Igor Perisic

Congrats VERSES (OTCQB:VRSSF) Gabriel René and Dan Mapes and Dr. Karl Friston for introducing an "efficient alternative to deep learning, reinforcement learning and generative AI called Renormalizing Generative Models (RGMs) that address foundational problems in artificial intelligence (AI), namely versatility, efficiency, explainability and accuracy, using a physics based approach." #ai #activeinference #RGM #spatialweb #AGI #cognitivecomputing https://lnkd.in/gcYSM4eN

23

I just released the April edition of my AI & the Web newsletter. Nathaniel Simard, founder & CEO of Tracel Technologies and inventor of the burn ML framework, spoke with me about how he got into ML, burn, Tracel and the future of AI. WebGPU, the successor of WebGL, will soon also be supported in Firefox and Safari. Small, powerful LLMs running right in the browser is not a distant future, but are the next big thing in cost-effective and privacy-friendly AI. Google is working hard on better WASM support in the browser that makes ML models running fast even on low-end devices without GPU. With Llama 3 by AI at Meta and Phi 3 by Microsoft AI we got some exciting new, small modells. Llama 3 was already ported to run in the browser and we'll see Phi 3 soon as well. MusicGen has been ported to the browser in transformers.js by Hugging Face engineers and developed an online Jukebox running right in the browser. #webgpu #webml #ai #ml #llama3 #wasm https://lnkd.in/drJCY7Ew

19

2 Comments

I co-hosted an AI engineering leaders dinner on Thursday with Grace Ge. Thank you to Beyang Liu from Sourcegraph, Devendra Chaplot from Mistral AI, Bryan Bischof from Hex, Marco Sanvido from Robust Intelligence, and Tim Shi from Cresta for joining us. Here are the a few key takeaways: 1️⃣ AI hardware has a long way to go, but folks in the community are generally excited about how it can influence our day-to-day. Right now, they're toys showing us glimpses into the next generation of consumer tech. I respect rabbit inc. / Humane for starting to make progress in this area, but I think there's a lot more to come 😁 2️⃣ We're at an interesting crossroads between engineering and research in the AI community. There's a lot of respect on *both* sides, but I find that these conversations often pendulum swing between specific research areas and very pragmatic scaling/tuning insights. This paper on Chinchilla by Rohan P. came up specifically as important new work: https://lnkd.in/gD6tmDRv 3️⃣ Everyone feels that AI can help them get out of the busy work, and help focus them on the high-leverage parts of their jobs. There does seem to be a tension, though: those solutions are still "not quite good enough" but the ones that make for cool demos look like problem-seeking solutions. The products mentioned were Granola and Agenda Hero. 4️⃣ Large categories of companies within AI are already being commoditized quickly. There was a lot of talk about which companies are taking the "smart bets" in some of these competitive categories, but it feels to me like these 5-7 year strategies will start paying dividends and creating category kings within the next year or so. Specific excitement around players like turbopuffer. 5️⃣ Data privacy remains a hot topic. It's not clear if the incentives line up to push in the privacy direction for most large companies, and I think a lot of R&D cost is being driven by what AI talent *wants to work on* rather than productionizing the whole ecosystem. Thanks again to Grace Ge for the wonderful venue and Sunil Dhaliwal + the Amplify Partners team for co-hosting with Orb!

73

7 Comments

From Azeem Azhar on Eric Schmidt’s recent talk at Stanford: “Agent-based systems, which autonomously execute multi-step tasks and adapt to environmental feedback, are poised to unlock significant economic value next. Building on this foundation, text-to-action capabilities will emerge, translating natural language into specific, working programs. Imagine your own personal programmer at your fingertips, ready to create whatever you envision. A dating app that matches people based on their Netflix viewing history and Spotify playlists? A prototype within minutes. These forces will converge. While he doesn’t say it explicitly, it seems that Eric expects the next two years to be faster and more turbulent than the previous two.” https://lnkd.in/gxWEy5Pf

8

Worth reading: How the Apache Arrow Format Accelerates Query Result Transfer by Ian Cook One of the hidden problems of intersystem communication is transcription of data formats. Ian does a wonderful job of explaining the "why" of Apache Arrow as well as the "how" of its advantages. Read on...

8

While I agree with the IMF on some of their concerns -- again need to differentiate LLMs from all AI, folks (institutions need training), I disagree with them a bit on their proposed solution. First of all, like most others, they are assuming LLMs will dominate when the evidence actually appears that they may not even survive. However, robotics, SLMs, neurosymbolic and other methods also provide risks to the labor force -- depending on how they are designed and deployed. The greatest risk for inequality isn't from job displacement, but rather a continuation of the recent sharp trend towards oligarchy in the U.S., unfettered expansion of the Big Tech oligopoly, and consolidation of wealth and power in other markets. I've long supported an increase in taxes on the ultra-wealthy as part of an overall plan for fiscal discipline, but the IMF suggestion would simply expand government. From a purely economic perspective, I'd focus on the actual results of ever-increasing government here in the U.S. As the government expanded, it also became more corrupt, and the wealth gap has expanded. The tyranny of good intentions results in part from gatekeepers galore during the process of wealth redistribution. By the time every stakeholder in the lobbying system is satisfied, wealth typically consolidates in the hands of those skimming off the top. I would urge the IMF and policymakers to rethink their position with an eye towards prevention. If for example we adopt responsible and safe AI systems in a manner that diversifies jobs and wealth rather than consolidates, that would be far more beneficial. Don't surrender to Big Tech dominance and oligarchs.

5

1 Comment

➰ APIContext's latest insights on cloud-based API performance have been featured on Nordic APIs! In the 2024 report, we delve into crucial aspects affecting API efficiency: 🔐 Authentication is key. Ensuring robust authentication mechanisms in APIs is crucial, yet offloading this function can introduce complex challenges. 📖 DNS configurations matter. How you handle DNS settings can significantly impact API performance, particularly when tied to authentication processes. 🌐 The role of third parties. Critical third-party services play a significant role in 🔗 API transactions, underscoring the need for careful integration and monitoring. ⛅ Not all clouds are created equal. Differences in cloud services can drastically affect API performance, an essential consideration as APIs account for over 80% of all internet traffic. #APIs #CloudComputing #TechInsights #APIContext

15

1 Comment

TRIPLE BAM! If you missed it, here's the recording of our live Q/A with Jay Alammar and Joshua Starmer PhD! We talked about many AI topics, answered lots of great questions, and talked about two great books: Hands-On Large Language Models: Language Understanding and Generation, and The Illustrated Guide to Neural Networks and AI!!! https://lnkd.in/dcPHs7YQ

44

1 Comment

💡 Last Thursday with Remi Takase, we had the opportunity to attend the #dotAI conference, and here are my main takeaways of the 1st Day: ✨ Stanislas Polu - Dust: AI is in a phase transition with profound societal impact. After the first phase, "Can See" (2012-2017), where CNN models accelerated tasks in seconds, we entered the "Can Generate" phase in 2017, with LLMs saving hours of work. Now, we're on the verge of a new paradigm: AI that "Can Think." ⚙️ Ines Montani - Explosion / Spacy / Prodigy : From Software 1.0 to Software 2.0… it's time to think beyond chatbots! LLMs are not just about text/image interactions for your users. You need to dive deep into the experience, understand the problem before finding the solution. Like the clock replaced the Knocking Machine! Thanks for the Knocking Machine test, I will keep this one! (deck link in comment ⬇) 🛠️ Thank you Merve Noyan for the deep dive into the open-source world with Hugging Face and how open-source models can catch up with larger proprietary models. Hugging Face demo was insightful! 🏦 Pierre Stock from Mistral AI introduced #Ministral, the new Edge model. Exciting perspectives ahead, especially with its focus on agents, function calling, and innovative approaches to compressing memory footprint and managing transformer attention on mobile devices. 🚀 Yann Leger - Koyeb on stage: "AI represents 28% of data center consumption in the US.(...) In the next four years, this consumption will increase by 500%!" 💻 Arthur Talpaert and Laure Seugé announced the release of Doctolib 's Consultation Assistant that suggests a medical summary based on voice audio from consultations and fills in the patient file directly in Doctolib software. How to empower doctors with technology to renew their interaction with patients, instead of focusing on typing into a computer 🛠️ A live demo by Katia Gil Guzman and Romain Huet from @OpenAI showcased the power of their real-time API when used with tool functions (Video Below ⬇ ) 🎉 Congratulations to Nessrine Berrama for bringing together such brilliant speakers and to Claire Lebarz for her flawless hosting throughout the event! #dotAI2024 #dotAI #AI dotConferences

77

6 Comments

Switch to Claude 3.5 Sonnet on AWS Bedrock for the most versatile, innovative, cost effective, and performant LLM model out there. With this model beating almost all benchmarks, and users world over going crazy over its mind boggling coding and visual abilities, it is a no-brainer at this point. "Claude 3.5 Sonnet, the latest and most capable model from artificial intelligence (AI) safety and research company Anthropic, is generally available today in Amazon Bedrock as part of Anthropic’s Claude family of AI models. Amazon Bedrock offers the broadest selection of high-performing foundation models (FMs) from leading AI companies, along with the capabilities and enterprise security customers need to quickly build and deploy generative AI applications. Data from Anthropic shows that Claude 3.5 Sonnet sets a new industry standard for intelligence. According to Anthropic, the new model outperforms other models available today,including OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro, as well as Anthropic’s own previously most capable model, Claude 3 Opus, in areas including expert knowledge, coding, and complex reasoning. Claude 3.5 Sonnet offers better intelligence and speed than Opus at one-fifth of the price. View Claude 3.5 Sonnet benchmarks."

7

1 Comment

In a recent podcast, Mike Dillinger, PhD and Larry Swanson make an excellent point: at least for the near term, LLMs need "adult supervision" to truly excel in specialized domains. Dillinger highlights how knowledge graphs can provide context to LLMs, acting as "jet fuel" for training. This approach is crucial for regulated industries where accuracy and context are especially important, but building a strong foundation with high-utility, compliant data at the onset is necessary. With our work at Integral, we put this concept into action with an automation that preserves analytic fidelity and compliance: 1. Accelerating insights with automated compliance for regulated data 2. Enhancing, customizing, and remediating data - enabling companies to build iterative, impactful, LLMs. Looking to accelerate or experiment with LLM development with context-rich data? Let's connect: useintegral.com #AI #KnowledgeGraphs #RegulatedData #PrivacyFirst

19

3 Comments

Recently I interviewed Nathaniel Simard, founder & CEO of Tracel Technologies and inventor of the burn ML framework: Jan: Nathaniel, we're excited to have you as our guest today! To kick off our conversation, could you please introduce yourself and share a bit about your background and what got you into ML? Nathaniel: Sure, I'm the creator of Burn, a deep learning framework written in Rust, and the founder of Tracel AI. I started coding in the first year of university, where I was studying mechanical engineering, but I quickly switched to software engineering after, since I instantly fell in love with programming. Then I explored different facets of the field from backend to frontend development, and I decided to start my career as a consultant focused on software quality. After some time, I wanted to go deeper into AI, since I was always interested in the process of learning, so I enrolled for a master's degree at MILA. J: According to Github you started to work on burn in summer 2022. Since then has already earned over 7000 stars on GH. What was your initial motivation to develop a new ML framework? N: I always had a side project going on, for fun mostly and to learn new things. I wanted to explore asynchronous and sparse neural network architectures, where each sub-network can learn and interact with other sub-networks asynchronously and independently. I wasn't able to actually create something useful because I needed fine control over the gradients and the concurrency primitives, which is not easily done with Python and PyTorch. At the same time, I was working on machine translation models at my current job, and it was quite painful to put models into production. I decided to switch my side project to a new deep learning framework, with more flexibility regarding gradients and concurrency primitives as well as being more reliable and easier to deploy on any system. J: Amazing to see that this was born out of a side project! I feel the struggle with concurrency in Python and this is something where Rust really shines. What are some of the other key features that makes burn special? N: I think there are two things that really set Burn apart. First, almost all neural network structures are generic over the backend. The goal is that you can ship your model with almost no dependency, and anybody can run it on their hardware with the most appropriate backend, even embedded devices without an operating system. Second, Burn really tries to push the boundaries of what is possible in terms of performance and flexibility. It offers a fully eager API, but also operation fusion and other optimizations that are normally only found in static graph frameworks. The objective is that you don't have to choose between portability, flexibility, and performance; you can have it all! To read the rest of the interview with Nathaniel's plans for burn & trends in AI here: https://lnkd.in/dKv9ztMm #ml #ai

3

The cost of training top models is increasing 2-3X per year, and at this rate will reach $1 billion per model by 2027. The findings are from Epoch AI and follow their recent analysis of increases in compute for foundation models running at 4-5X per year. They've also provided some interesting data on the relative breakdown of cost components like staff, hardware, and energy. A few thoughts after reading this: - I would assume staff costs won't increase as much as hardware and energy costs, so the relative contribution of human labor to future AI models will decrease, in yet another sign of capital's increasing dominance over labor in the industry (just look at NVIDIA's ever-rising stock price). - If compute is increasing 4-5X per year but cost is increasing 2-3X, is it safe to assume that training runs are getting more efficient? - Some people, like Anthropic CEO Dario Amodei, have said we'll see $1 billion training runs in the next year. Does this mean the 2-3X pace is about to increase? That Epoch is wrong in its estimate? Or that Amodei is overstating the cost, possibly to put fear into competitors who can't keep pace? - Only a few companies are going to be able to train future models. We're likely going to see further consolidation. Perhaps we get to a future with only two large general purpose models, like the Windows/MacOS and Android/iOS duopolies we've seen in the past. Article here: https://lnkd.in/g5SejXen

7

🧑‍💻 Hands-on Workshop: Building Enterprise-Grade RAG with AWS Bedrock Agents and SingleStore ➡️ 𝐀𝐖𝐒 𝐁𝐞𝐝𝐫𝐨𝐜𝐤: - As AI continues to revolutionize industries, 𝐀𝐖𝐒 𝐁𝐞𝐝𝐫𝐨𝐜𝐤 stands out as a key player in the AI ecosystem, offering a robust framework to deploy, scale, and optimize machine learning models effortlessly. AWS Bedrock enables developers to seamlessly integrate foundation models, making AI more accessible for enterprise applications and cutting-edge innovations. ➡️ 𝐀𝐖𝐒 𝐁𝐞𝐝𝐫𝐨𝐜𝐤 𝐀𝐠𝐞𝐧𝐭𝐬: - A new and powerful feature within Bedrock is 𝐁𝐞𝐝𝐫𝐨𝐜𝐤 𝐀𝐠𝐞𝐧𝐭𝐬—designed to automate workflows and reduce repetitive tasks, Bedrock Agents streamline business processes, enhancing productivity while driving down operational costs. These agents go beyond automation: they empower businesses with Retrieval-Augmented Generation (RAG) capabilities. Imagine a system that connects to your company's data sources and augments user requests with the exact information needed to generate accurate, context-aware responses. Whether it’s answering complex customer queries or ensuring compliance in claims management, Bedrock Agents have you covered. - Want to bring enterprise-grade RAG to your company? 𝐁𝐞𝐝𝐫𝐨𝐜𝐤 𝐀𝐠𝐞𝐧𝐭𝐬 offer the most efficient way to set up a scalable and secure RAG pipeline. For instance, if a user inquires about claims eligibility, a Bedrock Agent can securely access your knowledge base and return an accurate response like: “You need to turn in your driver’s license, pictures of the damaged car, and an accident report.” ➡️ 𝐔𝐩𝐜𝐨𝐦𝐢𝐧𝐠 𝐖𝐨𝐫𝐤𝐬𝐡𝐨𝐩: • Learn more in this workshop on Tuesday, October 15 @ 10-11am PT / 10:30-11:30pm IST where experts from AWS and SingleStore will walk you through the process of building enterprise-grade RAG systems using Amazon Bedrock Agents and SingleStore. This live session includes practical strategies, real-world examples, and a demo to set up and deploy RAG pipelines that handle data complexity and performance needs. ✅ The workshop will include a live demo and code-share session. If you've registered but can't attend it live, you'll get the code and webinar recording in your inbox. 🔗 Register here: https://lnkd.in/g33DVcwW #AWS #LLMs #RAG

23

2 Comments

101 real-world gen AI use cases featured at Google Cloud Next ’24 by @google snapshot of how 101 of these industry leaders are putting AI into production today, creating real-world profitable use cases! #rafAIgual #GenAI #usecases #enterprise #adoption

2

A must-read! Thanks, Tim Sanders for the recommendation and the book itself! Not just due to their Canada-connection, Ajay Agrawal, Joshua Gans, and Avi Goldfarb packed this ~200-page book with not just facts and figures, but frameworks and fascinating, fantastic food for thought. One key takeaway for me is to cut the use of 'insights' from my vocabulary when it comes to the impact that #AI brings to businesses (insights on the tech and trends are still great though, heck our team is called the Research Insights team). When we think about the power that AI is bringing to enterprises, it is all about its impact on 𝕡𝕣𝕖𝕕𝕚𝕔𝕥𝕚𝕠𝕟𝕤 and 𝕕𝕖𝕔𝕚𝕤𝕚𝕠𝕟𝕤. Saying that it helps businesses generate or gain insights is too surface level and not getting to the core of AI's capabilities. Business leaders, take note: focus on what predictions your businesses rely on and then figure out if, how, and when AI can come and improve them.

30

4 Comments

🚀 Join us at #KubeCon and be part of #Kubernetes in its 10th year! 🥳 Today is the last day to receive preferential pricing ahead of the onsite rate. Register here: https://lnkd.in/g3KttPYW ❓ What's the big deal about Kubernetes? Our latest research report describes the incredible impact of this #opensource project in transforming the world of #cloudnative computing across enterprises - and the role it plays in the lives of the amazing people who sustain it has been just as significant, captured in dozens of respondent "stories." Download Kubernetes Turns 10 report today, and see you at KubeCon! https://lnkd.in/gVpY6wWB

26