Weav AIis in the AI Infrastructure Industry

Weav AI

Fiction Students are hacking cars, pushing their vehicles and algorithms to the limits. Nvidia’s Drive PX 2 supercomputer is used to fuel a new kind of street race We travel to Cambridge Massachusetts, where an underground scene of renegades are tearing up the backroads. Multi-million dollar Bitcoin wagers and terabyte data dumps are fueling a new scene of street racing. With the advent of advanced driver assist (ADAS) in modern cars, (for example lane keeping and self-parking), autonomous driving has opened to the masses. Standard ADAS, paired with open source software like Comma.ai, Udacity, and Baidu’s Apollo is giving hackers complete control over their vehicles. A new breed of street racer is taking over, consisting of computer science majors despondent in the age of Snapchat dog filters, heartless Tinder dates, and dormroom binge drinking. Like the slicked back hot rod culture of the 1950s, a unique subculture is getting their adrenaline fix behind the wheel. I met with “Sean” at 11am on sunday. Sean’s name has been changed for obvious reasons. He’s a student, finishing a degree in CS, and from what I’ve been told, the mastermind behind this entire operation. What he was about to show me was a highly illegal operation. He told me later that the risk was worth it, “This is something no one has ever done before. Like Silk Road or Cryptocurrency, this is a technology that changes everything.” A completely new sport is emerging from this unnamed group’s lines of code. Sean flicked a cigarette outside a nondescript, run down industrial building in Cambridge, Massachusetts. He keyed us in, and led me through a door beside a steel garage door. Daft Punk thumped through a dimly lit hallway, ending in an open space. It looked like an old mechanics shop, 1970s bulletin boards now sported photos of cars and clippings from Wired Magazine. Three cars in various orders of assembly lay gleaming before me. Half a dozen students worked feverishly. Marijuana wafted from the vape-pen of someone coding next to a car. It took five weeks to get the introduction to even get in here, and the engineering students barely noticed I was there. Sean kissed his girlfriend and showed me his Acura. Cans of Rockstar energy and cigarrette buts littered the floor. A server rack covered the back wall. Ethernet cables of all color littered the ceiling, snaking out from the server rack. He walked me to his heavily modified car. The hood was open, but other than that, one couldn’t see anything special about the car. It was the ultimate sleeper. The guy smoking hash on his laptop was dubbed “Einstein”. He wrote several of the machine learning algorithms that took open source software and supercharged it. His black Macbook had a cable running into a custom designed T-Box, plugged directly into the CAN interface of the car. Any one of these engineers could have 6 figure salaries at Google, Uber, or any one of the dozens of companies pouring billions into the self driving car race. Instead they were doing doing this, pushing AV technology far beyond anyone else. “Something happens when you completely deregulate something,” Sean explained, “It moves it maximum speed.” In his case, the team was working around regulation. By now a few of the engineers had looked up at the stranger in their oasis of code and cars and hashish. I asked how Sean funded the venture and he remained silent. I changed tone, asking how this all began. “It started with the release of an AV source code being put on github,” He explained, looking down into the rebuilt engine of the Acura, “The damn code ran on a OnePlus 3 smartphone. It was too easy I had to give it a try.” Sean is referring to Comma.ai’s release of OpenPilot. The controversial and classically brash San Francisco startup is using the 820 GPU in a smartphone to capture data, and control a car. The DIY kit plugs into the vehicle’s CAN interface, which controls acceleration and steering. “So we screwed around with that, pushing it to it’s limits,” Sean explained, “Until some douche in a Mustang shot by me going 150,” he kicked an empty can of Rockstar on the floor, “That’s when this crazy idea hit me. What if we can drive better than any human ever could?” Sean had a way of pulling you into his words that couldn’t quite be conveyed. Over the past 8 months he got together a crack team of CS majors, gearheads, druggies, bassheads, and true innovators. All of them happened to be pretty good at Deep Learning. After 6 months of reverse engineering other people’s code, and finally building their own, they were ready for their first race. Sean sat behind the wheel of his “coded out”, white Acura. He pointed to two cameras in the corners of the window, along with the existing ADAS camera. “You’ve got to believe in the code,” He said almost religiously, setting his hands on the steering wheel, “She’s a machine that is truly self aware, and has to be treated with respect.” In the trunk of the car, two Nvidia PX supercomputers, with respective cooling systems were nestled into the back. The steering controller had to be swapped out for one with higher torque, and the VCU completely remapped. I asked Sean again, sitting in the passenger seat now, how he funded everything. He looked over at the server rack, then back to me, “Let’s just say… I made a lot of money with Bitcoin. The savings from are enough to put a wager on a race larger than anyone before.” He looked me in the eye, “Enough on that.” that We are living in an age of nondescript Bitcoin millionaires. Kids who mined or coded something stupid-clever in their teens, and end up completely set by the time they’re in university. He climbed out of the car, “So now that you know we’re not full of it, I’ll send you an email when we race.” I asked when. Sean laughed and glanced back at the server rack, “When the rest of the wagers have been made.” Ushering me out, I took one last look at the “coded out” cars in the shop. They looked almost boring from the outside. I asked one final question, why he let me see all of this. Sean was about to open the door to the street when he paused for a moment, “Because,” thinking, “We’re making history. These vehicles are gaining consciousness and I want the world to see it.” I couldn’t even tell that the cars behind me were unique. It would not be for a few weeks that an email arrived from a random address in my inbox. It simply said a time and a place, which turned out to be a library, along with “Don’t be late :)” I boarded the next flight from from SFO to Logan, leaving my South Boston AirBnb at just past 3am. It was July in Boston, and the summer nights were warm and pleasant. It reminded me of screwing around in high school, hopping barbed wire fences and drinking stolen alcohol. We met at the public library on the outskirts of Boston. Sean was waiting for me in his Acura. A few tiny cameras had been added in the back windows, along with a police cruiser style laptop holder, “Man your Lyft driver is slow. Hop in.” On his Macbook, Google Earth was open, showing a traced route of my cab ride extending about a mile from where we sat. “Don’t be offended, I grabbed you IMEI when we met weeks ago. The Brits call em IMSI Catchers, the feds, Stingrays. We use em to track cops.” Sean explained that every cop car has either a radio repeater, or a cop with a cell phone. The software defined radio (SDR) required to track IMEIs could easily be bought from China for a few hundred bucks. Each of the racecars had an SDR, and their signals triangulated to pinpoint police. I did not ask, but a fast enough SDR could be used to jam most police signals, including their cell phones. Sean turned down the brightness of the screen and checked his watch, a glowing Rolex GMT. The Acura started without him touching a control, the headlights turned on, the engine revved. We were on our way. A map appeared on the Macbook showing two other cars, along with a view around the car. It was eerie and surreal watching the car’s sensors compile a 3D view of the night around us. “We’re heading to Vermont,” Sean said quietly, he pressed a button on a Motorola radio hidden near the ignition. “Einstein, you pushed the latest build? I want to give it a go before 4G drops out.” Sean grabbed the wheel while he car updated to the latest firmware, and he laid back in his seat again. We didn’t speak much, just a few radio checks. Like he said, 4G became spotty, and eventually we were on our own. Once we were into Vermont, we joined the other two cars. They flashed their high beams behind us in excitement. Sean flicked off the car headlights, the other cars followed suit. Our world was submerged in darkness, driving at over 80mph, we hadn’t even reached first light. Sean began laughing into the radio, “Einstein, the Five-Eight link is working great”. On screen, the perceptual awareness of our vehicle had expanded threefold. He looked at me by the light of his macbook, keeping his eyes off the invisible road ahead, “Who needs headlights when you have RAW camera data?” The Macbook displayed an enormous, god’s eye view of the road ahead and behind. Each car was sharing data over a 150Mbps V2V Wi-Fi link, pulling enough data to drive in the dark. The cars didn’t even flinch. By first light we were deep into the forests of Vermont. For some reason, my Verizon phone wasn’t getting any signal. Lush forests and hills surrounded the highway, the sun peeked over the horizon. Everyone pulled over and stopped, jumping onto their cars and sipping coffee from a thermos. Sean had his gold Pixel phone out. Four guys and three girls stood around the cars, hugging campus hoodies and their lovers. “Ladies and Gentleman,” he looked at each of the forward vehicle cameras. He paused, looking surprised while reading his phone, “The wagers have been set!” He looked each one of his team members in the eye, “Four thousand two hundred Bitcoin,” One of the engineers spit out his coffee, “will be the prize for this race. We have people all over Earth wagering over one hundred and fifty thousand Bitcoin on this!” He paused, “This is unprecedented.” For reference, in July 2017, the value of 150,000 BTC is well over $1 Billion USD. A bit less if properly laundered. Everyone was quiet for a moment. “The funds,” Sean continued, “Have been collected, mixed, and are awaiting distribution.” He turned off the screen, holding his phone up, then checked his watch, “To those of you watching this live via satellite, the time is 10:17 UTC,” He was staring into the camera of his car now, “And this just might be the greatest wager in the history of auto racing.” Sean was dangerously alive with energy. He said quietly, calmly, and quickly, “Let the race begin!” Einstein lit up a joint and passed it to a girl with dreadlocks. Sean kissed his girlfriend, who rode in another car. Everyone was silent for a moment. The next few moments are difficult to remember. I recall climbing into Sean’s coded out Acura and thinking, . I tightened my race harness. this is complete madness Everyone lined up along the highway. Sean’s car stood last, and he calmly did a radio check, a police scan, and checked the satellite internet connection. “Syncing clocks…. Mark!” He hit enter. A countdown timer appeared in the command prompt of our vehicle. There are moments in your life where reality distorts and you are thrust by incredible acceleration into a completely new world. A tremendous noise erupted from all around. The road was cast into light as three tricked out, coded up supercars passed from 0–60mph in less than 4 seconds. I looked back to the blur of tire smoke, as my body pressed into the racing seat. The laptop’s screen before me was cut in half, showing a 3D map around the car, and a hypnotic swirl of code. Billions of possibilities were analyzed and thrown away. The outcomes deemed acceptable were converted via fiber optic into explosions of gasoline, pressing the seat into my back. Before I could blink, the cars were up to 100Mph. I rocked side to side in my 5 point harness seat. Sean was ecstatic, with his hands behind his head. We neared 150Mph, drifting into the hilly turns. If the code failed now, it would be too late to do anything. Why worry one’s self? The other cars would drive quite literally a few inches ahead. Every moment seemed we would crash to our death, but the computers found that risk perfectly acceptable. Perhaps the engine noise was the most hypnotic. Each gear shift was artistic, the way the vehicle committed to its turns was enigmatic. I wanted more. The onscreen map showed several possible trajectories, one turned green. We shot past the car ahead, around a blind turn. A minivan driving head on blared it headlights! Our car swerved, losing traction, and corrected without a hitch. I was beginning to feel nauseous. Sean swore out loud, this is what he lived for. This is also what he would die for, at least at this rate. Making great attempts to hold my stomach, we dove into deadly hairpin turns at over 100mph. I found the computers knew how to drift. The engine began to redline as we accelerated into a straightaway. Something strange happened. Sean’s car centered itself across the dashed center line. The car ahead moved to the right. Behind us, the last car moved into the space beside. Instead of blocking the way, Sean’s car was letting the others pass. The cars weren’t programmed to compete with one another, but work together, pushing their specs to the max. They were making way for each other to see which car could be pushed the hardest. Wild vibration shook our car. The cars besides us struggled to maintain straight, as the wheels vibrated side to side. The speedometer read 160, the satellite link read closer to 200. On the map, a turn appeared. The car began analyzing millions of trajectories, taking longer than usual to pick a path. Sean was breathing heavily now, sweat ran down his face, “What are you gonna to do baby? What are you going to do?” It was a game of chicken, to see which car would make way for the other to turn first. The computer committed, throwing our bodies into the turn. *** Sean and I met a few weeks after the race, at Philz coffee in Palo Alto. Oddly enough the building was owned by Palantir, a symbol of for-profit, big-data mass surveillance. Sean felt right at home, sipping a mint mojito, wearing reflective Ray-Ban sunglasses. The team had been disbanded, most of the members had graduated. Sean moved as far away as he could from the icy winters of New England. That race had left each member a multi millionaire. Besides the winning prize, each member took a substantial cut of the bets made globally. Not a single person had to work another day in their lives, yet almost all of them had either joined a company, or were building their own. I knew it would be my last meeting with Sean. We joked about the race, and conversed on where the technology was heading. He explained that most of the current AV software was rule based. The systems that would learn on their own, without any human input would win in the end. “It’s not a game of rules, it a game of learning.” I asked him what next. “Well, I’ve got more ‘coin than god. And an algorithm better than Google. What do I do now?” Sean sipped his coffee. He looked at me with a smirk, then up to the security camera in the corner, “Find a way to make it seem like I pay taxes!” X Garrett Kinsman Writes Technology out of San Francisco and Bangalore India. This unfortunately is a work of fiction

Hacking the Underground Autonomous Street Racing Scene

There is a difference between being involved in writing hundreds of product descriptions and drowning yourself in the writing of a novel in the “romance” genre. While both of these writings need skills, passion, and personality - the former saps energy. The reasons are - tough deadlines, monotonous nature, and lack of freedom to be experimental. So, how do you get back the main character's energy? Well, there is no new formula to make a monotonous task interesting overnight. But, we can leverage AI tools like ChatGPT to help us write those descriptions, make those translations, and effortlessly personalize the content for us. 5 Ways Generative AI Can Change The Way You Work On Product Catalogs Go from months to minutes: AI as a headstart to write  product descriptions In an e-commerce business, there are a bunch of marketers working on the content for the product catalogs. There are description writers, SEO writers, SEO analysts, growth marketers, and more who work on creating the perfect description for each product. This consumes days or even months depending on the complexity of the product and the number of SKUs involved. However, Generative AI flips this scenario by giving marketers a headstart and helping them escape “ content choke”. Let’s say you are a fashion brand selling a sweatshirt and you want to come up with a description. Then, you may want to write a prompt that needs little to no human intervention (think further edits). Make sure your prompt includes: Your brand voice The necessary specifications of the product Your target audience The keyword Durability AI Prompt: Help me write a product description for a sweatshirt that I am selling on my online store. The description should have a casual but confident tone. My product is red and is a soft fabric. Women above the age of 18 are my target audience and I'm trying to focus on the keyword "Red sweatshirt". It is a daily wear. AI Response: Step into comfort and style with our Red Sweatshirt, your go-to daily wear for women aged 18 and above. Unwind in the soft embrace of its premium fabric, wrapping yourself in cozy confidence. This red beauty effortlessly blends casual charm with a touch of sophistication, making it the perfect companion for your everyday adventures. Whether you're out for a coffee run or enjoying a laid-back day, our Red Sweatshirt adds a pop of color to your wardrobe. Elevate your comfort game with this must-have piece – because feeling good has never looked this chic. Embrace the ease of daily wear, and let the vibrant red hue be your style statement. It's not just a sweatshirt; it's your daily dose of comfort and confidence. It took me less than 2 minutes to get that going 🥳 Never see translation as a barrier: Slay product content in any language Let’s stick to the red sweatshirt example. If you are selling the red sweatshirt globally online then you may want to tailor it for that specific region and in that specific language. In the absence of a Generative AI tool, this could involve a lot of time, resources, and effort to translate content and make it sound native. AI Prompt: Help me translate this English product description into Spanish: “Step into comfort and style with our Red Sweatshirt, your go-to daily wear for women aged 18 and above. Unwind in the soft embrace of its premium fabric, wrapping yourself in cozy confidence. This red beauty effortlessly blends casual charm with a touch of sophistication, making it the perfect companion for your everyday adventures. Whether you're out for a coffee run or enjoying a laid-back day, our Red Sweatshirt adds a pop of color to your wardrobe. Elevate your comfort game with this must-have piece – because feeling good has never looked this chic. Embrace the ease of daily wear, and let the vibrant red hue be your style statement. It's not just a sweatshirt; it's your daily dose of comfort and confidence”. AI Response: "Sumérgete en la comodidad y el estilo con nuestra Sudadera Roja, tu elección diaria para mujeres de 18 años en adelante. Relájate en el suave abrazo de su tela premium, envolviéndote en una confianza acogedora. Esta belleza roja combina sin esfuerzo el encanto casual con un toque de sofisticación, convirtiéndose en la compañera perfecta para tus aventuras diarias. Ya sea que salgas a tomar un café o disfrutes de un día relajado, nuestra Sudadera Roja añade un toque de color a tu guardarropa. Eleva tu estilo de comodidad con esta pieza imprescindible, porque sentirse bien nunca ha lucido tan chic. Acepta la facilidad de la vestimenta diaria y deja que el vibrante tono rojo sea tu declaración de estilo. No es solo una sudadera; es tu dosis diaria de comodidad y confianza." ==== I’ve always wanted a robot as a child to help me do my assignments. It’s so cool to see that happening. How about you? 😎 ==== Recommend personalizing product content to your boss without the fear of workload as the reward We all know that the key to more conversion is personalization. Marketers talk about this easily but to get this done is no piece of cake. So, the next time you pitch personalizing product content to your boss, don’t fret about the workload that follows. Generative AI can help you create personalized content for all your target audiences and segments with simple prompts. AI Prompt: Help me personalize the product description for a sweatshirt that I am selling on my online store. The description should have a casual but confident tone. My product is red and is a soft fabric. Women above the age of 18 are my target audience and I'm trying to focus on the keyword "Red sweatshirt". It is a daily wear. Write 5 descriptions each personalized as follows: My audience is American women who would prefer this for workplace My audience is Canadian women for whom I’m doing a Holiday discount My audience is Indian women who are interested in specific details about my product My audience is Australian women for whom I am giving 25% off My audience is Spanish women and I want it to be in Spanish AI Response: For American Women - Workplace Focus: Step into polished comfort at the workplace with our Red Sweatshirt – a versatile daily wear choice for dynamic American women aged 18 and above. Crafted from irresistibly soft fabric, this red gem seamlessly blends casual charm with a touch of sophistication, ensuring you stay stylish throughout your busy day. Elevate your office wardrobe with a pop of vibrant color, making a confident statement with every wear. For Canadian Women - Holiday Discount: Celebrate the season in style with our Red Sweatshirt, a cozy daily companion perfect for Canadian women aged 18 and above. Immerse yourself in the plush comfort of its soft fabric, and revel in the festive spirit with an exclusive Holiday discount just for you. This red beauty effortlessly combines comfort and style, ensuring you stay warm and chic throughout the winter festivities. For Indian Women - Details Enthusiasts: Explore unparalleled comfort and style with our Red Sweatshirt, tailored for fashion-forward Indian women aged 18 and above. Dive into the luxurious softness of the fabric as you discover the intricate details that make this red marvel a wardrobe essential. From casual elegance to specific design features, this sweatshirt promises to elevate your daily fashion game. For Australian Women - 25% Off Special: Wrap yourself in Australian charm with our Red Sweatshirt, designed for women aged 18 and above. Enjoy the soft embrace of this red stunner, now even more irresistible with a special 25% off. Whether you're heading out for a coffee or embracing a laid-back day, this daily wear adds a touch of Aussie style to your wardrobe, making comfort and confidence your signature. For Spanish-Speaking Women - In Spanish: Sumérgete en el confort y el estilo con nuestra Sudadera Roja, la elección diaria perfecta para mujeres españolas de 18 años en adelante. Relájate en el suave abrazo de su tela premium, fusionando sin esfuerzo el encanto casual con un toque de sofisticación. Ya sea para un café rápido o un día relajado, esta sudadera roja añade un toque de color a tu guardarropa, haciéndote sentir bien con confianza a diario. ¡Haz de la comodidad tu estilo y deja que el vibrante tono rojo hable por ti! == This feels like an 8-hour workday task while with AI it did not even take 8 minutes 🫣 == Sweat to chill: Your catalog creatives can be taken over by sweet AI AI empowers e-commerce businesses by automating and improving the handling of images and media assets. With image recognition algorithms, businesses can automatically tag and categorize vast image libraries, simplifying content management. AI-driven tools enhance product searchability, providing users with a more intuitive and personalized experience. AI algorithms automate the process of tagging images, and streamlining content organization Image recognition improves product searchability, enhancing user navigation and reducing bounce rate AI optimizes media assets for different devices, ensuring a consistent and high-quality visual experience AI tools can dynamically edit images, adjusting aspects like lighting and backgrounds to enhance visual appeal There are tools like Genus AI, that can help you enhance plain product images into more exciting and engaging images. I’d think twice before scheduling photoshoots in exotic locations 📸 Combining a PIM with Generative AI helps you stay consistent across channels Combining a Product Information Management (PIM) system with Generative AI proves to be a strategic synergy for businesses aiming to maintain consistency across various channels. The PIM system acts as a centralized hub for managing and organizing product information, ensuring uniformity and accuracy. When integrated with Generative AI, the process is elevated to new heights, automating the generation of consistent and high-quality content for diverse channels. This powerful combination not only streamlines workflows but also enhances efficiency by automating tasks such as content creation, thereby reducing the risk of inconsistencies. As a result, businesses can present a unified brand identity, accurate product information, and engaging content across all channels, fostering a seamless and cohesive customer experience. When power comes together, there’s nothing like it 🤝 Final thoughts Generative AI revolutionizes e-commerce catalog creation, streamlining processes, enhancing creativity, and providing a scalable solution. The key takeaway is that businesses leveraging generative AI not only achieve operational efficiency but also unlock the potential for more engaging and personalized customer experiences, ultimately reshaping the future of online retail.

The Role of Generative AI in Helping E-commerce Businesses Create Product Catalogs on Autopilot

Last year, I wrote my first . I liked the experience, so I'm trying one more time. Let the future decide if it will become a trend or not. yearly retrospective Before diving into our safe technological world, my thoughts go to Ukraine, to my friends who had to flee their own country, to other friends who fought on the front to defend it from an imperial power, and to all victims of an old kleptocrat who clings to power despite the cost to others. The free world needs to support Ukraine more. I hope 2024 will be the year of Ukrainian victory. The AI revolution Last year, I kept the post focused on what I did. However, AI is pervasive in our tech world, if not the whole world, and deserves a dedicated section. Nobody even remotely connected to tech can ignore the buzz surrounding AI. Even friends and families who don't work near tech probably talk about it. So far, I didn't weave any AI-related thing into any of my talks, despite the huge incentive to do so: having AI in your Call For Paper proposal vastly increases your chances of being selected. And yet, it doesn't mean that I'm not playing with it on a personal level. Here's my experience so far. I've been playing with Dall-E as a use case to try out web development in Rust. I found the results mind-blowing. However, the limitations are enormous; it seems the size is always square, and the set of possible dimensions is limited. I've used ChatGPT in several areas. First, I tried to generate conference abstracts. I fed an entire blog post and asked ChatGPT to turn it into an abstract. Despite my lack of prompt engineering skills, I've found results severely lacking. Abstracts felt artificial, like any content created with ChatGPT, but worse, the abstract revealed too much or not enough. I also tried to use it to refactor two nested Kotlin statements, one nested in the other, to the null-safe construct. The first result didn't compile; the second one forgot a branch. It led me to the correct solution, though. if else let Finally, I recently started using GitHub Copilot as an extension inside of IntelliJ IDEA. Its behavior is fascinating: most of the time, it does nothing, but once in a while, it offers a snippet of a couple of lines, which is either entirely correct or at least very close to the target solution. Even better, the suggestions seem to be more frequent and even more relevant with time. All in all, I'm far from impressed by the current state of AI. However, I like Copilot a lot: I prefer rare advice, which is relevant, than the opposite. Technical content Next February will mark the two-year milestone that I'm working for: on . I'm still very pleased about both. It allows me to do things I like a lot, such as writing posts and giving talks. API7.ai Apache APISIX In 2023, I published fifty blog posts on : one each week on Sunday, but on Christmas and the New Year - for obvious reasons. Here are the top most viewed pages: this blog Rank Post Views Avg. time on page #1 Leverage the richness of HTTP status codes 14,848 :29 #2 A list of cache providers 12,770 1:15 #3 Calling Rust from Python 8,780 :52 #4 My final take on Gradle (vs. Maven) 8,238 2:11 #5 Learning by doing: An HTTP API with Rust 6,642 1:00 I continue to cross-post on different sites. Here are the numbers compared with last year's: Site 2023 2022 Medium 741 564 Dev.to 8156 1,838 Hashnode 89 80 Absolute numbers are not that interesting, but comparing them is. Interestingly, numbers on are growing wildly, while on Hashnode, they plateau. Note that other sites provide no precise follower count or no count. dev.to Besides, I created a script to track daily metrics across sites and social media, just as I did for Apache APISIX. It displays interesting results: As above, numbers are much less important than the trend. Growth visibly happens mainly on and Bluesky for reasons I cannot fathom (yet?). dev.to Finally, I started a weekly newsletter, unoriginally named A Java Geek weekly. So far, I've written a . I list the posts and videos I found interesting during the week. Note that they contain the same links I post on , , and , with a bit more content, either a summary or my opinion. couple of them LinkedIn Mastodon BlueSky Open Source contributions In 2023, after over twenty years in software, I finally became ! I'm both excited and impressed; it's like belonging to a group of mythical beings I'd only heard about. an Apache committer Of course, working on Apache APISIX helped a lot. Yet, the exciting bit is that all my contributions are entirely unrelated to code; they are blog posts, reviews of blog posts written by others, issues, comments on issues, etc. Here's the GitHub summary: Conclusion Last year, my resolutions were: To deepen my understanding of the Apache APISIX ecosystem Write as many blog posts as in 2022 - it's hard to do better, anyway Design at least three new talks I fulfilled the two first goals but unfortunately failed the last item. I have only created two new talks, both based on previous posts. I'll keep them secret for now; they should appear soon on my page in case they are selected. Speaking I lack imagination, so I'll keep the same goals for this year as last year's and add exploring the ecosystem as well. Let's see how it fares. Happy New Year! API7.ai Originally on on January 7th, 2024 published A Java Geek

Everything I Achieved in 2023

Building the AI-powered bank of the future necessitates a comprehensive organizational transformation, spanning customer engagement, decision-making processes, and core technology infrastructure. Banks must embrace AI technologies on a broad scale to stay competitive and meet the ever-increasing expectations of customers. This entails offering intelligent solutions and seamless integration into partner ecosystems. AI plays a pivotal role in guiding customer interactions and optimizing operational costs. To succeed, banks must cultivate a scalable, resilient, and adaptable core technology foundation, accompanied by an operating model that combines the right talent, culture, and organizational structure. The COVID-19 pandemic has accelerated these trends, positioning banks at a crucial juncture to capitalize on this momentum. Can Banks Rise to the AI Challenge? Indeed, they can. By investing in core technologies such as cloud computing, data APIs, and application programming interfaces, banks can enhance scalability, flexibility, and speed while reducing IT overhead. Effective data management promotes timely access, reusability, and development acceleration. APIs streamline operations, foster collaboration with external partners, and enrich the customer experience. AI-powered decision-making for the Bank of the Future involves the utilization of Artificial Intelligence (AI) and Machine Learning (ML) models to analyze extensive and diverse datasets in near real-time. This empowers banks to assess the eligibility of new customers for credit services, determine loan terms and pricing, and mitigate the risk of fraud. To unlock value across the organization, several prerequisites must be met. These include: a centralized data lake, modular technology for transitioning solutions to production, performance management, risk controls integrated into models, edge capabilities like natural language processing and computer vision, and feedback mechanisms for assessing offer effectiveness. Additionally, banks should advance their analytics capabilities to maintain a competitive edge. The Absence of a Clear AI Strategy Banks encounter formidable challenges on multiple fronts. They must achieve the agility, speed, and flexibility typical of fintech companies while navigating the scale, security standards, and regulatory demands associated with traditional financial enterprises. Challenges arise from the absence of a clear AI strategy, underdeveloped core technology and data infrastructure, and outdated operating models and talent strategies. Furthermore, legacy systems pose limitations regarding capacity, flexibility, support for variable computing requirements, data processing, and real-time analysis. The building blocks of an AI-driven bank encompass: customer engagement, AI-powered decision-making, core technology, data infrastructure, and a platform-based operating model Becoming "AI-First" Is Essential for Banks to Remain Competitive in the Digital Age. AI technologies streamline operations, enhance automation, and often outperform human decision-making in terms of speed and accuracy. Embracing AI is an opportunity to unlock substantial additional value, especially as digital engagement trends accelerate and big-tech companies encroach on financial services. AI technologies contribute to revenue growth by delivering personalized services to customers and employees, reducing costs through automation and resource optimization, and identifying new opportunities through enhanced data processing and insights. This can lead to higher profits, personalized experiences at scale, omnichannel excellence, and rapid innovation cycles, with the potential annual value reaching as high as $1 trillion. Banks are expanding their AI usage by adopting AI for: rapid loan approvals, biometric authentication, virtual assistants, and various digital-commerce innovations. They aim to build new value propositions based on advanced AI and analytics capabilities and transition to an "AI-first" strategy. , AI transforms banking by offering personalized omnichannel experiences that are intelligent, tailored, and available across various channels. AI recognizes spending patterns, facilitates frictionless payments through facial recognition, and integrates seamlessly with non-banking applications. For retail customers , AI provides highly personalized offers, efficient omnichannel experiences, and secure banking operations. For small- or medium-sized enterprises It enables split-second loan approvals, biometric authentication, and virtual assistants, allowing banks to understand customer needs better and compete effectively with neobanks and nonbank challengers. Final Word Obstacles that hinder the deployment of AI capabilities at scale include the need for fintech-like agility within a traditional financial enterprise's scale, security, and regulatory constraints. These obstacles often result from the lack of a clear AI strategy, weak core technology and data infrastructure, and outdated operating models and talent strategies. Legacy systems' limitations also impede the adoption of AI technologies, necessitating enhancements in capacity and flexibility.

Strategies, Challenges, and Opportunities for the Financial Industry in the Digital Age

There is no greater topic being discussed in the tech world today than Artificial Intelligence (AI), and there is a good reason for it. The technology is literally changing the world as we know it. While the nightmare scenario that involves what’s known as “singularity,” whereby super-intelligent machines take over and permanently alter human existence through enslavement or eradication is also a highly discussed topic, this really should not be the focus. Yes, Tesla and SpaceX founder Elon Musk, among others, that Artificial General Intelligence (AGI) is one of the greatest risks to human civilization, but again, we need to focus on what benefits AI holds for us in the future and not on theoretical doomsday scenarios. has warned In 2018, MIT president L. Rafael Reif commissioned the . The purpose of the task force was to understand the relationships between emerging technologies and work, to help shape public discourse around realistic expectations of technology, and to explore strategies to enable a future of shared prosperity. MIT Task Force on the Work of the Future In a January 2022 , the task force’s David Autor, David A. Mindell, and Elisabeth B. Reynolds recommend “investing and innovating in skills and training, improving job quality, including modernizing unemployment insurance and labor laws, and enhancing and shaping innovation by increasing federal research and development spending, rebalancing taxes on capital and labor, and applying corporate income taxes equally.” article Ramón López de Mántaras, of the Artificial Intelligence Research Institute (IIIA), in Bellaterra, Spain, that the “final goal of artificial intelligence (AI)—that a machine can have a type of intelligence similar to a human’s—is one of the most ambitious ever proposed by science. In terms of difficulty, it is comparable to other great scientific goals, such as explaining the origin of life or the Universe, or discovering the structure of matter.” explains general He quotes Descartes who wondered whether a complex mechanical system of gears, pulleys, and tubes could possibly emulate thought. “Two centuries later,” writes de Mántaras, “the metaphor had become telephone systems, as it seemed possible that their connections could be likened to a neural network. Today, the dominant model is computational and is based on the digital computer.” De Mántaras focuses on weak and strong AI and notes that the distinction between the two was first introduced by philosopher John Searle in an article criticizing AI in 1980, “which provoked considerable discussion at the time, and still does today. Strong AI would imply that a properly designed computer does not simulate a mind but , and should, therefore, be capable of an intelligence equal, or even superior to human beings.” actually is one He also delineates the differences between symbolic and connectionist AI models. According to Ashok Goel at the Georgia Institute of Technology, “While symbolic AI posits the use of knowledge in reasoning and learning as critical to producing intelligent behavior, connectionist AI postulates that learning of associations from data (with little or no prior knowledge) is crucial for understanding behavior.” , the subtle and major differences between the various AI models, including evolutionary and corporeal, suggests that it is “increasingly important for businesses to understand how these models work and the potential implications of using them.” According to Nico Klingler In his titled, “How Artificial Intelligence Will Change the World”, author Mike Thomas explains how AI will influence our lives in a changing world. He describes how AI “has come a long way since 1951, when the was written by Christopher Strachey, whose checkers program completed a whole game on the Ferranti Mark I computer at the University of Manchester.” article first documented success of an AI computer program Thomas notes transportation, manufacturing, healthcare, education, media, and customer service as main benefactors of AI in the near future. These specific areas will be most affected in the immediate future by AI systems, career changes, and changing characteristics. According to De Mántaras, the road to truly intelligent AI “will continue to be long and difficult.” He quotes Gabriel García Márquez’s 1936 speech (“The Cataclysm of Damocles”): “Since the appearance of visible life on Earth, 380 million years had to elapse in order for a butterfly to learn how to fly, 180 million years to create a rose with no other commitment than to be beautiful, and four geological eras in order for us human beings to be able to sing better than birds, and to be able to die from love.”

An Academic Approach to the Future of AI

This paper is available on arxiv under CC 4.0 license. Authors: (1) Matu´s Medo, Department for BioMedical Research, Inselspital, Bern University Hospital, University of Bern, Department of Radiation Oncology, Inselspital, Bern University Hospital, University of Bern and matus.medo@unifr.ch; (2)  Michaela Medova, Department for BioMedical Research, Inselspital, Bern University Hospital, University of Bern, and Department of Radiation Oncology, Inselspital, Bern University Hospital, University of Bern. Table of Links Abstract & Introduction Results Discussion Materials and methods Acknowledgement & References Supporting Information 3. Discussion The broad range of tools for fitting mutational signatures makes it difficult to understand which tool to choose for a given project. In this work, we provide a comprehensive assessment of eleven different tools on synthetic mutational catalogs. Using mutational catalogs where only one signature is active allows us to quantify the differences in the fitting difficulty of individual signatures. We find that flat signatures whose average similarity with other signatures is high are the most difficult to fit. To assess fitting tools, we use synthetic mutational catalogs whose signature activities are modeled based on eight distinct cancer types. We find that when the number of mutations is small (100 mutations per sample), SigProfilerSingleSample is the best tool to use for all cancer types. As the number of mutations increases, SigProfilerAssignment becomes the best tool for all cancer types with mmsig best for some cancer types for an intermediate number of mutations (2,000 mutations per sample). The results change little when Pearson correlation is used instead of fitting error for the evaluation (Figs. S22 and S23 in SI). The risk of overfitting the data by including too many signatures in the reference catalog is often discussed in the literature (Maura et al., 2019, Koh et al., 2021, Degasperi et al., 2022). We show that three methods (SigProfilerSingleSample, SigProfilerExtractor, and mmsig) are robust to such overfitting when the number of mutations per sample is not too small (Figs. S12 and S13 in SI). Crucially, we find that the common practice of excluding signatures from the reference catalog because they do not seem to be active (or are little active) in the analyzed data is little effective or even harmful, unless the number of mutations per sample is small (Fig. S16 in SI). In most cases it is preferable to let the built-in statistical methods of these tools to decide using the complete COSMIC catalog as a reference. While our work gives clear recommendations on which tools to consider and which to avoid, many issues can be addressed in the future to further improve the fitting of mutational signatures. First, a similar assessment of fitting tools can be done for other types of signatures: doublet base substitutions, small insertions and deletions, and copy number variations. Then, some tools (e.g., sigfit and mmsig) can compute confidence intervals for their estimates. For sigfit, we used these estimates in the way recommended by the authors: When the lower bound of the confidence interval for the relative signature weight is below 0.01, the signature is marked as absent in the sample (its relative weight is set to zero). Our results show that this recommended practice indeed improves the results achieved by sigfit. By bootstrapping (resampling the original mutational catalogs with replacement), confidence intervals could be computed also for the tools that do not compute them by themselves. It would be worth investigating whether thus-determined confidence intervals could also improve the performance of other fitting tools. To fit mutational signatures, it is commonly required that each sample has at least 50 (Riva et al., 2020) or 200 (Blokzijl et al., 2018) single base substitutions. However, different signatures are differently difficult to fit (Figure 1), so it is unlikely that such a universal threshold is meaningful. Based on the simulation framework established here, it would be possible to study the minimum necessary number of mutations for different signatures of interest and different cancer types. To best match the needs of practitioners, it would be possible to extend the simulation framework so that it finds the best-performing fitting tool for a given cancer type, a list of signatures of interest, and a given distribution of sample mutation counts. When an active signature is missing from the reference catalog because it is new or is falsely deemed inactive in a given set of samples, most fitting tools cannot cope with this situation and distribute all (or nearly all) mutations among signatures from the reference set (Fig. S15 in SI). Three tools do not suffer from this drawback—deconstructSigs, signature.tools.lib, and sigfit—but their performance is generally weak. One could aim to enhance one of the bestperforming tools, such as SigProfilerAssignment, with an additional step that would quantify whether given signatures are indeed likely to produce given mutational catalogs. Another approach to this problem is offered in (Maura et al., 2019) where the authors suggest a multistep process to simultaneously avoid under- and over-fitting. Mutational signatures are first extracted de novo, then they are assigned to existing signatures from a reference catalog, and thusdetermined active signatures are finally used to fit the original input data. While there are well-defined statistical principles for the first and third steps of this process, the second assignment step is largely ad hoc. A more principled approach could improve the outcome. More generally, COSMIC signatures are the result of analyzing tumor tissues from many different organs, so they can be viewed as an average across them (Degasperi et al., 2022). It might be more appropriate to construct organ-specific reference catalogs that would better reflect mutational processes that occur in these biologically diverse systems. The fast-growing number of sequenced samples would make such an effort possible. This paper is available on under CC 4.0 license. Arxiv

Discussing Tools for Fitting Mutational Signatures: A Comprehensive Comparison

Did you know that 85.1% of people who use AI today apply it to writing articles and creating content? And these figures are set to increase even further. Let's start with some fascinating facts: However, the real question remains: does AI-generated content meet quality standards? Let's explore this further. The Backstory: My Own Challenges With AI Greetings, hackers! My name is Alex, and I'm the founder of Flipner. I'm excited to share with you the journey that led me to launch my AI copywriting startup. AI has become a big part of our daily tasks, helping with writing, managing social media, and more. ChatGPT is one of these tools, known for doing a lot. However, when it generates text, it's not always as rich or genuine as what a person might write. The words make sense and are organized well, but they can miss the deeper connection or unique touch that comes from human creativity. This kind of AI writing might be okay for quick, straightforward tasks, but it doesn't always engage readers on a deeper level. The content can end up feeling a bit bland and repetitive. Also, it's tough to plan content over time with today's AI tools, which focus on creating content instantly without the ability to delay or schedule. Imagine wanting to draft a post or article and having various notes or ideas you wish to compile. Right now, this means saving thoughts and links in different ways—some jot down notes on their phone, others might send voice messages on social networks, start a chat with themselves on WhatsApp, or even email drafts to themselves. This process is pretty scattered, leading to a bit of chaos when trying to organize thoughts. I've found myself caught in this mess, often losing track of my notes. My Initial Step: Beginning With Dialogue Initially, I set out to explore the idea further by talking to various people who regularly produce content, including bloggers, professional writers, copywriters, and authors of academic and scientific papers. Unsurprisingly, almost all of them highlighted how creating text is a demanding task, emphasizing that the essence of content lies in the thoughts and ideas of the author. This led to a realization: people are looking for more than just a content generation platform. They need a space to store, organize, and structure their ideas. : why not develop a platform that not only facilitates text generation but also offers a space to save notes, thoughts, and voice messages? A hub where content can be not just created but curated, with features allowing users to manage styles for their desired text format. This sparked an idea More critically, I envisioned a platform where AI plays a supportive, technical role—synthesizing the user's own thoughts and notes inputted into the system, rather than generating content on its own. This approach would ensure the content remains authentically the author's, differentiating it from current AI-generated texts by making it more human and reflective of the author's intent. Such a platform would transform AI into a tool that assembles author-driven content, preserving the unique voice and insights of the individual. Second Step: Understanding the Problem and Crafting a Solution If you were to ask journalists or professional writers today about their use of artificial intelligence in their work, about 80% would likely say they don't use it. The main reason? AI as it currently exists tends to produce content that lacks originality, often seen as unable to truly replace the human touch. The key to overcoming this challenge lies in repositioning AI as a supportive tool rather than a replacement, emphasizing its role in aiding humans to weave their unique thoughts into cohesive texts. This approach offers a new avenue for the journalism industry, where AI serves as an assistant that helps consolidate individual ideas into polished, ready-to-publish content. By leveraging such platforms, content creators can not only enhance the quality of their work but also streamline their workflow, saving precious time. This sparked the concept for Flipner AI, a portable assistant tailored to your content needs. Flipner is designed to simplify your workflow by allowing you to save and amalgamate your notes, transforming them into complete texts. It introduces a three-tiered approach to content creation: Firstly, Flipner serves as a central hub for content creators of all types, whether you're a blogger, a writer, or simply jotting down personal notes. You can add content to Flipner in the form of text notes or voice memos, with the latter being automatically converted from speech to text. This feature is especially handy for capturing your thoughts on the fly, enabling you to record any idea that comes to mind, no matter where you are. The second tier focuses on the organization and enhancement of your collected notes. Flipner not only stores but also helps you merge and refine these notes, offering functionalities for easy editing and translation into different languages as needed. On the third level, Flipner takes your accumulated notes for a specific project—be it a social media update, an article, or academic work—and artfully assembles them into a cohesive piece. It rephrases your notes while maintaining the original intent, presenting your thoughts in a polished and structured format. In essence, Flipner AI aims to be more than just a tool; it seeks to be a companion in your creative process, ensuring that your original voice and ideas are not just preserved but showcased in the best possible light. Third Step: Building a Functional MVP To create an MVP that was both comprehensive and approachable, I focused on including features that would make the beta version stand out, based on real-world needs and industry insights: A user-friendly interface that's available whenever and wherever you need it, ensuring accessibility and ease of use. Style management options that accommodate a variety of content requirements, allowing users to tailor their output to different audiences and formats. The ability to edit generated text directly within the system, removing the hassle of having to use external tools for modifications. Support for multiple languages, opening up the platform for international use and facilitating global content creation. An automatic feature that converts voice recordings into text, streamlining the note-taking process and making it more efficient. Fourth Step: Launching and Testing the Market We decided to launch Flipner's first public version on Product Hunt for two main reasons: the platform hosts a significant portion of our target audience, and it offers a chance for initial traction and user feedback. The preparation phase lasted around three months, balancing marketing strategies and technical development simultaneously. The effort paid off spectacularly when Flipner AI was named the on Product Hunt. #1 Product of the Day and the #2 Product of the Week Moreover, we gathered over 30 suggestions for new features and improvements from the feedback. This success served as a strong validation for my initial concept, motivating me to further refine and evolve Flipner. I am committed to continuous development, aiming to revolutionize text creation with AI assistance. Keep an eye on my progress with #buildinpublic on and for the latest updates. In the meantime, I invite you to try out Flipner AI for yourself at no cost by following this ! X (formerly Twitter) LinkedIn link

How I Faced Challenges in AI Writing & Founded My Copywriting Startup that Became #1 on Product Hunt

আমি কীভাবে এআই রাইটিংয়ে চ্যালেঞ্জের মুখোমুখি হয়েছি এবং আমার কপিরাইটিং স্টার্টআপ প্রতিষ্ঠা করেছি যেটি প্রোডাক্ট হান্টে #1 হয়েছে

Wie ich mich den Herausforderungen beim KI-Schreiben stellte und mein Copywriting-Startup gründete, das zur Nummer 1 bei der Produktsuche wurde

Cómo enfrenté los desafíos en la redacción de IA y fundé mi startup de redacción publicitaria que se convirtió en la número 1 en la búsqueda de productos

Comment j'ai relevé les défis liés à la rédaction d'IA et fondé ma startup de rédaction qui est devenue n°1 sur Product Hunt

कैसे मैंने एआई राइटिंग में चुनौतियों का सामना किया और अपना कॉपी राइटिंग स्टार्टअप स्थापित किया जो प्रोडक्ट हंट में #1 बन गया

AI ライティングの課題にどのように立ち向かい、プロダクト ハントでナンバー 1 になったコピーライティングのスタートアップを設立したか

AI 글쓰기의 어려움에 직면하고 제품 검색에서 1위가 된 카피라이팅 스타트업을 설립한 방법

Como enfrentei desafios na redação de IA e fundei minha startup de redação que se tornou número 1 em busca de produtos

Как я столкнулся с проблемами в написании текстов с использованием искусственного интеллекта и основал свой стартап по копирайтингу, который стал номером 1 на Product Hunt

Yapay Zeka Yazımında Zorluklarla Nasıl Karşılaştım ve Ürün Avında 1 Numara Olan Metin Yazarlığı Girişimimi Nasıl Kurdum?

Làm thế nào tôi đối mặt với những thách thức trong việc viết AI và thành lập công ty khởi nghiệp viết quảng cáo của tôi và trở thành số 1 trong cuộc săn tìm sản phẩm

我如何面对人工智能写作的挑战并创立我的文案初创公司，该公司在 Product Hunt 中排名第一

Hey there, fellow tech enthusiasts! My name is and today, I want to share with you some insides of mine. As a seasoned developer who's seen the ebb and flow of the digital tide, let's dive into the swirling vortex of what the future might hold for us in 2024. Grab your digital crystal ball, folks, because here come some spicy predictions! Alex Cloudstar, The Era of Augmented Reality 🌐 Picture this: virtual worlds blending seamlessly with our reality. That’s right, folks, AR is about to take a massive leap in 2024. From enhancing customer experiences in retail to transforming education, this tech is set to reshape how we interact with the world. Get ready for a mix of Pokémon GO-like experiences and practical applications that’ll blow your socks off! But I can’t say that I am a fan of this. Just like with other forms of technology, there's a risk of over-reliance and addiction. People might become so engrossed in AR experiences that they neglect real-world responsibilities or become detached from reality. While AR can create immersive experiences, it might also lead to further social detachment. If individuals start preferring virtual interactions over real-world interactions, it could potentially worsen issues related to social isolation and decrease physical human contact. AI, AI, AI Everywhere 🤖 Artificial Intelligence is no longer the next big thing; it's the now big thing. In 2024, we'll see AI continue to weave itself into our daily lives, from predictive analytics guiding business decisions to AI-powered assistants handling more complex tasks. But fear not, humans, for we shall remain the masters behind the machines! Quantum Computing: Unlocking the Unthinkable 🔐 Hold onto your hats, folks, because quantum computing is inching closer to reality in 2024. We're talking about mind-boggling computational power that could revolutionize industries, from healthcare to finance, solving problems that were once deemed unsolvable. It's like upgrading from a bicycle to a warp-speed spaceship! Sustainable Tech 🌱 With great tech comes great responsibility. In 2024, the tech world is taking a giant leap toward sustainability. From eco-friendly data centers to renewable energy-powered devices, the focus on reducing our digital carbon footprint is becoming more crucial than ever. Agrotech and Food Sustainability are essential. Technologies focusing on sustainable agriculture practices, precision farming, and resource-efficient techniques like vertical farming or smart irrigation play a crucial role in ensuring food security while reducing environmental impact. The Rise of Privacy Tech 🔒 Privacy concerns have been making headlines for years, and in 2024, they're taking center stage. As users become more aware of their data rights, we'll witness a surge in privacy-focused tech. Encryption, decentralized systems, and user-centric platforms will shine as guardians of our digital identities. Privacy is the new black, my friends! Now, you might be wondering, "How do I prepare for this tech revolution?" Well, fear not, intrepid developers, for I've got a few tips up my sleeve: Embrace change like a long-lost friend. Keep learning, exploring new technologies, and adapting like a chameleon in a digital jungle. Stay Agile, Stay Relevant: Engage with the tech community. Attend conferences, join forums, and collaborate on projects. The collective wisdom of the crowd is a treasure trove of insights. Community is Key: With great tech comes great vulnerabilities. Prioritize security in your code and stay updated on the latest cybersecurity trends. Focus on Security: As tech evolves, consider the ethical implications of your creations. Build responsibly, with an eye toward the greater good. Ethics Matter: The future is unwritten. Embrace uncertainty, for it’s where innovation thrives. Embrace the Unknown: So, dear readers, fasten your seatbelts, charge your devices, and get ready for a rollercoaster ride through the tech-scape of 2024. It's a thrilling time to be a part of this ever-evolving industry! Until next time, happy coding, and may your bugs be minimal and your breakthroughs monumental!

WEAV AI

weav.ai

-- employees

Since n.d.

COMPANY RANKING

WEAV AI

EVERGREEN INDEX #11775

Representing this company? Click here to claim and customize this page!

Weav AI WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

WEAV AI

weav.ai

-- employees

Since n.d.

COMPANY RANKING

WEAV AI

EVERGREEN INDEX #11775

RELATED COMPANIESRANK

STORIES ABOUT WEAV AI

Representing this company? Click here to claim and customize this page!

Weav AI WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!