Mike Sokol’s Post

We need the right people in the drivers’ seats as AI speeds ahead. Want to hear some big AI news? ⬇️ Evidently the “fastest AI chip of all time” (@500,000 t/s) is being released by Etched called ‘Sohu’, allowing products to be built that wouldn’t be possible with GPUs (like NVIDIA’s). These specialized chips are the first ASIC (application specific integrated circuit; or chips built to solve a single hash algorithm) for transformer models, which is a BIG DEAL. Transformers are the “T” in ChatGPT. Etched claims that ONE 8xSohu server replaces 160xH100s (NVIDIA) which is squarely (pun intended) between squared & cubed — an order & a half of magnitude (OOM^2.5). As we need more & more compute but need it faster, more energy efficient, cheaper, more accurate etc. Speciailized Chips are the solution to this problem, & companies like etched & Groq are setting the bar incredibly high. I recently posted about Groq, as well as a radically different approach to this solution, “bio-computers” here: https://lnkd.in/dxtZ2KzM. The other big news is that there’s a new “best AI (llm) available publically, for free” which is Anthropic Claude 3.5 Sonnet* - & a TOTALLY ACCURATE CLAIM. Although it doesn’t create images or video nor have internet connectivity, it is so fast, accurate, original, & is incredible at coding — even things you don’t know needed to be or could be coded — & shows its work, reasoning & much more. Additionally, I’ve got it to make “zero-shot” (no tweaks needed) coding & deployment of the games Snake & Tetris, Mancala, Tic-Tac-Toe, as well as FPS (first-person-shooter) games, which is incredible. It can create simple presentations, add simple graphics, & will have agentic capabilities as well as compatible with MS copilot. Candidly, I think Claude is my new go-to AI, although I’ll still be using ChatGPT & Gemini daily, & Perplexity yet too. Also, aside from the absence of more multimodality, this is as good as I’d hoped ChatGPT 5 would be. It’s a literal GAME CHANGER (& I don’t say that often or lightly). Lastly, this video shows a Tesla getting lapped by a Tahoe, & I’m not a fan of f/elon or most of his projects, but their AI arm, xAI just raised a few billion dollars, so perhaps we will finally see something from them sooner or later. Want to stay ahead of & on top of all things AI including how to use & leverage it? Check out Tech For Good (free) here: https://lnkd.in/d6AT-8fG. Thanks for checking this out. Please leave your thoughts & perspectives in the comments below. * hopefully a Claude 3.5 Opus soon too! ** Claude 3.5 Sonnet is only BEST when you enable Artifacts; also it’s FREE! 🎥 VC: @tberg.media IG #techforgood #ai #ainews #coryconnects #agi #asi #llms #gpus #asic #gai #innovation #tech4good #lpus #sohu

Over the last year, interest in artificial intelligence has boomed. Generative artificial intelligence was one of the most disruptive technologies in 2023, thanks OpenAI’s ChatGPT which took the world by storm.⚡ Powered by Nvidia’s powerful graphic processing units (GPUs), demand for smart solutions, particularly AI, has risen exponentially. Despite the low downturn in consumer #electronics and the growth in excess electronic component inventory, tech giants worldwide sought to make their own ChatGPT golden goose. Learn how AI will fuel 2024 technology trends: https://hubs.li/Q02dfJTP0 #Sourcengine #semiconductors #AIboom #disruptivetechnologies

GenAI? ChatGPT? OpenAI? Nvidia? GPUs? Hallucinations??? Welcome to the alphabet soup for today's Marketers and leaders. I had the pleasure of joining Yogesh Chavda to just scratch the surface of AI in Marketing. Here's just a snippet below: https://lnkd.in/g5QizCZX Or listen to the entire session here: https://lnkd.in/gSut9ksA Let me know what you're learning and how you're staying up to date. #aiandbusiness #marketing2024 #cmoinsights

What was on my mind is now AI taking over. What I used to love about writing was my confidence and experience of resources/sources. We no longer know the sources of AI except that it's a hodgepodge of dissimilar thoughts on the web. Since Chatgpt launched on the web on November 30th, 2022 it has accelerated its use, and 2 weeks ago onto Linkedin to write your thoughts. NVIDIA's new models and model architectures are introduced to developers to optimize their models with the latest NVIDIA has to offer. The cutting-edge implementation is masked muti-head attention for the context and generation phases of GPT model execution. This is going to take time and hopefully find our new way of working. The ecosystem is innovating rapidly as larger models unleash new capabilities and uses. Some of the largest, most advanced language models, require multiple GPUs working in concert to deliver responses in real-time. A type of model parallelism in which matrices are split across devices enables efficient inference at scale-with each model running in parallel across multiple GPUs connected across multiple servers without developer intervention or model changes. WELCOME TO THE BRAVE NEW WORLD...

Zuckerberg has gone ultra intense on human level ai development, buying up 340k of Nvidia's 500k H100 gpus. ●Around 20billion worth of gpu https://lnkd.in/epyT8UAT ●Zuck's own words in a talk: https://lnkd.in/eTMQq-SP ♡For those who may be unaware, human level ai/agi is often cited as the last invention humanity would need to make. Google, Microsoft etc are all racing to create human level ai. ChatGpt is an example of "emergingAgi" as cited in prior papers. ●I wrote about this happening in my 2018 and 2019 local newspaper gleaner articles: [2018, agi discussion] https://lnkd.in/gbWp4irN [2019, mass automation - Why Ja needs a minister of Ai] https://lnkd.in/es8fMTE

Announcement of AI benchmark test results with new additions #MLPerfInference 🤝 Follow us on Discord 🔜: https://lnkd.in/gt823Zd3 🤝 Follow us on Whatsapp 🔜 https://wapia.in/wabeta _ ❇️ Summary: MLCommons has introduced the MLPerf Inference v4.0 benchmark test to measure AI processing performance of hardware, including Meta's Llama 2 70B language model and Stable Diffusion XL image generation AI. The benchmark highlights the rapid growth of generative AI models, with NVIDIA leading in performance. More than 8,500 results were submitted by 23 organizations, with a focus on Edge and Datacenter systems. The top performer in generative AI was an NVIDIA system, showcasing significant improvements in performance compared to previous models. Hashtags: #chatGPT #AIbenchmark #MLPerfInference

Since the public release of ChatGPT last November, there is a heightening interest from enterprises and individuals in Generative AI, as it is expected to significantly change how we work and play in the imminent future.    In fact, analysts expect this transformative new technology to contribute more than $15 trillion to global economy by 2030. Business leaders adopt #GenAI as they acknowledge the tangible benefits it can provide them to:   👉 Accelerate the time to results and increase value 👉 Automate the processes and scale their growth 👉 Leverage the real value of their data for actionable insights   Practically all sectors of the economy can benefit from GenAI. Read in the following article how industries such as #Banking, #Telecommunications and #PublicSector plan to use GenAI to optimize their #efficiencies. #DellTechCEE #iWork4Dell

There is something new under the sun of AI. Nvidia has just released Chat with RTX: an AI chatbot that runs locally on your local PC It searches and analyzes our PC's files, including text documents, PDFs, and even YouTube videos. Differently from the cloud-based approach (ChatGPT et al), this chatbot runs locally on the user's computer, providing quick access to personalized information while, by design, keeps the data private. This is game changer and a master move from a company which, at a market cap of 1.8 Trillion USD, already outshines the competition (10x Intel's market cap). #ai #aichatbots #aiadvancements #aiassistant

Groq: The Next-Gen AI Platform Revolutionizing Performance! Embark on a journey into the future of AI with Groq, where cutting-edge technology meets lightning-fast performance. Our innovative Language Processing Unit (LPU), crafted by former Google engineers, is redefining the possibilities of AI processing. While established players like ChatGPT rely on traditional Nvidia GPUs, Groq's LPU sets a new standard. Say goodbye to sluggish processing speeds of 30–60 tokens per second. With Groq, experience an exhilarating 8–10 fold increase in performance, reaching an astounding 500 tokens per second for smaller models and an impressive 250 for larger ones. #AI #Groq #Innovation #LLM