider to the GitHub API in order to be able to use it in deploying GitHub Infrastructure (create repos, etc) inside a GitHub organization. The terraform documentation gives 2 options,  GitHub app and OAuth/personal access token. authenticate terraform provider I have taken the inputs required by the GitHub app authentication (i.e. app id and app private key) and developed a GitHub action capable of using AWS KMS for storing the private key and generating the installation access token. GitHub action Once the token is generated, you can override the GITHUB_TOKEN workflow environment variable with the token produced by the action. Here is a terraform example on how to use the action as well as the respective workflow file . Here workflow file Enjoy!

Access GitHub API Securely Using GitHub Actions from an Organization Perspective - A Case Study

The Facial Recognition Industry: A Developer's Perspective

AI Consciousness is Inevitable: A Theoretical Computer Science Perspective: References

Addressing Kevin Mitchell’s Questions From the Perspective of Conscious Turing Machine Robots

Authors:
(1) Lenore Blum (lblum@cs.cmu.edu);
(2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References ABSTRACT We look at consciousness through the lens of Theoretical Computer Science, a branch of mathematics that studies computation under resource limitations. From this perspective, we develop a formal machine model for consciousness. The model is inspired by Alan Turing’s simple yet powerful model of computation and Bernard Baars’ theater model of consciousness. Though extremely simple, the model aligns at a high level with many of the major scientific theories of human and animal consciousness, supporting our claim that machine consciousness is inevitable. 1 Introduction We study consciousness from the perspective of Theoretical Computer Science (TCS), a branch of mathematics concerned with understanding the underlying principles of computation and complexity, including the implications and surprising consequences of resource limitations. By taking resource limitations into account, the TCS perspective is distinguished from the earlier Turing Theory of Computation (TOC) where limitations of time and space did not figure. TOC distinguishes computable from not computable. It does not distinguish between computable and not efficiently computable. [1] We highlight the importance of this separation for tackling consciousness and related topics such as the paradox of free will. Elsewhere (Blum & Blum, 2021), we describe the Conscious Turing Machine (CTM), a simple formal machine model of consciousness inspired in part by Alan Turing’s simple formal machine model of computation (Turing, 1937), and by Bernard Baars’ theater model of consciousness (Baars, Bernard J., 1997). In (Blum & Blum, 2022), we consider how a CTM could exhibit various phenomena associated with consciousness (e.g., blindsight, inattentive blindness, change blindness) and present CTM explanations that agree, at a high level, with cognitive neuroscience literature. In contrast to Turing, we take resource limitations into account, both in designing the CTM model and in how resource limitations affect (and help explain) feelings of consciousness. Our perspective differs even more. What gives the CTM its feeling of consciousness is not its input output map, nor its computing power, but what’s under the hood.[2] In this chapter we take a brief look under the hood. In addition, we show how the CTM naturally aligns with and integrates features considered key to human and animal consciousness by many of the major scientific theories of consciousness.[3] These theories consider different aspects of consciousness and often compete with each other (Lenharo, 2024). Yet their alignment with the CTM at a high level helps demonstrate their compatibility and/or complementarity. But, even more, their alignment with the CTM, a simple machine model that exhibits phenomena associated with consciousness, supports our claim that a conscious AI is inevitable. David Chalmers’ introduction of the Hard Problem (Chalmers, 1995) helped classify most notions of consciousness into one of two types. The first type, variously called access consciousness (Block, 1995) or functional (computational) or cognitive consciousness, we call conscious attention. The second type (associated with the Hard Problem) is called subjective or phenomenological consciousness and is generally associated with feelings or qualia. We call it conscious awareness. Chalmers’ Hard Problem can be viewed as a challenge to show that subjective consciousness is “functional”. We contend that consciousness writ large requires both conscious attention and conscious awareness, each informing the other to various degrees. We contend that a machine that interacts with its worlds (inner and outer) via input sensors and output actuators, that constructs models of these worlds enabling planning, prediction, testing, and learning from feedback, and that develops a rich internal multimodal language, can have both types of consciousness. In particular, we contend that subjective consciousness is computational and functional. We emphasize that the CTM is a formal machine model designed to explore and understand consciousness from a TCS perspective. It is not intended to model the brain nor the neural correlates of consciousness. Nevertheless, the CTM is inspired by cognitive and neuroscience theories of consciousness. Specifically, as we have mentioned, the CTM is inspired by cognitive neuroscientist Bernard Baars’ theater model of consciousness (Baars, Bernard J., 1997), the global workspace (GW) theory of consciousness. However, here again, the CTM is not a standard GW model. The CTM differs from GW in a number of important ways: its competition for global broadcast is formally defined and completely replaces the ill-defined Central Executive of other GW models; its special processors including especially its Model-of-the-World processor construct and employ models of its (inner and outer) worlds; its rich multimodal internal language, Brainish, for creating labeled sketches in its world models and for communicating between processors; and its predictive dynamics (cycles of prediction, testing, feedback and learning, locally and globally). The CTM also interacts with its outer world via input sensors and output actuators. To emphasize CTM’s embodied, embedded, enacted and extended mind, we call it here the CTM Robot (CtmR). While working on this chapter, we became aware of Kevin Mitchell’s blog post in Wiring the Brain (Mitchell, 2023) in which he makes a point similar to one that we make, namely, that many of the major theories of consciousness are compatible and/or complementary. For a similar conclusion, see (Storm & et.al., 2024). Even more, Mitchell presents “a non-exhaustive list of questions … that a theory of consciousness should be able to encompass”. He declares that “even if such a theory can’t currently answer all those questions, it should at least provide an overarching framework[4] (i.e., what a theory really should be), in which they can be asked in a coherent way, without one question destabilizing what we think we know about the answer to another one.” Mitchell’s questions are thoughtful, interesting, and important. At the end of this chapter, we offer preliminary answers from the perspective of the Conscious Turing Machine Robot (CtmR). Our answers both supplement and highlight material in the brief Overview of CtmR that we now present.[5] This paper is available on arxiv under CC BY 4.0 DEED license. [1] For a brief history of TOC and TCS see Appendix 7.1. [2] This is important. We claim that simulations that modify CTM’s key internal structures and processes will not necessarily experience what CTM does. We are not claiming that the CTM is the only possible machine model to experience feelings of consciousness. [3] These theories include: The Global Workspace/Global Neuronal Workspace (GW/GNW), Attention Schema Theory (AST), Predictive Processing (PP), Integrated Information Theory (IIT), Embodied, Embedded, Enacted and Extended (EEEE) theories, Evolutionary theories, and the Extended Reticulothalamic Activating System + Free Energy Principle Theory (ERTAS + FEP). [4] Italics ours. [5] In the Overview, we annotate paragraphs that refer to Kevin Mitchell’s queries. As an example, if a paragraph has a label [KM1], then it refers to Mitchell’s first query, KM1. Conversely, if Mitchell’s query is labeled with an asterisk such as KM1*, then it refers to [KM1] in the Overview. Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Authors: Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 5 Summary and Conclusions 6 Acknowledgements 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References References ABSTRACT We look at consciousness through the lens of Theoretical Computer Science, a branch of mathematics that studies computation under resource limitations. From this perspective, we develop a formal machine model for consciousness. The model is inspired by Alan Turing’s simple yet powerful model of computation and Bernard Baars’ theater model of consciousness. Though extremely simple, the model aligns at a high level with many of the major scientific theories of human and animal consciousness, supporting our claim that machine consciousness is inevitable. 1 Introduction We study consciousness from the perspective of Theoretical Computer Science (TCS), a branch of mathematics concerned with understanding the underlying principles of computation and complexity, including the implications and surprising consequences of resource limitations. By taking resource limitations into account, the TCS perspective is distinguished from the earlier Turing Theory of Computation (TOC) where limitations of time and space did not figure. TOC distinguishes computable from not computable. It does not distinguish between computable and not efficiently computable. [1] We highlight the importance of this separation for tackling consciousness and related topics such as the paradox of free will. Elsewhere (Blum & Blum, 2021), we describe the Conscious Turing Machine (CTM), a simple formal machine model of consciousness inspired in part by Alan Turing’s simple formal machine model of computation (Turing, 1937), and by Bernard Baars’ theater model of consciousness (Baars, Bernard J., 1997). In (Blum & Blum, 2022), we consider how a CTM could exhibit various phenomena associated with consciousness (e.g., blindsight, inattentive blindness, change blindness) and present CTM explanations that agree, at a high level, with cognitive neuroscience literature. In contrast to Turing, we take resource limitations into account, both in designing the CTM model and in how resource limitations affect (and help explain) feelings of consciousness. Our perspective differs even more. What gives the CTM its feeling of consciousness is not its input output map, nor its computing power, but what’s under the hood.[2] In this chapter we take a brief look under the hood. In addition, we show how the CTM naturally aligns with and integrates features considered key to human and animal consciousness by many of the major scientific theories of consciousness.[3] These theories consider different aspects of consciousness and often compete with each other (Lenharo, 2024). Yet their alignment with the CTM at a high level helps demonstrate their compatibility and/or complementarity. But, even more, their alignment with the CTM, a simple machine model that exhibits phenomena associated with consciousness, supports our claim that a conscious AI is inevitable. our claim that a conscious AI is inevitable. David Chalmers’ introduction of the Hard Problem (Chalmers, 1995) helped classify most notions of consciousness into one of two types. The first type, variously called access consciousness (Block, 1995) or functional (computational) or cognitive consciousness, we call conscious attention. The second type (associated with the Hard Problem) is called subjective or phenomenological consciousness and is generally associated with feelings or qualia. We call it conscious awareness. Chalmers’ Hard Problem can be viewed as a challenge to show that subjective consciousness is “functional”. We contend that consciousness writ large requires both conscious attention and conscious awareness, each informing the other to various degrees. We contend that a machine that interacts with its worlds (inner and outer) via input sensors and output actuators, that constructs models of these worlds enabling planning, prediction, testing, and learning from feedback, and that develops a rich internal multimodal language, can have both types of consciousness. In particular, we contend that subjective consciousness is computational and functional. We emphasize that the CTM is a formal machine model designed to explore and understand consciousness from a TCS perspective. It is not intended to model the brain nor the neural correlates of consciousness. Nevertheless, the CTM is inspired by cognitive and neuroscience theories of consciousness. Specifically, as we have mentioned, the CTM is inspired by cognitive neuroscientist Bernard Baars’ theater model of consciousness (Baars, Bernard J., 1997), the global workspace (GW) theory of consciousness. However, here again, the CTM is not a standard GW model. The CTM differs from GW in a number of important ways: its competition for global broadcast is formally defined and completely replaces the ill-defined Central Executive of other GW models; its special processors including especially its Model-of-the-World processor construct and employ models of its (inner and outer) worlds; its rich multimodal internal language, Brainish, for creating labeled sketches in its world models and for communicating between processors; and its predictive dynamics (cycles of prediction, testing, feedback and learning, locally and globally). The CTM also interacts with its outer world via input sensors and output actuators. To emphasize CTM’s embodied, embedded, enacted and extended mind, we call it here the CTM Robot ( CtmR ). CtmR While working on this chapter, we became aware of Kevin Mitchell’s blog post in Wiring the Brain (Mitchell, 2023) in which he makes a point similar to one that we make, namely, that many of the major theories of consciousness are compatible and/or complementary. For a similar conclusion, see (Storm & et.al., 2024). Even more, Mitchell presents “a non-exhaustive list of questions … that a theory of consciousness should be able to encompass”. He declares that “even if such a theory can’t currently answer all those questions, it should at least provide an overarching framework[4] (i.e., what a theory really should be), in which they can be asked in a coherent way, without one question destabilizing what we think we know about the answer to another one.” Mitchell’s questions are thoughtful, interesting, and important. At the end of this chapter, we offer preliminary answers from the perspective of the Conscious Turing Machine Robot (CtmR). Our answers both supplement and highlight material in the brief Overview of CtmR that we now present.[5] This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [1] For a brief history of TOC and TCS see Appendix 7.1. [2] This is important. We claim that simulations that modify CTM’s key internal structures and processes will not necessarily experience what CTM does. We are not claiming that the CTM is the only possible machine model to experience feelings of consciousness. [3] These theories include: The Global Workspace/Global Neuronal Workspace (GW/GNW), Attention Schema Theory (AST), Predictive Processing (PP), Integrated Information Theory (IIT), Embodied, Embedded, Enacted and Extended (EEEE) theories, Evolutionary theories, and the Extended Reticulothalamic Activating System + Free Energy Principle Theory (ERTAS + FEP). [4] Italics ours. [5] In the Overview, we annotate paragraphs that refer to Kevin Mitchell’s queries. As an example, if a paragraph has a label [KM1], then it refers to Mitchell’s first query, KM1. Conversely, if Mitchell’s query is labeled with an asterisk such as KM1*, then it refers to [KM1] in the Overview.

AI Consciousness is Inevitable: A Theoretical Computer Science Perspective

Vědomí umělé inteligence je nevyhnutelné: perspektiva teoretické informatiky

KI-Bewusstsein ist unvermeidlich: Eine theoretische Informatikperspektive

La conciencia de la IA es inevitable: una perspectiva teórica desde la informática

A conciencia da IA é inevitable: unha perspectiva teórica da informática

La coscienza dell'intelligenza artificiale è inevitabile: una prospettiva teorica dell'informatica

AI の意識は不可避: 理論的なコンピュータ サイエンスの観点

AI ცნობიერება გარდაუვალია: კომპიუტერული მეცნიერების თეორიული პერსპექტივა

Kesedaran AI Tidak Dapat Dielakkan: Perspektif Sains Komputer Teoritikal

A Consciência da IA é Inevitável: Uma Perspectiva Teórica da Ciência da Computação

Ndërgjegjja e AI është e pashmangshme: Një perspektivë teorike e shkencës kompjuterike

AI-medvetande är oundvikligt: ett teoretiskt datavetenskapligt perspektiv

Свідомість штучного інтелекту неминуча: теоретична перспектива інформатики

The above image is a beautiful image of a butterfly meshed in vibrant colors with musical notes all across it. With so much going on in it, what one person sees in it (at least at first) can be very different from another person. A little while ago, I was working on a neural style transfer project. This is one of the images that I used as the style image. However, the resultant image didn’t display as much difference from the original base image as I’d thought. Even so, I gaily showed it to a couple of friends of mine. I asked what they thought of it, and they said that the base image and the butterfly image both had a significant chunk of white in them which was why the color transfer wasn’t obvious. I honestly stared at the image for a couple of minutes before I realized that what they had referred to as the ‘butterfly image’ was in fact my style image. See, the thing that had initially attracted me to the picture was the color explosion in combination with the musical notes and the writings. It was all so electric. But this wasn’t what my friends saw at first glance, and like I said, I didn’t see the butterfly until they pointed it to me even though, from a logical perspective it really should have been. The point of this entire rambling is that people have different perspectives, and when multiple people are working on a project, in order to make it a success, each individual’s perspective and way of thinking should be clear to the rest of the team. This is the definition of coordination in a functional system. Achieving this level of coordination is hard enough face-to-face, but it doubles down on the internet. That, to me, seems to be one of the issues plaguing the decentralization community. What Is a Blockchain Network? I will give a brief overview of what decentralization and blockchain networks are so that everyone can understand my point more clearly. The blockchain is like a ledger that everyone can read - at least if it’s public. If it’s private or semi-public, then you have to be one of the authorized people to be able to view and change anything in the ledger. But even for public blockchains, a transaction performed on it needs to be verified. The more times it is verified, the more its authenticity is assured. The trouble is these verifications require solving complex mathematical problems which, in turn, require vast amounts of computational resources. Moreover, people working on these are individuals who do not, in real life, know each other. They are only connected through their mutual interest in decentralization. Therefore, in order to ascertain cohesive coordination among these individuals to assure the formation of a well-coordinated decentralized system, protocols are established by the network administrators to clarify the vision and mission of the network. Now understandably, most of these networks’ missions and protocols revolve around data privacy and security because from what I’ve been able to understand--courtesy of Google’s Gemini--is that it is completely possible for a group of hackers or cybercriminals to create a private blockchain network of their own where they freely store and share all kinds of hacked data. So clearly, the blockchain still needs some work to rid it of criminal activity. Now, I understand that my example may be a little extreme and that law-enforcement agencies probably have similar sophisticated technologies to counteract these crimes, and things like digital fingerprints exist, as judiciously explained by Gemini. Nevertheless, these are problems that the blockchain community is rightfully interested in resolving. But that leads to another problem: tunnel vision. The Big Picture The blockchain community is a very exclusive community, composed mostly of young, enthusiastic, tech-savvy individuals. However, the idea of data privacy and security is for everyone, not just the techies. So, it is important that this technology is used and understood as much by regular individuals as it is by its developers. Imagine this scenario. A bank manager or administrator of a small, local bank decides that he wants to convert his banking system to a decentralized one to better the security for the bank customers. He hires a young developer to help and guide him through this change. The developer talks to him about authentication codes, verification, mining data, and other resources needed to form an effective blockchain network. The manager--even if he is relatively young--will probably understand that the system access should be limited to a few trusted individuals, and maybe a little bit more. However, unless he himself is a computer scientist--unlikely in the case of a bank manager--it’s doubtful that he will understand the whole concept of decentralization. But the tipping point would be the computational costs required. Remember, the bank is a small, local bank with a limited budget and resources. When you add this to the fact that the manager may or may not have completely understood how the system works, the odds that the bank will end up installing an effective decentralized system are slim. The Gaps One of the main issues stopping decentralization from becoming a global phenomenon is the lack of suitable hardware, closely followed by inadequate education for the masses. My neural style transfer project was actually not the project I’d originally meant to do. Instead, I had intended to create a Generative AI project using transfer learning to make a GAN algorithm create some new images. My dataset--which was composed of the works of lesser-known artists--wasn’t a very big one, so I knew that it was possible that the end result may not be as novel as it would have been with a bigger dataset. But what I hadn’t realized was that even with that small dataset the time, space, and computation required would be so enormous that, for my purposes, it would be better if I switched to a slightly different type of algorithm. That’s kind of my point. I use a regular laptop for most of my work, and for the most part, even for other A.I. projects, it’s usually enough. But it seems that for projects that require any kind of ‘generation’--even to a small extent--this can be a very big problem. In my case, it wasn’t that big of a deal since it was more of a personal project than anything else, but for the task of providing privacy and security to the clients of a bank, this may be a bigger issue because while their computational resources may be better than my tiny, lonely laptop, it’s still unlikely to have the computational power even remotely resembling that which is necessary for maintaining blockchain networks. Furthermore, the blockchain community really needs to understand that most people in positions responsible for the privacy and security of organizations are at the very least in their forties. Even if they are younger, they’re probably not very aware of how decentralization systems work which is understandable given its complexity. One of the reasons that social media platforms are so universally popular across all demographics is that aside from knowing what a few essential buttons do, people don’t need to know very much about it. This is especially so in cases of individuals over the age of 60, and while your average bank manager may not be 60, he/she isn’t going to be 20 or even 30 either. Conclusion There is a lot of money being spent on making decentralized systems more reliable and eco-friendly, but maybe what is necessary is some of that money to also be spent on creating hardware that is not so costly and on educating individuals, i.e., marketing according to their demographic and positions in life. I mean, I can use my social media account from any phone and the phone needn’t be a good quality one, or even a new one. As long as the wifi and my app work, I don’t need anything else. All in all, ensuring online coordination requires in and of itself an extensive system. If you’re only looking at the shape of the butterfly then you’re missing the colors and beauty associated with it, and if you only look at the colors, you might end up missing the main thing. So, maybe what the blockchain and decentralization community needs right now for optimal coordination and cooperation are not just networks that focus on increasing privacy and security but instead a system of network of networks, where one network may be responsible for ensuring privacy and security, the other for hardware, and still other which works on marketing and educating plans for the masses all connected towards one unified goal. In this way, privacy and security will no longer be just a catchphrase for the tech-savvy but an asset utilized by everyone. The above image is a beautiful image of a butterfly meshed in vibrant colors with musical notes all across it. With so much going on in it, what one person sees in it (at least at first) can be very different from another person. A little while ago, I was working on a neural style transfer project. This is one of the images that I used as the style image. However, the resultant image didn’t display as much difference from the original base image as I’d thought. Even so, I gaily showed it to a couple of friends of mine. I asked what they thought of it, and they said that the base image and the butterfly image both had a significant chunk of white in them which was why the color transfer wasn’t obvious. I honestly stared at the image for a couple of minutes before I realized that what they had referred to as the ‘butterfly image’ was in fact my style image. See, the thing that had initially attracted me to the picture was the color explosion in combination with the musical notes and the writings. It was all so electric. But this wasn’t what my friends saw at first glance, and like I said, I didn’t see the butterfly until they pointed it to me even though, from a logical perspective it really should have been. The point of this entire rambling is that people have different perspectives, and when multiple people are working on a project, in order to make it a success, each individual’s perspective and way of thinking should be clear to the rest of the team. This is the definition of coordination in a functional system. Achieving this level of coordination is hard enough face-to-face, but it doubles down on the internet. That, to me, seems to be one of the issues plaguing the decentralization community. What Is a Blockchain Network? I will give a brief overview of what decentralization and blockchain networks are so that everyone can understand my point more clearly. The blockchain is like a ledger that everyone can read - at least if it’s public. If it’s private or semi-public, then you have to be one of the authorized people to be able to view and change anything in the ledger. But even for public blockchains, a transaction performed on it needs to be verified. The more times it is verified, the more its authenticity is assured. The trouble is these verifications require solving complex mathematical problems which, in turn, require vast amounts of computational resources. Moreover, people working on these are individuals who do not, in real life, know each other. They are only connected through their mutual interest in decentralization. Therefore, in order to ascertain cohesive coordination among these individuals to assure the formation of a well-coordinated decentralized system, protocols are established by the network administrators to clarify the vision and mission of the network. Now understandably, most of these networks’ missions and protocols revolve around data privacy and security because from what I’ve been able to understand--courtesy of Google’s Gemini--is that it is completely possible for a group of hackers or cybercriminals to create a private blockchain network of their own where they freely store and share all kinds of hacked data. So clearly, the blockchain still needs some work to rid it of criminal activity. Now, I understand that my example may be a little extreme and that law-enforcement agencies probably have similar sophisticated technologies to counteract these crimes, and things like digital fingerprints exist, as judiciously explained by Gemini. Nevertheless, these are problems that the blockchain community is rightfully interested in resolving. But that leads to another problem: tunnel vision. The Big Picture The blockchain community is a very exclusive community, composed mostly of young, enthusiastic, tech-savvy individuals. However, the idea of data privacy and security is for everyone, not just the techies. So, it is important that this technology is used and understood as much by regular individuals as it is by its developers. Imagine this scenario. A bank manager or administrator of a small, local bank decides that he wants to convert his banking system to a decentralized one to better the security for the bank customers. He hires a young developer to help and guide him through this change. The developer talks to him about authentication codes, verification, mining data, and other resources needed to form an effective blockchain network. The manager--even if he is relatively young--will probably understand that the system access should be limited to a few trusted individuals, and maybe a little bit more. However, unless he himself is a computer scientist--unlikely in the case of a bank manager--it’s doubtful that he will understand the whole concept of decentralization. But the tipping point would be the computational costs required. Remember, the bank is a small, local bank with a limited budget and resources. When you add this to the fact that the manager may or may not have completely understood how the system works, the odds that the bank will end up installing an effective decentralized system are slim. The Gaps One of the main issues stopping decentralization from becoming a global phenomenon is the lack of suitable hardware, closely followed by inadequate education for the masses. My neural style transfer project was actually not the project I’d originally meant to do. Instead, I had intended to create a Generative AI project using transfer learning to make a GAN algorithm create some new images. My dataset--which was composed of the works of lesser-known artists--wasn’t a very big one, so I knew that it was possible that the end result may not be as novel as it would have been with a bigger dataset. But what I hadn’t realized was that even with that small dataset the time, space, and computation required would be so enormous that, for my purposes, it would be better if I switched to a slightly different type of algorithm. That’s kind of my point. I use a regular laptop for most of my work, and for the most part, even for other A.I. projects, it’s usually enough. But it seems that for projects that require any kind of ‘generation’--even to a small extent--this can be a very big problem. In my case, it wasn’t that big of a deal since it was more of a personal project than anything else, but for the task of providing privacy and security to the clients of a bank, this may be a bigger issue because while their computational resources may be better than my tiny, lonely laptop, it’s still unlikely to have the computational power even remotely resembling that which is necessary for maintaining blockchain networks. Furthermore, the blockchain community really needs to understand that most people in positions responsible for the privacy and security of organizations are at the very least in their forties. Even if they are younger, they’re probably not very aware of how decentralization systems work which is understandable given its complexity. One of the reasons that social media platforms are so universally popular across all demographics is that aside from knowing what a few essential buttons do, people don’t need to know very much about it. This is especially so in cases of individuals over the age of 60, and while your average bank manager may not be 60, he/she isn’t going to be 20 or even 30 either. Conclusion Conclusion There is a lot of money being spent on making decentralized systems more reliable and eco-friendly, but maybe what is necessary is some of that money to also be spent on creating hardware that is not so costly and on educating individuals, i.e., marketing according to their demographic and positions in life. I mean, I can use my social media account from any phone and the phone needn’t be a good quality one, or even a new one. As long as the wifi and my app work, I don’t need anything else. All in all, ensuring online coordination requires in and of itself an extensive system. If you’re only looking at the shape of the butterfly then you’re missing the colors and beauty associated with it, and if you only look at the colors, you might end up missing the main thing. So, maybe what the blockchain and decentralization community needs right now for optimal coordination and cooperation are not just networks that focus on increasing privacy and security but instead a system of network of networks, where one network may be responsible for ensuring privacy and security, the other for hardware, and still other which works on marketing and educating plans for the masses all connected towards one unified goal. In this way, privacy and security will no longer be just a catchphrase for the tech-savvy but an asset utilized by everyone.

Stories, News & Mentions about PERSPECTIVE

Perspective’s Stories on HackerNoon

Perspective's News & Mentions