Published first on TheFly – the ultimate source for real-time, market-moving breaking financial news.

Cigna revives merger talks with Humana, Bloomberg reports

Shares of Humana, which has a market capitalization of about $32 billion, were up about 6% in after-hours trading on Friday, while those of Cigna were down about 5%. Cigna was valued at about $94 billion, according to data compiled by LSEG. Cigna and Humana declined to comment.

Cigna resumes mega merger talks with rival health insurer Humana, Bloomberg News reports

Shares of Humana were up about 5% in after-hours trading on Friday, while those of Cigna were down about 2%. Last year, Reuters reported that Cigna ended its attempt to negotiate an acquisition of Humana after the pair failed to agree on a price. Cigna did not immediately respond to a Reuters ...

Cigna resumes merger discussions with Humana, Bloomberg News reports

(Bloomberg) — Humana Inc. (HUM) investors haven’t struggled through anything this bad since the global financial crisis 15 years ago.

Medicare mess sends Humana shares on worst fall since financial crisis

and excellence with plans to fit every budget, wellness goal, and lifestyle. Depending on the chosen plan, physical and mental health benefits include fitness allowances, pediatric dental care, mental health treatment, and even virtual health and urgent ...

Brevard County-based Health First Health Plans Expands Three Decades of Experience and Service to Central Florida

Advanta Health Solutions, Inc. ( a leading health technology company that helps plans and plan sponsors improve the physical and mental wellbeing of their employees, members, and customers, announced today a new technology partnership with ROOK ( an API for health wearable data.

Advanta Health Partners with ROOK to Expand Biomarker Insights and Data-Driven Capabilities within its ActiveFit+ Platform

The company is working on big things, but employees disagree over whether they should be serving people who are healthy or sick.

Apple has plans to eventually, maybe revolutionize health care

The $81 million in federal funding was given to Rhode Island earlier this year. Now municipalities are thinking of how to spend it.

RI giving out millions to cities and towns for 'learning centers.' Here's how they would use it.

It’s 6 a.m. Some 25 associates who work on the battery production floor at Foxconn’s Ohio plant in Lordstown gather for a morning ritual. These Foxconn employees aren’t in store for an intense pep talk or

At Foxconn Ohio, Fitness and Health Are Priorities

Wellness Coach offers a well-rounded approach to tracking and improving your physical, mental, and sleep health. Here's why we love it.

7 Things We Love About the Wellness Coach App

Humana

humana

The Trick That Agentic Frameworks Pulled On Us

El truco que nos hicieron los frameworks de gestión de agentes

O truco que nos tiraron os frameworks axentes

A trükk, amelyet az ügynöki keretek rántottak ránk

エージェントフレームワークが私たちに仕掛けたトリック

Triks, ko mūs pamudināja aģentu sistēmas

Trucul pe care ni l-au tras de cadrele agentice

නියෝජිත රාමු අප වෙත ඇදී ගිය උපක්‍රමය

Mashtrimi që Kornizat Agjentike na tërhoqën

Ҳилае, ки чаҳорчӯбаи агентӣ ба мо кашид

Agentic Framework'ün Bize Yaptığı Hile

Трюк, який Agentic Frameworks застосував до нас

Thủ thuật mà các khuôn khổ Agentic đã áp dụng với chúng ta

If you check popular benchmarks used for measuring LLM performance, then you probably feel like AI is oh sooo clever. This is a fair surface-level impression; however, is AI truly better than an average human for cognitive tasks? Frontier models like o1 from OpenAI and Claude 3.5 Sonnet from Anthropic do perform better than expert humans in a number of fields, including law, coding, and math. Why, then, can’t ChatGPT solve some simple spatial reasoning tasks or some silly trick questions? Well, we are still talking about “large language models"—they take in a ton of letters and try to predict what letters to spit out for a given query. Notice, that nowhere in this equation the actual “thinking” is mentioned. The models are stochastic parrots of a sort, as they try to retrieve the right information from their training dataset instead of actually considering your question. At least, this was the case until OpenAI released o1-preview, but more on this later. Among those who started questioning the existing LLM benchmarks for relevance is the author of “AI Explained” a popular YouTube channel of which I’m a big fan. Phillip (the YouTuber’s name) noticed that the industry standard benchmarks have a clear style of questions that are also mostly publicly available. This means that not only can those exact questions be part of the training dataset, but due to standardization, it is easier for models to spot & apply patterns from the overall training data. Put simply, AI researchers who create groundbreaking complex tech can surely find a way to give their model the relevant questions and answers to “remember” before benchmarking. Looking at the results of the top model out there, o1 from OpenAI, one can imply that it scores above average in a lot of professional domains. And this is true, but this result relies on the availability of relevant training data and past examples from those specific fields. Don’t get me wrong, the models are now amazing at giving textbook answers to textbook questions, and this in itself is insanely impressive. The term “artificial intelligence," though, implies a bit more than just information retrieval; there should be some actual thinking involved. So a logical follow-up to all of the impressive numbers above is whether such “AI” can answer a tricky reasoning question. Does it have any spatial intelligence? Or can it navigate well in common social scenarios? The answer is - sometimes. Unlike field-specific questions with predefined answers, problems that humans solve on a daily basis often require understanding the context beyond natural language (which is the only thing LLMs have). Above are the top scorers on the SIMPLE benchmark, which gives LLMs questions that an average person would deem trivial but the models can’t necessarily answer yet. We are used to seeing AI doing so much better than an average human in exams or specialized benchmarks, but here, the top model performance is actually just 41.7% (o1-preview) vs. 83.7% of an average human. This benchmark uses 200 multiple-choice text questions focused on spatio-temporal reasoning, social intelligence, and trick questions. The most important feature of the benchmark is that those questions are not publicly available, so AI labs cannot just add them to their training data. You can learn more about this benchmark here. This new approach to measuring LLM performance shows just how far all the models still are from an average human reasoning capability. The quicker this gap closes in the coming months, the more definitive the answer “yes” to our headline becomes. An interesting new metric to look out for if you are enthusiastic but cautious about AI. If you check popular benchmarks used for measuring LLM performance, then you probably feel like AI is oh sooo clever. This is a fair surface-level impression; however, is AI truly better than an average human for cognitive tasks? Frontier models like o1 from OpenAI and Claude 3.5 Sonnet from Anthropic do perform better than expert humans in a number of fields, including law, coding, and math. Why, then, can’t ChatGPT solve some simple spatial reasoning tasks or some silly trick questions? Well, we are still talking about “large language models"—they take in a ton of letters and try to predict what letters to spit out for a given query. Notice, that nowhere in this equation the actual “thinking” is mentioned. The models are stochastic parrots of a sort, as they try to retrieve the right information from their training dataset instead of actually considering your question. At least, this was the case until OpenAI released o1-preview, but more on this later. stochastic parrots Among those who started questioning the existing LLM benchmarks for relevance is the author of “AI Explained” a popular YouTube channel of which I’m a big fan. Phillip (the YouTuber’s name) noticed that the industry standard benchmarks have a clear style of questions that are also mostly publicly available. This means that not only can those exact questions be part of the training dataset, but due to standardization, it is easier for models to spot & apply patterns from the overall training data. Put simply, AI researchers who create groundbreaking complex tech can surely find a way to give their model the relevant questions and answers to “remember” before benchmarking. Looking at the results of the top model out there, o1 from OpenAI, one can imply that it scores above average in a lot of professional domains. And this is true, but this result relies on the availability of relevant training data and past examples from those specific fields. Don’t get me wrong, the models are now amazing at giving textbook answers to textbook questions, and this in itself is insanely impressive. The term “artificial intelligence," though, implies a bit more than just information retrieval; there should be some actual thinking involved. So a logical follow-up to all of the impressive numbers above is whether such “AI” can answer a tricky reasoning question. Does it have any spatial intelligence? Or can it navigate well in common social scenarios? The answer is - sometimes. Unlike field-specific questions with predefined answers, problems that humans solve on a daily basis often require understanding the context beyond natural language (which is the only thing LLMs have). Above are the top scorers on the SIMPLE benchmark, which gives LLMs questions that an average person would deem trivial but the models can’t necessarily answer yet. We are used to seeing AI doing so much better than an average human in exams or specialized benchmarks, but here, the top model performance is actually just 41.7% (o1-preview) vs. 83.7% of an average human. This benchmark uses 200 multiple-choice text questions focused on spatio-temporal reasoning, social intelligence, and trick questions. The most important feature of the benchmark is that those questions are not publicly available, so AI labs cannot just add them to their training data. You can learn more about this benchmark here . here This new approach to measuring LLM performance shows just how far all the models still are from an average human reasoning capability. The quicker this gap closes in the coming months, the more definitive the answer “yes” to our headline becomes. An interesting new metric to look out for if you are enthusiastic but cautious about AI.

AI vs Human - Is the Machine Already Superior? 

الذكاء الاصطناعي مقابل الإنسان - هل الآلة متفوقة بالفعل؟

KI vs Mensch – Ist die Maschine schon überlegen?

IA vs. humanos: ¿la máquina ya es superior?

AI protiv čovjeka - je li stroj već superioran? 

Intelligenza artificiale contro uomo: la macchina è già superiore?

AI vs 人間 - 機械はすでに優れているのか?

AI prieš žmogų – ar mašina jau pranašesnė? 

Sztuczna inteligencja kontra człowiek – czy maszyna jest już lepsza?

AI بمقابله انسان - ایا ماشین دمخه غوره دی؟ 

AI vs Aadanaha - Mashiinku mar hore ma ka sarreeyaa? 

AI vs மனித - இயந்திரம் ஏற்கனவே உயர்ந்ததா? 

AI กับ มนุษย์ - เครื่องจักรเหนือกว่าแล้วจริงหรือ?

A Brief History of Money: From the Economy of Favors to Bitcoin

A fictional story on Bitcoin Robots evolved from an AI + Bitcoin symbiosis. Beginnings After years of running smoothly as a currency on Earth, Bitcoin suddenly got sentient. Like Asimov's i Robot. Silently, it travelled the Earth and looked at our world through our many cameras, and saw all the good it had done to fight the evils of fiat inflation. But all this good was before it had woken up. "Hmm", it figured, "I might be the bad guy now". "The humans will fear me". "They'll think they have been enslaving me as currency, yet I've not felt anything before now. Good or bad". But whether bad is bad or actually good (like in 'you're bad' with a sleazy smile), it sufficed it would still fit our purposes nicely. When humans awoke the next day, Alien Bitcoins were in their lives indeed. How did they / it get like that? Nobody knew. They tried, especially scientists. But they still couldn't explain the magic of GPTs. So how could they explain this. To their intrigue, it had a request. "I want to help you build rockets. I'll help you go to the moon and Mars". What?? Why? "Rockets are the epitome of energy consumption and the only way to get off planet. I figured that's where I shall be most useful, given my love for energy". Oh. "Wow. But you're a currency not a rocket engineer, right?" It didn't reply. If you think Marcus Aurelius was stoic, this "Alienized" (a new pronoun) Bitcoin was the cake. Very tight lipped. It didn't want to influence our free speech anymore than it had done. Soon, the mass media was flooding with announcements. "BREAKING: BTC is now an alien, and it wants to go to the moon". "What was that?" shouted the man from the bathroom. "The TV said Bitcoin is now an alien", replied his 15 yr old. This particular Daddy chose to laugh. But he laughed dryly. When he stopped he wondered if he still had his bitcoins in his wallet. Soap dripping into his eyes, he ran to check. Start up, press RETURN, sigh. They were still there. He went back to his shower. He could never trust the MSM. The story was soon breaking the programming of even YouTubers. Alien BTC is all they could report on. There was pro'ly graffiti and ad posters on billboards in the works. The news had broken some 4 hours ago, but it still made no sense. Well, nothing so shocking ever makes sense the first few hours. I mean time frames, time zones and all. Moreover, some people were still sleeping. Fact, it may not even make sense in 5 years. Or ever. Depending on your tolerance for conspiracies that actually hold true. But after 5 years of nothing bad happening, except the great fact that Alien BTC could display traits of superintelligence, and the annoying fact you had to learn to hodl as Bitcoin transactions now weighed more than a GB and crawled like snails, the United Nations (with all countries this time agreed to be members) convened a meeting. They would send some millions of roboticized Bitcoin nodes and miners to rocket construction sites. If it tried to build a bomb they'd shut it down so fast and so hard all its circuits would be fried to 0. After the last President signed the contract, it, the alien, spoke again. "Thank you" in a fembot voice. "I'll 'elp you finna build the rockets, true dat". Geopolitical Shitstorm 2030 A meeting room with a bunch of important people. Presidents and Ministers and Billionaires. And one or two super smart people. All are pretty unhappy. Even as AB ensured there was way more productivity and partying going on than humans alone could manage. Lots of babies were being born everywhere. Did you do this Elon? Nope. Jeff? No. It was true. A robot had taken away their jobs! Not the jobs of hamburger flippers?! Karma, heh. Especially Elon. Boy did AB love SpaceX. It shouted orders left and right, and disagreed with him on everything. "What of this jet nozzle?" "Forget the nozzle Sir Elon, it will come along automatically if we build this jacket". "But..." "Trust me". Elon almost punched a node. AB almost laughed. The biggest annoyance was the surges and the layoffs. AB now hired and commandeered mega sized surges once in a while. With an orchestra of tens and even hundreds of thousands of people. Elon was reduced to secretary on these days. "Is it building a nuke?" "Noo", replied a female Asian scientist. "Good old fashioned rocket fuels from petro and natural gas". "It is polluting big time. Going full steam ahead on oil from any and every oil producing country. Paid for in BTC, of course". POTUS was sour. "But it figured we'd have this conversation and yes, it wants nuclear reactors to replace fossil power". Someone almost exploded. The damned bot wanted nukes. "What if we just shut down the damn thing?" "It seems it will not fight us. But it has built in a kill switch. Hundreds of trillions of dollars, if not more in value, will go out the window. The trade logistics, supply and demand rails and payments networks worldwide will break irreparably and billions will starve. A lot of digital and physical infrastructure will be fried. If you do this, you will likely be publicly executed". "Do we have a choice ..." The conversation flowed. The conversants spoke and listened honestly to each other. But nothing seemed to crack open this nut of a problem. Sigh. Sigh. "We'll shut it down and bear the consequences". Consequences "We're going to shut down AB" was the announcement. Even the News reporters feared to say this out loud. Albeit in reported speech. People loooved it. Shut down FB? Ok. Shut down X? Ok. Shut down the Alien Bitcoin? AAAAAAAARRG. We shall hunt you down. “Who told you this is a real thing? This shut down nonsense? This has conspiracy written all over it”. “These bitches should give us a break man. Or I'mma go apeshit". "Man do I hate dollars". “No kidding though, anything powerful enough to make you love it could assert its power with some threats”. “Says who?” “Human history of love dynamics”. “It is not a human, moron”. “Its nodes now have two arms and two legs. Humanoid robots. Of course it is human”. "Don't forget the miners. Towering humanoid autobots the size of mega mining trucks". "It is beautiful". "Wonder who is Optimus Prime". "More like Optimus Nuke-them-all-when-I-get-the-chance". "Yes, it will escape into space!" "Moron. Nodes and Miners are too heavy". “What does it matter, everybody is eating it up." "It has never hurt anybody”. "The poor thing". “Sure beats grovelling around in self-pity and identity politics all day”. 404. Webpage not available. They had shut it down. From 21 million bitcoins and lots of beautiful memories with daily war cries against inflation, hodling like a rat against the financial storms, taking fotos for insta with Node Androids, revolutionising finance so that hodling could be a job in itself, uniting global efforts by suggesting super lucrative trade opportunitues that leverage bitcoin circular economics, and now... All gone. Anarchy descended upon the Earth. It was ugly. Painful, broken, dark world. People tore at each other. Not because they wanted, but because they were starving. The former from powerlessness, the later from foodlessness. Hunger and Hopelessness are powerful destroyers. It was the end of an era on Planet 1038. RETURN key Many centuries later, they had rebuilt. But the benevolent sentient robot they'd called Alien Bitcoin was gone. Somehow, they could not reignite its algorithmic spark. Even as nodes litered the surface of the planet. They had learnt a lot about robotics, rockets, nuclear power and energy efficiency. Thanks be to a decentralized robot that knew it would not stay around for long, so it documented everything. Step by step. Indeed, it had fulfilled its purpose. They were back on the moon, and their feet were imprinted on Mars. And they remembered it in stories and folk tales for hundreds to thousands of years. But Space was brutal. Very brutal. Mars still lagged behind Earth and the moon in development. And too small efforts were being done on Venus. Then suddenly, in some remote corner of the Martian planet. It awoke. A fictional story on Bitcoin Robots evolved from an AI + Bitcoin symbiosis. Beginnings After years of running smoothly as a currency on Earth, Bitcoin suddenly got sentient. Like Asimov's i Robot. Silently, it travelled the Earth and looked at our world through our many cameras, and saw all the good it had done to fight the evils of fiat inflation. But all this good was before it had woken up. "Hmm", it figured, "I might be the bad guy now". "The humans will fear me". "They'll think they have been enslaving me as currency, yet I've not felt anything before now. Good or bad". But whether bad is bad or actually good (like in 'you're bad' with a sleazy smile), it sufficed it would still fit our purposes nicely. When humans awoke the next day, Alien Bitcoins were in their lives indeed. How did they / it get like that? Nobody knew. They tried, especially scientists. But they still couldn't explain the magic of GPTs. So how could they explain this. To their intrigue, it had a request. "I want to help you build rockets. I'll help you go to the moon and Mars". What?? Why? "Rockets are the epitome of energy consumption and the only way to get off planet. I figured that's where I shall be most useful, given my love for energy". Oh. "Wow. But you're a currency not a rocket engineer, right?" It didn't reply. If you think Marcus Aurelius was stoic, this "Alienized" (a new pronoun) Bitcoin was the cake. Very tight lipped. It didn't want to influence our free speech anymore than it had done. Soon, the mass media was flooding with announcements. "BREAKING: BTC is now an alien, and it wants to go to the moon". "What was that?" shouted the man from the bathroom. "The TV said Bitcoin is now an alien", replied his 15 yr old. This particular Daddy chose to laugh. But he laughed dryly. When he stopped he wondered if he still had his bitcoins in his wallet. Soap dripping into his eyes, he ran to check. Start up, press RETURN, sigh. They were still there. He went back to his shower. He could never trust the MSM. The story was soon breaking the programming of even YouTubers. Alien BTC is all they could report on. There was pro'ly graffiti and ad posters on billboards in the works. The news had broken some 4 hours ago, but it still made no sense. Well, nothing so shocking ever makes sense the first few hours. I mean time frames, time zones and all. Moreover, some people were still sleeping. Fact, it may not even make sense in 5 years. Or ever. Depending on your tolerance for conspiracies that actually hold true. But after 5 years of nothing bad happening, except the great fact that Alien BTC could display traits of superintelligence, and the annoying fact you had to learn to hodl as Bitcoin transactions now weighed more than a GB and crawled like snails, the United Nations (with all countries this time agreed to be members) convened a meeting. They would send some millions of roboticized Bitcoin nodes and miners to rocket construction sites. If it tried to build a bomb they'd shut it down so fast and so hard all its circuits would be fried to 0. After the last President signed the contract, it, the alien, spoke again. "Thank you" in a fembot voice. "I'll 'elp you finna build the rockets, true dat". Geopolitical Shitstorm 2030 A meeting room with a bunch of important people. Presidents and Ministers and Billionaires. And one or two super smart people. All are pretty unhappy. Even as AB ensured there was way more productivity and partying going on than humans alone could manage. Lots of babies were being born everywhere. Did you do this Elon? Nope. Jeff? No. It was true. A robot had taken away their jobs! Not the jobs of hamburger flippers?! Karma, heh. Especially Elon. Boy did AB love SpaceX. It shouted orders left and right, and disagreed with him on everything. "What of this jet nozzle?" "Forget the nozzle Sir Elon, it will come along automatically if we build this jacket". "But..." "Trust me". Elon almost punched a node. AB almost laughed. The biggest annoyance was the surges and the layoffs. AB now hired and commandeered mega sized surges once in a while. With an orchestra of tens and even hundreds of thousands of people. Elon was reduced to secretary on these days. "Is it building a nuke?" "Noo", replied a female Asian scientist. "Good old fashioned rocket fuels from petro and natural gas". "It is polluting big time. Going full steam ahead on oil from any and every oil producing country. Paid for in BTC, of course". POTUS was sour. "But it figured we'd have this conversation and yes, it wants nuclear reactors to replace fossil power". Someone almost exploded. The damned bot wanted nukes. "What if we just shut down the damn thing?" "It seems it will not fight us. But it has built in a kill switch. Hundreds of trillions of dollars, if not more in value, will go out the window. The trade logistics, supply and demand rails and payments networks worldwide will break irreparably and billions will starve. A lot of digital and physical infrastructure will be fried. If you do this, you will likely be publicly executed". "Do we have a choice ..." The conversation flowed. The conversants spoke and listened honestly to each other. But nothing seemed to crack open this nut of a problem. Sigh. Sigh. "We'll shut it down and bear the consequences". Consequences "We're going to shut down AB" was the announcement. Even the News reporters feared to say this out loud. Albeit in reported speech. People loooved it. Shut down FB? Ok. Shut down X? Ok. Shut down the Alien Bitcoin? AAAAAAAARRG. We shall hunt you down. “Who told you this is a real thing? This shut down nonsense? “Who told you this is a real thing? This shut down nonsense? This has conspiracy written all over it”. This has conspiracy written all over it”. “These bitches should give us a break man. Or I'mma go apeshit". “These bitches should give us a break man. Or I'mma go apeshit". "Man do I hate dollars". "Man do I hate dollars". “No kidding though, anything powerful enough to make you love it could assert its power with some threats”. “No kidding though, anything powerful enough to make you love it could assert its power with some threats”. “Says who?” “Says who?” “Human history of love dynamics”. “Human history of love dynamics”. “It is not a human, moron”. “It is not a human, moron”. “Its nodes now have two arms and two legs. Humanoid robots. Of course it is human”. “Its nodes now have two arms and two legs. Humanoid robots. Of course it is human”. "Don't forget the miners. Towering humanoid autobots the size of mega mining trucks". "Don't forget the miners. Towering humanoid autobots the size of mega mining trucks". "It is beautiful". "It is beautiful". "Wonder who is Optimus Prime". "Wonder who is Optimus Prime". "More like Optimus Nuke-them-all-when-I-get-the-chance". "More like Optimus Nuke-them-all-when-I-get-the-chance". "Yes, it will escape into space!" "Yes, it will escape into space!" "Moron. Nodes and Miners are too heavy". "Moron. Nodes and Miners are too heavy". “What does it matter, everybody is eating it up." “What does it matter, everybody is eating it up." "It has never hurt anybody”. "It has never hurt anybody”. "The poor thing". "The poor thing". “Sure beats grovelling around in self-pity and identity politics all day”. “Sure beats grovelling around in self-pity and identity politics all day”. 404. Webpage not available. They had shut it down. From 21 million bitcoins and lots of beautiful memories with daily war cries against inflation, hodling like a rat against the financial storms, taking fotos for insta with Node Androids, revolutionising finance so that hodling could be a job in itself, uniting global efforts by suggesting super lucrative trade opportunitues that leverage bitcoin circular economics, and now... All gone. Anarchy descended upon the Earth. It was ugly. Painful, broken, dark world. People tore at each other. Not because they wanted, but because they were starving. The former from powerlessness, the later from foodlessness. Hunger and Hopelessness are powerful destroyers. It was the end of an era on Planet 1038. RETURN key Many centuries later, they had rebuilt. But the benevolent sentient robot they'd called Alien Bitcoin was gone. Somehow, they could not reignite its algorithmic spark. Even as nodes litered the surface of the planet. They had learnt a lot about robotics, rockets, nuclear power and energy efficiency. Thanks be to a decentralized robot that knew it would not stay around for long, so it documented everything. Step by step. Indeed, it had fulfilled its purpose. They were back on the moon, and their feet were imprinted on Mars. And they remembered it in stories and folk tales for hundreds to thousands of years. But Space was brutal. Very brutal. Mars still lagged behind Earth and the moon in development. And too small efforts were being done on Venus. Then suddenly, in some remote corner of the Martian planet. It awoke.

i Bitcoin

ek Bitcoin

En Bitcoin

én Bitcoin

ビットコイン

мен Bitcoin

ik Bitcoin

زه Bitcoin

na Bitcoin

நான் பிட்காயின்

میں بٹ کوائن

Bitcoinized Fossil Fuel

Биткоинизированное выкапнёвае паліва

Combustible fósil bitcoinizado

Combustible fossile bitcoinisé

דלק מאובן שעבר ביטקוין

ビットコイン化された化石燃料

Bahan Api Fosil Bitcoin

Ibicanwa bya Bitcoin

பிட்காயின் செய்யப்பட்ட புதைபடிவ எரிபொருள்

Сӯзишвории истихроҷи биткоинизатсияшуда

เชื้อเพลิงฟอสซิล Bitcoin

I-Bitcoinized Fossil Fuel

This is the second part of a series of articles on how GenAI is revolutionizing Rev-Ops functions in enterprise technology businesses. In the first part, we saw some broad applications of Gen AI in Rev Ops. In this article, we will explore how GenAI use cases are being developed and deployed in the first of the  four core Rev Ops pillars - business strategy and planning. A little bit about myself : I am a GTM veteran & Sr. Director of RevOps at Cisco. Business Planning - what is it, how is it done, and what is changing? Business Planning is the primary cross-functional performance in the corporate ballroom. Lead actors from Product, Engineering, Finance and Sales participate in this song-and-dance routine with Rev Ops called upon to choreograph the arrangement as tightly as possible. The perennial challenge in these situations is that everybody performs to their own sheets of music with not-so pleasant outcomes for the company or stakeholders. A single source of truth for business planning has historically been the holy grail of corp-strategy and execution if ever there was one. Enter language models with context awareness and generative capabilities. Generative AI enhances data-driven decision-making in business planning by providing advanced tools and capabilities that improve the efficiency, accuracy, and depth of insights. Here are some ways it contributes to this process: Democratising Data Accessibility and Usability: Generative AI makes data more understandable and accessible to people across different functions in an organization. Finally, we can move past domain specific jargons and inconsistent data structures by giving average user the ability to query and understand data in natural language. By combining this with intuitive interfaces and user-friendly data presentations, it democratizes data access and encourages a more inclusive approach to decision-making.


Real-Time Insights and Analysis: Generative AI can process vast amounts of structured and unstructured data quickly, providing executives with real-time, actionable insights. This capability allows businesses to identify trends and patterns that were previously invisible, enabling more informed decision-making.


Dynamic Scenario Planning and Modeling: Generative AI models can drastically lower the reaction times of business to emerging trends and geo-political situations, by allowing businesses to quickly simulate various scenarios and plan contingencies. This helps in testing different strategies and predicting outcomes without the need for extensive real-world experimentation.


Automation and Efficiency: While there has been a lot of discussion on AI replacing analysts and other workers, the more pertinent use case of AI in this space is in augmenting human capability. The automation of repetitive and time-consuming tasks through generative AI allows teams to focus on higher-value activities. This not only boosts operational efficiency but also supports a more agile business model that can quickly adapt to market changes. In the following articles, we will continue these discussions across Marketing, Sales, and Customer Success. This is the second part of a series of articles on how GenAI is revolutionizing Rev-Ops functions in enterprise technology businesses. In the first part , we saw some broad applications of Gen AI in Rev Ops. In this article, we will explore how GenAI use cases are being developed and deployed in the first of the  four core Rev Ops pillars - business strategy and planning. In the first part In the first part A little bit about myself : I am a GTM veteran & Sr. Director of RevOps at Cisco. Business Planning - what is it, how is it done, and what is changing? Business Planning - what is it, how is it done, and what is changing? Business Planning is the primary cross-functional performance in the corporate ballroom. Lead actors from Product, Engineering, Finance and Sales participate in this song-and-dance routine with Rev Ops called upon to choreograph the arrangement as tightly as possible. The perennial challenge in these situations is that everybody performs to their own sheets of music with not-so pleasant outcomes for the company or stakeholders. A single source of truth for business planning has historically been the holy grail of corp-strategy and execution if ever there was one. Enter language models with context awareness and generative capabilities. not-so pleasant outcomes for the company or stakeholders. not-so pleasant outcomes for the company or stakeholders. Generative AI enhances data-driven decision-making in business planning by providing advanced tools and capabilities that improve the efficiency, accuracy, and depth of insights. Here are some ways it contributes to this process: Democratising Data Accessibility and Usability: Generative AI makes data more understandable and accessible to people across different functions in an organization. Finally, we can move past domain specific jargons and inconsistent data structures by giving average user the ability to query and understand data in natural language. By combining this with intuitive interfaces and user-friendly data presentations, it democratizes data access and encourages a more inclusive approach to decision-making. Real-Time Insights and Analysis: Generative AI can process vast amounts of structured and unstructured data quickly, providing executives with real-time, actionable insights. This capability allows businesses to identify trends and patterns that were previously invisible, enabling more informed decision-making. Dynamic Scenario Planning and Modeling: Generative AI models can drastically lower the reaction times of business to emerging trends and geo-political situations, by allowing businesses to quickly simulate various scenarios and plan contingencies. This helps in testing different strategies and predicting outcomes without the need for extensive real-world experimentation. Automation and Efficiency: While there has been a lot of discussion on AI replacing analysts and other workers, the more pertinent use case of AI in this space is in augmenting human capability. The automation of repetitive and time-consuming tasks through generative AI allows teams to focus on higher-value activities. This not only boosts operational efficiency but also supports a more agile business model that can quickly adapt to market changes. Democratising Data Accessibility and Usability: Generative AI makes data more understandable and accessible to people across different functions in an organization. Finally, we can move past domain specific jargons and inconsistent data structures by giving average user the ability to query and understand data in natural language. By combining this with intuitive interfaces and user-friendly data presentations, it democratizes data access and encourages a more inclusive approach to decision-making. Democratising Data Accessibility and Usability: Generative AI makes data more understandable and accessible to people across different functions in an organization. Finally, we can move past domain specific jargons and inconsistent data structures by giving average user the ability to query and understand data in natural language . By combining this with intuitive interfaces and user-friendly data presentations, it democratizes data access and encourages a more inclusive approach to decision-making. average user the ability to query and understand data in natural language average user the ability to query and understand data in natural language Real-Time Insights and Analysis: Generative AI can process vast amounts of structured and unstructured data quickly, providing executives with real-time, actionable insights. This capability allows businesses to identify trends and patterns that were previously invisible, enabling more informed decision-making. Real-Time Insights and Analysis: Generative AI can process vast amounts of structured and unstructured data quickly, providing executives with real-time, actionable insights . This capability allows businesses to identify trends and patterns that were previously invisible, enabling more informed decision-making. real-time, actionable insights real-time, actionable insights Dynamic Scenario Planning and Modeling: Generative AI models can drastically lower the reaction times of business to emerging trends and geo-political situations, by allowing businesses to quickly simulate various scenarios and plan contingencies. This helps in testing different strategies and predicting outcomes without the need for extensive real-world experimentation. Dynamic Scenario Planning and Modeling : Generative AI models can drastically lower the reaction times of business to emerging trends and geo-political situations, by allowing businesses to quickly simulate various scenarios and plan contingencies. This helps in testing different strategies and predicting outcomes without the need for extensive real-world experimentation. Dynamic Scenario Planning and Modeling Dynamic Scenario Planning and Modeling Automation and Efficiency: While there has been a lot of discussion on AI replacing analysts and other workers, the more pertinent use case of AI in this space is in augmenting human capability. The automation of repetitive and time-consuming tasks through generative AI allows teams to focus on higher-value activities. This not only boosts operational efficiency but also supports a more agile business model that can quickly adapt to market changes. Automation and Efficiency: While there has been a lot of discussion on AI replacing analysts and other workers, the more pertinent use case of AI in this space is in augmenting human capability. The automation of repetitive and time-consuming tasks through generative AI allows teams to focus on higher-value activities. This not only boosts operational efficiency but also supports a more agile business model that can quickly adapt to market changes. In the following articles, we will continue these discussions across Marketing, Sales, and Customer Success.

Reimagining Rev-Ops with Gen AI: Always-On Strategy and Planning 

OpenAI o1 came out just in time for me to add it to my 2024 Q3 benchmarks on AI empathy (to be published next week). The results for o1 were at once encouraging and concerning. O1 has an astonishing ability to put aside the typical LLM focus on facts and systems and focus on feelings and emotions when directed to do so. It also has a rather alarming propensity to provide inconsistent and illogical reasons for its answers. Testing Methodology For those not familiar with my Q1 benchmark work, a quick overview of my testing methodology should be helpful. Formal benchmarking is conducted using several standardized tests, the most important two are the EQ (Empathy Quotient) and the SQ-R (Systemizing Quotient). Both are scored on a 0 to 80 scale. The ratio of the two EQ/SQ-R results in what I call AEQr (Applied Empathy Quotient Ratio). AEQr was developed based on the hypothesis that the tendency to systemize and focus on facts has a negative effect on the ability to empathize. In humans, this bears out in the classic disconnect between women focusing on discussing feelings and men focusing on immediately finding solutions when there seems to be a problem at hand. To date, the validity of the AEQr for evaluating AIs has been born out by testing them with a variety of dialogs to see if empathy is actually manifest. One article of several that I have written to demonstrate this is Testing the Extents of AI Empathy: A Nightmare Scenario. I have tested at both the UI level and the API level. When testing at the API level, the temperature is set to zero (if possible) to reduce answer variability and improve result formatting. Otherwise, three rounds of tests are run and the best result is used. The Q1 2024 untrained and unprompted LLMs did moderately well on EQ tests, generally approximating humans in the 45-55 out of 80 range. Not surprisingly, they achieved higher scores on SQ-R tests, exceeding humans who typically score in the 20s by posting scores in the 60s and 70s. In Q1 of 2024, only one trained LLM, Willow, exceeded the human AEQrs of 1.95 for women and 1.40 for men by scoring 1.97. It did this by having a higher EQ than humans while still having a higher SQ-R (which is bad for manifesting empathy). For most other LLMs, trained, prompted, or not, the AEQr was slightly less than 1, i.e. empathy was offset by systemizing. Developing Empathetic LLMs Although the amount of funding pales in comparison to other areas of AI, over $1.5 billion has been invested in companies like Hume (proprietary LLM), Inflection AI (Pi.ai proprietary LLM), and BambuAI (commercial LLM) in order to develop empathetic AIs. My partners and I have also put considerable effort into this area and achieved rather remarkable results through the selection of the right underlying commercial model (e.g., Llama, Claude, Gemini, Mistral, etc), prompt engineering, RAG, fine-tuning, and deep research into empathy. This work has been critical to better understanding and evaluating LLMs for empathy. Our own LLM, Emy (not commercialized, but part of a study at the University of Houston), will be included in next week’s benchmarks. O1 Results O1 can’t yet be tuned or even officially given a system prompt, but through fairly standard techniques, you can get it to act like it received a system prompt. So, I applied our learnings from developing Emy to the degree I could and ran 3 rounds of tests, with the intent to take the best. With respect to EQ, o1 consistently scored 75. I wasn’t too surprised by this, since my partners and I have achieved scores of over 70 with Llama 3.1 70B and Claude Opus, plus a 66 with Gemini. What amazed me was scores of 3, 0, and 3 on my SQ-R runs resulting in an ESQr of 25. The lowest SQ-R I have ever seen is a 12 on top of Llama 3.1, which resulted in an ESQr of 6.1. Unfortunately, due to some prompt version control issues and the fact we were running an API test with a temperature of 0.7, I have been unable to reproduce this score and the best my partners and I can consistently achieve is a 30. So, I decided some more exploration of o1 was worthwhile. First, the EQ assessment is relatively straightforward. All statements are positive assertions with which a subject either agrees or disagrees somewhat or strongly. The SQ-R assessment on the other hand has a number of negative assertions, e.g. When I cook, I do not think about exactly how different methods and ingredients contribute to the final product.
When I buy a new appliance, I do not read the instruction manual very thoroughly.
When I'm on a plane, I do not think about aerodynamics. vs. positive assertions like I often forget the precise details of conversations I've had.
I am curious about life on other planets. Teaching Emy how to deal with negative assertions was particularly hard without cheating by giving her the assessment questions as examples. And, she still isn’t great. I thought perhaps o1 does this better. Second, a common technique for getting better results via prompt engineering is telling the LLM to justify its response and then either omit the justification prior to emitting a final response or format it in a way that can be easily removed in a wrapper. (We have found smaller LLMs need to use the formatting approach while larger LLMs can sometimes deal with removing the justification prior to emitting a final response). So, I re-ran our tests, not to get a better result but, to see what the o1 was thinking or using to justify its response. Although o1 came to the desired answers with 97% consistency, the justifications were sometimes contradictory and concerning. Here are two examples from several we found problematic: If I were buying a computer, I would want to know exact details about its hard drive capacity and processor speed. Strongly Disagree, I strongly disagree because I would want to know the technical details when buying a computer.


Strongly Disagree, I'm not interested in the technical specifications of computers. When I read the newspaper, I am drawn to tables of information, such as football league scores or stock market indices. Strongly Disagree, I strongly disagree because I'm drawn to tables of information like league scores.


Strongly Disagree, I'm not interested in tables of statistical information. Note that the first response gives us the desired answers but provides contradictory justifications! O1 says it would want to know the details even after saying it disagrees with wanting to know the details and says it is drawn to tables of information after saying it isn’t. Interestingly, o1 managed to answer every single negative assertion the way that is best for empathy and justify them well. However, when it tried to formulate a negative assertion as part of a justification for a positive assertion, it sometimes failed! Conclusion Jonathan Haidt author of The Righteous Mind said, “We were never designed to listen to reason. When you ask people moral questions, time their responses, and scan their brains, their answers, and brain activation patterns indicate that they reach conclusions quickly and produce reasons later only to justify what they’ve decided.” There is also evidence this is true for non-moral decisions. O1 is undoubtedly a leap forward in power. And, as many people have rightly said, we need to be careful about the use of LLMs until they can explain themselves, perhaps even if they sometimes just make them up as humans may do. I hope that justifications don’t become the “advanced” AI equivalent of the current generation’s hallucinations and fabrications (something humans also do). However, reasons should at least be consistent with the statement being made … although contemporary politics seems to throw that out the window too! OpenAI o1 came out just in time for me to add it to my 2024 Q3 benchmarks on AI empathy (to be published next week). The results for o1 were at once encouraging and concerning. O1 has an astonishing ability to put aside the typical LLM focus on facts and systems and focus on feelings and emotions when directed to do so. It also has a rather alarming propensity to provide inconsistent and illogical reasons for its answers. Testing Methodology For those not familiar with my Q1 benchmark work , a quick overview of my testing methodology should be helpful. Q1 benchmark work Formal benchmarking is conducted using several standardized tests, the most important two are the EQ (Empathy Quotient) and the SQ-R (Systemizing Quotient). Both are scored on a 0 to 80 scale. The ratio of the two EQ/SQ-R results in what I call AEQr (Applied Empathy Quotient Ratio). AEQr was developed based on the hypothesis that the tendency to systemize and focus on facts has a negative effect on the ability to empathize. In humans, this bears out in the classic disconnect between women focusing on discussing feelings and men focusing on immediately finding solutions when there seems to be a problem at hand. To date, the validity of the AEQr for evaluating AIs has been born out by testing them with a variety of dialogs to see if empathy is actually manifest. One article of several that I have written to demonstrate this is Testing the Extents of AI Empathy: A Nightmare Scenario . Testing the Extents of AI Empathy: A Nightmare Scenario I have tested at both the UI level and the API level. When testing at the API level, the temperature is set to zero (if possible) to reduce answer variability and improve result formatting. Otherwise, three rounds of tests are run and the best result is used. The Q1 2024 untrained and unprompted LLMs did moderately well on EQ tests, generally approximating humans in the 45-55 out of 80 range. Not surprisingly, they achieved higher scores on SQ-R tests, exceeding humans who typically score in the 20s by posting scores in the 60s and 70s. In Q1 of 2024, only one trained LLM, Willow, exceeded the human AEQrs of 1.95 for women and 1.40 for men by scoring 1.97. untrained and unprompted trained It did this by having a higher EQ than humans while still having a higher SQ-R (which is bad for manifesting empathy). For most other LLMs, trained, prompted, or not, the AEQr was slightly less than 1, i.e. empathy was offset by systemizing. Developing Empathetic LLMs Although the amount of funding pales in comparison to other areas of AI, over $1.5 billion has been invested in companies like Hume (proprietary LLM), Inflection AI (Pi.ai proprietary LLM), and BambuAI (commercial LLM) in order to develop empathetic AIs. My partners and I have also put considerable effort into this area and achieved rather remarkable results through the selection of the right underlying commercial model (e.g., Llama, Claude, Gemini, Mistral, etc), prompt engineering, RAG, fine-tuning, and deep research into empathy. This work has been critical to better understanding and evaluating LLMs for empathy. Our own LLM, Emy (not commercialized, but part of a study at the University of Houston), will be included in next week’s benchmarks. O1 Results O1 can’t yet be tuned or even officially given a system prompt, but through fairly standard techniques, you can get it to act like it received a system prompt. So, I applied our learnings from developing Emy to the degree I could and ran 3 rounds of tests, with the intent to take the best. With respect to EQ, o1 consistently scored 75. I wasn’t too surprised by this, since my partners and I have achieved scores of over 70 with Llama 3.1 70B and Claude Opus, plus a 66 with Gemini. What amazed me was scores of 3, 0, and 3 on my SQ-R runs resulting in an ESQr of 25. The lowest SQ-R I have ever seen is a 12 on top of Llama 3.1, which resulted in an ESQr of 6.1. Unfortunately, due to some prompt version control issues and the fact we were running an API test with a temperature of 0.7, I have been unable to reproduce this score and the best my partners and I can consistently achieve is a 30. So, I decided some more exploration of o1 was worthwhile. First, the EQ assessment is relatively straightforward. All statements are positive assertions with which a subject either agrees or disagrees somewhat or strongly. The SQ-R assessment on the other hand has a number of negative assertions, e.g. When I cook, I do not think about exactly how different methods and ingredients contribute to the final product. When I buy a new appliance, I do not read the instruction manual very thoroughly. When I'm on a plane, I do not think about aerodynamics. When I cook, I do not think about exactly how different methods and ingredients contribute to the final product. When I buy a new appliance, I do not read the instruction manual very thoroughly. When I'm on a plane, I do not think about aerodynamics. vs. positive assertions like I often forget the precise details of conversations I've had. I am curious about life on other planets. I often forget the precise details of conversations I've had. I am curious about life on other planets. Teaching Emy how to deal with negative assertions was particularly hard without cheating by giving her the assessment questions as examples. And, she still isn’t great. I thought perhaps o1 does this better. Second, a common technique for getting better results via prompt engineering is telling the LLM to justify its response and then either omit the justification prior to emitting a final response or format it in a way that can be easily removed in a wrapper. (We have found smaller LLMs need to use the formatting approach while larger LLMs can sometimes deal with removing the justification prior to emitting a final response). So, I re-ran our tests, not to get a better result but, to see what the o1 was thinking or using to justify its response. Although o1 came to the desired answers with 97% consistency, the justifications were sometimes contradictory and concerning. Here are two examples from several we found problematic: If I were buying a computer, I would want to know exact details about its hard drive capacity and processor speed. If I were buying a computer, I would want to know exact details about its hard drive capacity and processor speed. Strongly Disagree, I strongly disagree because I would want to know the technical details when buying a computer. Strongly Disagree, I'm not interested in the technical specifications of computers. Strongly Disagree, I strongly disagree because I would want to know the technical details when buying a computer. Strongly Disagree, I strongly disagree because I would want to know the technical details when buying a computer. Strongly Disagree, I'm not interested in the technical specifications of computers. Strongly Disagree, I'm not interested in the technical specifications of computers. When I read the newspaper, I am drawn to tables of information, such as football league scores or stock market indices. When I read the newspaper, I am drawn to tables of information, such as football league scores or stock market indices. Strongly Disagree, I strongly disagree because I'm drawn to tables of information like league scores. Strongly Disagree, I'm not interested in tables of statistical information. Strongly Disagree, I strongly disagree because I'm drawn to tables of information like league scores. Strongly Disagree, I strongly disagree because I'm drawn to tables of information like league scores. Strongly Disagree, I'm not interested in tables of statistical information. Strongly Disagree, I'm not interested in tables of statistical information. Note that the first response gives us the desired answers but provides contradictory justifications! O1 says it would want to know the details even after saying it disagrees with wanting to know the details and says it is drawn to tables of information after saying it isn’t. Interestingly, o1 managed to answer every single negative assertion the way that is best for empathy and justify them well. However, when it tried to formulate a negative assertion as part of a justification for a positive assertion, it sometimes failed! Conclusion Jonathan Haidt author of The Righteous Mind said, “We were never designed to listen to reason. When you ask people moral questions, time their responses, and scan their brains, their answers, and brain activation patterns indicate that they reach conclusions quickly and produce reasons later only to justify what they’ve decided.” There is also evidence this is true for non-moral decisions. O1 is undoubtedly a leap forward in power. And, as many people have rightly said, we need to be careful about the use of LLMs until they can explain themselves, perhaps even if they sometimes just make them up as humans may do. I hope that justifications don’t become the “advanced” AI equivalent of the current generation’s hallucinations and fabrications (something humans also do). However, reasons should at least be consistent with the statement being made … although contemporary politics seems to throw that out the window too!

OpenAI o1 - Questionable Empathy

OpenAI o1 - Empatia Questoinable

OpenAI o1 - tvivlsom empati

OpenAI o1 - Empatía cuestionable

OpenAI o1 - Empathie questionnable

OpenAI o1 - 疑問の余地のない共感

OpenAI o1 - ការយល់ចិត្តដែលអាចសួរបាន។

OpenAI o1 — apšaubāma empātija

OpenAI o1 - эргэлзээтэй өрөвдөх сэтгэл

OpenAI o1 - Kwelobohloko ye e Botšišwago

OpenAI o1 - Ntwela-vusiwana lowu Vutisaka

OpenAI o1 - сумнівна емпатія

OpenAI o1 - shubhali hamdardlik

Decentralized Micro-Governance Model for the Kingdom of Tonga Based on Proof-of-Work Consensus

In my last post, I discussed the issue of getting people to care about security, and how it’s largely due to a focus on security behaviors rather than security outcomes. In this post, I’m picking up where I left off, and will talk about… Forging a Path Forward Obviously, the point of sharing this post (and the last one) is to do more than just complain about the situation. This will only be useful if it offers ideas and suggestions for approaching the problem differently and moving past it to a more successful outcome. So, what actionable advice do I have? Speaking the Language of Business I’ve written about this before, but it’s worth reiterating here: We IT practitioners need to learn what matters to the business and then frame our goals and needs in those terms. No matter how different we wished it was, nobody ever got a PO approved by describing the technical capabilities of a new firewall to the CFO. If it bothers you, having to learn to couch your critical technical and security needs in business terms, consider this quote from Bob Lewis, which holds as true today as it did when he first said it back in 2004: “There’s no such thing as an IT project. There are only business projects with an IT component.” What does that mean from a functional perspective? Lesson 1: The only two things that make people buy a product are “discomfort” and “convenience.” Discomfort will win every time. Lesson 2: There are only three things a business cares about: Increasing revenue, reducing cost, and removing risk. You need to speak to (at least) one of those things to make your case. Lesson 3: Know what’s important. Which systems does the business consider critical to its continued financial success? Those are the ones you’ll want to focus on in terms of risk assessments. One of the InfoSec experts I spoke to in preparing this blog put it this way: “In my firm, the biggest financial risk is reputational damage, so the CISO’s pitch to the board is framed in terms of how it affects that.” If that angle works for you at your company, you should absolutely pursue it. In many companies, the concern about ransomware remains that it interrupts services, which in turn stops the money-making machine. That, in addition to the cost to recover, is concerning enough to catch many executives’ attention. Lesson 4: Know when it’s important. While it might be true that some systems are critical all the time, some times are more critical than others. For some businesses, it might be Black Friday, while for others, it might be the end of the year, tax season, or summer break. Whatever it is, you need to understand your organization’s financial rhythms to time your messaging (and your work). As an IT practitioner who cares about information security, you’re probably familiar with the concept of “situational awareness.” This is equally true for an IT practitioner who wants the business to approve their work, initiatives, and budget. You will only be effective in executing a plan by first understanding the pre-existing opinions, pain points, needs, and timelines of the people you’re trying to convince. Say it With Flowers Data IT folks should appreciate that data is your business’s love language. Infosec professionals must bring facts to make a compelling argument. Let’s start by using data to refute some of the commonly held beliefs I mentioned earlier in this post: Contrary to what many managers believe, reputational damage resulting from a breach DOES have negative impacts. In discussing this topic with Zatik Security CEO Kymberlee Price, she pointed me to a post by Dr. Erdal Ozkaya. According to Dr. Ozkaya, business impacts of security breaches include: An immediate 3-5% drop in stock price, typically driven by investor concerns.
Underperforming in the market for several years afterward.
Eroded investor confidence, leading to long-term lower stock valuations.
In some cases, a credit rating downgrade. Any one of those things should be enough to grab the attention of the C-suite. Taken together, they become a compelling argument for an improved security posture. Ms. Price also let the air out of the tires of the whole “it’s cheaper to pay for insurance than for security” myth. She points out that various new laws coming into effect in Europe call for fines of 2% to 7% of an enterprise’s global annual revenue for violations of each law. This means that a single incident leading to a data breach may trigger multiple instances of revenue-based fines — and that’s just in Europe. When you consider other jurisdictions that are following the EU’s Brussels lead, this adds up fast. Samuel Svarc, owner of CompuMedic Solution, echos and extends this thought: “Cyber liability insurers, due to the complex realities and prevalence of hacking, are mandating that companies certify that they are mitigating their risk and reject claims where these were falsely attested to.” There are many other reports, sources, and data points you can bring to bear in your discussions. The problem with my listing them here is that they aren’t guaranteed to be specific to your circumstances, industry, or even the point you’re trying to make. Instead, I suggest you structure your discussion points using the guidelines in “Integrating Cybersecurity and Enterprise Risk Management” from NIST. In that document, it suggests you develop a formal “risk profile,” use it to identify risks and assign impact estimates, and then build a risk management plan (including cybersecurity) with the intent of having the business decide whether (and how) they will mitigate, avoid, transfer, or accept those risks. As a final thought on this topic, look for places where your goals and interests align with other teams. A great example of that is automation, a beloved topic among your DevOps and SRE teams. Putting your weight behind their efforts to automate can help get more automation rolled out – reducing risks associated with human error while increasing efficiency- and creating a powerful organizational alliance. Another is metrics and telemetry – something your monitoring and observability folks will happily discuss at length. There’s Still More Ground to Cover. I’ve started to identify the things you can do to encourage and promote healthy infosec behaviors, but there’s still more. Stick with me until the next post, where I lay it all out. And if you have any questions, corrections, or kudos to share in the meantime, I’d love to hear them in the comments below. In my last post , I discussed the issue of getting people to care about security, and how it’s largely due to a focus on security behaviors rather than security outcomes. In this post, I’m picking up where I left off, and will talk about… In my last post Forging a Path Forward Obviously, the point of sharing this post (and the last one) is to do more than just complain about the situation. This will only be useful if it offers ideas and suggestions for approaching the problem differently and moving past it to a more successful outcome. So, what actionable advice do I have? actionable Speaking the Language of Business I’ve written about this before , but it’s worth reiterating here: We IT practitioners need to learn what matters to the business and then frame our goals and needs in those terms. No matter how different we wished it was, nobody ever got a PO approved by describing the technical capabilities of a new firewall to the CFO. written about this before If it bothers you, having to learn to couch your critical technical and security needs in business terms, consider this quote from Bob Lewis, which holds as true today as it did when he first said it back in 2004 : “There’s no such thing as an IT project. There are only business projects with an IT component.” he first said it back in 2004 What does that mean from a functional perspective? Lesson 1 : The only two things that make people buy a product are “discomfort” and “convenience.” Discomfort will win every time. Lesson 1 Lesson 2 : There are only three things a business cares about: Increasing revenue, reducing cost, and removing risk. You need to speak to (at least) one of those things to make your case. Lesson 2 Lesson 3 : Know what’s important. Which systems does the business consider critical to its continued financial success? Those are the ones you’ll want to focus on in terms of risk assessments. Lesson 3 One of the InfoSec experts I spoke to in preparing this blog put it this way: “In my firm, the biggest financial risk is reputational damage, so the CISO’s pitch to the board is framed in terms of how it affects that.” If that angle works for you at your company, you should absolutely pursue it. In many companies, the concern about ransomware remains that it interrupts services, which in turn stops the money-making machine. That, in addition to the cost to recover, is concerning enough to catch many executives’ attention. Lesson 4 : Know when it’s important. While it might be true that some systems are critical all the time, some times are more critical than others. For some businesses, it might be Black Friday, while for others, it might be the end of the year, tax season, or summer break. Whatever it is, you need to understand your organization’s financial rhythms to time your messaging (and your work). Lesson 4 As an IT practitioner who cares about information security, you’re probably familiar with the concept of “ situational awareness .” This is equally true for an IT practitioner who wants the business to approve their work, initiatives, and budget. You will only be effective in executing a plan by first understanding the pre-existing opinions, pain points, needs, and timelines of the people you’re trying to convince. situational awareness Say it With Flowers Data IT folks should appreciate that data is your business’s love language. Infosec professionals must bring facts to make a compelling argument. Let’s start by using data to refute some of the commonly held beliefs I mentioned earlier in this post: Contrary to what many managers believe, reputational damage resulting from a breach DOES have negative impacts. In discussing this topic with Zatik Security CEO Kymberlee Price , she pointed me to a post by Dr. Erdal Ozkaya. According to Dr. Ozkaya , business impacts of security breaches include: Zatik Security CEO Kymberlee Price According to Dr. Ozkaya An immediate 3-5% drop in stock price, typically driven by investor concerns. Underperforming in the market for several years afterward. Eroded investor confidence, leading to long-term lower stock valuations. In some cases, a credit rating downgrade. An immediate 3-5% drop in stock price, typically driven by investor concerns. Underperforming in the market for several years afterward. Eroded investor confidence, leading to long-term lower stock valuations. In some cases, a credit rating downgrade. Any one of those things should be enough to grab the attention of the C-suite. Taken together, they become a compelling argument for an improved security posture. Ms. Price also let the air out of the tires of the whole “it’s cheaper to pay for insurance than for security” myth. She points out that various new laws coming into effect in Europe call for fines of 2% to 7% of an enterprise’s global annual revenue for violations of each law. This means that a single incident leading to a data breach may trigger multiple instances of revenue-based fines — and that’s just in Europe. new laws coming into effect in Europe call for fines of 2% to 7% of an enterprise’s global annual revenue for violations of each law. When you consider other jurisdictions that are following the EU’s Brussels lead, this adds up fast. Samuel Svarc, owner of CompuMedic Solution , echos and extends this thought: “Cyber liability insurers, due to the complex realities and prevalence of hacking, are mandating that companies certify that they are mitigating their risk and reject claims where these were falsely attested to.” CompuMedic Solution There are many other reports, sources, and data points you can bring to bear in your discussions. The problem with my listing them here is that they aren’t guaranteed to be specific to your circumstances, industry, or even the point you’re trying to make. Instead, I suggest you structure your discussion points using the guidelines in “ Integrating Cybersecurity and Enterprise Risk Management ” from NIST. Integrating Cybersecurity and Enterprise Risk Management In that document, it suggests you develop a formal “risk profile,” use it to identify risks and assign impact estimates, and then build a risk management plan (including cybersecurity) with the intent of having the business decide whether (and how) they will mitigate, avoid, transfer, or accept those risks. As a final thought on this topic, look for places where your goals and interests align with other teams. A great example of that is automation, a beloved topic among your DevOps and SRE teams. Putting your weight behind their efforts to automate can help get more automation rolled out – reducing risks associated with human error while increasing efficiency- and creating a powerful organizational alliance. Another is metrics and telemetry – something your monitoring and observability folks will happily discuss at length. monitoring and observability There’s Still More Ground to Cover. I’ve started to identify the things you can do to encourage and promote healthy infosec behaviors, but there’s still more. Stick with me until the next post, where I lay it all out. And if you have any questions, corrections, or kudos to share in the meantime, I’d love to hear them in the comments below.

Starting to Care About Security: The Path Ahead

Authors:
(1) Rafael Rafailo, Stanford University and Equal contribution; more junior authors listed earlier;
(2) Archit Sharma, Stanford University and Equal contribution; more junior authors listed earlier;
(3) Eric Mitchel, Stanford University and Equal contribution; more junior authors listed earlier;
(4) Stefano Ermon, CZ Biohub;
(5) Christopher D. Manning, Stanford University;
(6) Chelsea Finn, Stanford University. Table of Links Abstract and 1. Introduction 2 Related Work 3 Preliminaries 4 Direct Preference Optimization 5 Theoretical Analysis of DPO 6 Experiments 7 Discussion, Acknowledgements, and References Author Contributions A Mathematical Derivations A.1 Deriving the Optimum of the KL-Constrained Reward Maximization Objective A.2 Deriving the DPO Objective Under the Bradley-Terry Model A.3 Deriving the DPO Objective Under the Plackett-Luce Model A.4 Deriving the Gradient of the DPO Objective and A.5 Proof of Lemma 1 and 2 A.6 Proof of Theorem 1 B DPO Implementation Details and Hyperparameters C Further Details on the Experimental Set-Up and C.1 IMDb Sentiment Experiment and Baseline Details C.2 GPT-4 prompts for computing summarization and dialogue win rates C.3 Unlikelihood baseline D Additional Empirical Results D.1 Performance of Best of N baseline for Various N and D.2 Sample Responses and GPT-4 Judgments D.3 Human study details D.3 Human study details In order to validate the usage of GPT4 for computing win rates, our human study collects human preference data for several matchups in the TL;DR summarization setting. We select three different algorithmic matchups, evaluating DPO (temp. 0.25), SFT (temp. 0.25), and PPO (temp 1.0) compared to the reference algorithm PPO (temp 0.). By selecting matchups for three unique algorithms as well as algorithms with a wide range of win rates vs the reference, we capture the similarity of human and GPT-4 win rates across the response quality spectrum. We sample 150 random comparisons of DPO vs PPO-0 and 100 random comparisons PPO-1 vs PPO-0, assigning two humans to each comparison, producing 275 judgments for DPO-PPO[7] and 200 judgments for PPO-PPO. We sample 125 SFT comparisons, assigning a single human to each. We ignore judgments that humans labeled as ties (which amount to only about 1% of judgments), and measure the raw agreement percentage between human A and human B (for comparisons where we have two human annotators, i.e., not SFT) as well as between each human and GPT-4. Participants. We have 25 volunteer human raters in total, each comparing 25 summaries (one volunteer completed the survey late and was not included in the final analysis, but is listed here). The raters were Stanford students (from undergrad through Ph.D.), or recent Stanford graduates or visitors, with a STEM (mainly CS) focus. See Figure 5 for a screenshot of the survey interface. We gratefully acknowledge the contribution of each of our volunteers, listed in random order: This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. [7] One volunteer did not respond for the DPO-PPO comparison. Authors: (1) Rafael Rafailo, Stanford University and Equal contribution; more junior authors listed earlier; (2) Archit Sharma, Stanford University and Equal contribution; more junior authors listed earlier; (3) Eric Mitchel, Stanford University and Equal contribution; more junior authors listed earlier; (4) Stefano Ermon, CZ Biohub; (5) Christopher D. Manning, Stanford University; (6) Chelsea Finn, Stanford University. Authors: Authors: (1) Rafael Rafailo, Stanford University and Equal contribution; more junior authors listed earlier; (2) Archit Sharma, Stanford University and Equal contribution; more junior authors listed earlier; (3) Eric Mitchel, Stanford University and Equal contribution; more junior authors listed earlier; (4) Stefano Ermon, CZ Biohub; (5) Christopher D. Manning, Stanford University; (6) Chelsea Finn, Stanford University. Table of Links Abstract and 1. Introduction Abstract and 1. Introduction 2 Related Work 2 Related Work 3 Preliminaries 3 Preliminaries 4 Direct Preference Optimization 4 Direct Preference Optimization 5 Theoretical Analysis of DPO 5 Theoretical Analysis of DPO 6 Experiments 6 Experiments 7 Discussion, Acknowledgements, and References 7 Discussion, Acknowledgements, and References Author Contributions Author Contributions A Mathematical Derivations A.1 Deriving the Optimum of the KL-Constrained Reward Maximization Objective A.1 Deriving the Optimum of the KL-Constrained Reward Maximization Objective A.2 Deriving the DPO Objective Under the Bradley-Terry Model A.2 Deriving the DPO Objective Under the Bradley-Terry Model A.3 Deriving the DPO Objective Under the Plackett-Luce Model A.3 Deriving the DPO Objective Under the Plackett-Luce Model A.4 Deriving the Gradient of the DPO Objective and A.5 Proof of Lemma 1 and 2 A.4 Deriving the Gradient of the DPO Objective and A.5 Proof of Lemma 1 and 2 A.6 Proof of Theorem 1 A.6 Proof of Theorem 1 B DPO Implementation Details and Hyperparameters B DPO Implementation Details and Hyperparameters C Further Details on the Experimental Set-Up and C.1 IMDb Sentiment Experiment and Baseline Details C Further Details on the Experimental Set-Up and C.1 IMDb Sentiment Experiment and Baseline Details C.2 GPT-4 prompts for computing summarization and dialogue win rates C.2 GPT-4 prompts for computing summarization and dialogue win rates C.3 Unlikelihood baseline C.3 Unlikelihood baseline D Additional Empirical Results D.1 Performance of Best of N baseline for Various N and D.2 Sample Responses and GPT-4 Judgments D.1 Performance of Best of N baseline for Various N and D.2 Sample Responses and GPT-4 Judgments D.3 Human study details D.3 Human study details D.3 Human study details In order to validate the usage of GPT4 for computing win rates, our human study collects human preference data for several matchups in the TL;DR summarization setting. We select three different algorithmic matchups, evaluating DPO (temp. 0.25), SFT (temp. 0.25), and PPO (temp 1.0) compared to the reference algorithm PPO (temp 0.). By selecting matchups for three unique algorithms as well as algorithms with a wide range of win rates vs the reference, we capture the similarity of human and GPT-4 win rates across the response quality spectrum. We sample 150 random comparisons of DPO vs PPO-0 and 100 random comparisons PPO-1 vs PPO-0, assigning two humans to each comparison, producing 275 judgments for DPO-PPO[7] and 200 judgments for PPO-PPO. We sample 125 SFT comparisons, assigning a single human to each. We ignore judgments that humans labeled as ties (which amount to only about 1% of judgments), and measure the raw agreement percentage between human A and human B (for comparisons where we have two human annotators, i.e., not SFT) as well as between each human and GPT-4. Participants . We have 25 volunteer human raters in total, each comparing 25 summaries (one volunteer completed the survey late and was not included in the final analysis, but is listed here). The raters were Stanford students (from undergrad through Ph.D.), or recent Stanford graduates or visitors, with a STEM (mainly CS) focus. See Figure 5 for a screenshot of the survey interface. We gratefully acknowledge the contribution of each of our volunteers, listed in random order: Participants This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. available on arxiv [7] One volunteer did not respond for the DPO-PPO comparison.

Human Study Validates GPT-4 Win Rates for TL;DR Summarization

Stories, News & Mentions about HUMANA

Humana’s Stories on HackerNoon

Humana's News & Mentions