Ari Kamlani

Thank you @stephzhan for the chat and @sequoia for hosting, pleasure to come by!

With many 🧩 dropping recently, a more complete picture is emerging of LLMs not as a chatbot, but the kernel process of a new Operating System. E.g. today it orchestrates: - Input & Output across modalities (text, audio, vision) - Code interpreter, ability to write & run programs - Browser / internet access - Embeddings database for files and internal memory storage & retrieval A lot of computing concepts carry over. Currently we have single-threaded execution running at ~10Hz (tok/s) and enjoy looking at the assembly-level execution traces stream by. Concepts from computer security carry over, with attacks, defenses and emerging vulnerabilities. I also like the nearest neighbor analogy of "Operating System" because the industry is starting to shape up similar: Windows, OS X, and Linux <-> GPT, PaLM, Claude, and Llama/Mistral(?:)). An OS comes with default apps but has an app store. Most apps can be adapted to multiple platforms. TLDR looking at LLMs as chatbots is the same as looking at early computers as calculators. We're seeing an emergence of a whole new computing paradigm, and it is very early.

Claude 2.1 with 200k context just got released by @AnthropicAI with 40% price decrease. Making it cheaper than @OpenAI GPT-4 turbo. 🤑 I updated the LLM token-based pricing sheet. 👉 #gid=0" target="_blank" rel="nofollow noopener">docs.google.com/spreadsheets/d…

Invest the time early on to discard good ideas in pursuit of a great one. Take -1 to 0 seriously so you can increase your chances of going 0 to 1. Read more on what I mean here ⬇️ blog.southparkcommons.com/what-is-negati…

Please help Kindra reach the fundraising goals for the 2023 TCS New York City Marathon Fundraising! haku.ly/7e1240878d

A tremendous job from the @Titans social media team.

Oops haven't tweeted too much recently; I'm mostly watching with interest the open source LLM ecosystem experiencing early signs of a cambrian explosion. Roughly speaking the story as of now: 1. Pretraining LLM base models remains very expensive. Think: supercomputer + months. 2. But finetuning LLMs is turning out to be very cheap and effective due to recent PEFT (parameter efficient training) techniques that work surprisingly well, e.g. LoRA / LLaMA-Adapter, and other awesome work, e.g. low precision as in bitsandbytes library. Think: few GPUs + day, even for very large models. 3. Therefore, the cambrian explosion, which requires wide reach and a lot of experimentation, is quite tractable due to (2), but only conditioned on (1). 4. The de facto OG release of (1) was Facebook's sorry Meta's LLaMA release - a very well executed high quality series of models from 7B all the way to 65B, trained nice and long, correctly ignoring the "Chinchilla trap". But LLaMA weights are research-only, been locked down behind forms, but have also awkwardly leaked all over the place... it's a bit messy. 5. In absence of an available and permissive (1), (2) cannot fully proceed. So there are a number of efforts on (1), under the banner "LLaMA but actually open", with e.g. current models from @togethercompute, @MosaicML ~matching the performance of the smallest (7B) LLaMA model, and @AiEleuther , @StabilityAI nearby. For now, things are moving along (e.g. see the 10 chat finetuned models released last ~week, and projects like llama.cpp and friends) but a bit awkwardly due to LLaMA weights being open but not really but still. And most interestingly, a lot of questions of intuition remain to be resolved, e.g. especially around how well finetuned model work in practice, even at smaller scales.

Senior AI Architect, Ari Kamlani is featured in Reworked discussing why responsible AI should be on the agenda of every enterprise. Read the article to learn more about the public expectations of AI and the ethics of responsible AI. bit.ly/3Hzvcmb

I heard about this Dashboard Design Patterns paper, and man, it's cool. One of the hardest parts of gathering requirements for anything is getting two people to imagine the same end outcome. I'm now realizing how often that's true for dashboards arxiv.org/pdf/2205.00757…

If only I had access to this data 😜

@SeifertESPN I assume none of this data is available to the public, doh!!!

Data compiled for the NFL and NFLPA shows that injury rates on non-contact injuries to lower extremities are statistically the same on artificial turf vs. natural surfaces. Important context in the ongoing debate.

Have you wondered about the magic behind generative AI art tools like DALL·E and Midjourney? Seeing the outputs is fun, and learning the algorithms can be even more rewarding. Today we look at 4 popular Machine Learning models — Autoencoder, Transformer, Diffusion Model, and GAN.

Beyond Limits Senior AI Solutions Architect and Data Scientist, Ari Kamlani had the opportunity to interview with Tyler Gallagher to discuss the future of artificial intelligence. Read the full interview in Authority Magazine below. bit.ly/3xlxPSx

AI Magazine features Beyond Limits’ Senior AI Solutions Architect & Data Scientist, Ari Kamlani, in a Q/A-style article with a special focus on Ari’s role and responsibilities. bit.ly/3Nu4wT0

Some of my learnings from consulting different companies on their machine learning efforts:

@mark_riedl I believe many athletes get custom designs on their training shoes and cleats. Some of them may have affiliations with artist brands like #kickasso

I have discovered that I can design bespoke shoes with neural network architecture diagrams from my papers. Send help

Beyond Limits Senior AI Solutions Architect/Data Scientist Ari Kamlani is quoted by insideBIGDATA, discussing how hybrid AI can help human operators improve decision making and innovate at a faster pace than ever before. Read the full comment here: bit.ly/3M6gfHt

@akamlani, Senior Data Scientist @beyondlimitsai, was present at #BigDataAITO 2021 to provide a POV on presenting a set of patterns to structure your AI projects and portfolio from the lens of sports analogies. Watch the talk here 📽️ ow.ly/FNHl50Hzpwx

Inspired by @__mharrison__ and the Styling chapter in Effective Pandas. Embed @matplotlib plots inside your @pandas_dev dataframe. #dataviz 😍