At this point I feel like we understand pretty well what's going on with LLMs:

- Outputs are roughly equivalent to kernel smoothing over positional embeddings (arxiv.org/pdf/1908.11775…)
- The learned computation model is *probably* bounded by RASP-L (arxiv.org/pdf/2310.16028…)
-

NEW for FORTUNE: Open source AI is booming, but OpenAI’s GPT-4 is still the big winner with corporate customers—for now
Thanks to Sarah Wang at a16z, Ali Ghodsi at Databricks for chatting on this 🙏
fortune.com/2024/04/08/ope…

Hey Matt Turck, it'll be easier to read the map if you start each box with: AWS, Azure, GCP!!

DBRX by Databricks ...it's REALLY good!!

The New MoE 132b parameter model is open-source and costs $10 m to train. Thank you, Databricks, for your contribution to OS.

Check out the full explanation and testing:

🎥👇

It keeps getting smarter, we look at errors and are going to have the “diagnose error” functionality everywhere. Why is my job failing? Diagnose Error. Why does this dashboard not render? Diagnose error…

Is $20/month for an LLM chatbot too much or too little? It's the perfect price if you're reading 2.5 books a day, i.e. 17 hours continuous reading every day of the year!

Here is an oversimplified gross margin calculation. OpenAI offers GPT3.5-turbo API and charge $2 per million

There is so much focus on the standard LLM benchmarks (MMLU, ARC, GSM8k etc), but for enterprises the only thing that matters is how well the AI does on the domain specific tasks. Check out a comparison between DBRX and GPT4 on these domain-specific benchmark datasets.

This of awesome. Try it out on perplexity.

Super awesome SVA!

We all share responsibility for building AI that improves lives & unlocks a better future for humanity. @SVAngel makes this pledge, and we are proud to initiate this Open Letter: Build AI for a Better Future. Please join OpenAI, Meta, Google, @YCombinator, @HuggingFace,

I think this will mark an important milestone for Gen AI. The spotlight has been on the capabilities of LLMs (scaling laws, leaderboards, etc). But it's now clear that LLM performance alone will be meaningless. You will need a Compound AI system to get the best performance out of

Your enterprise --> Your domain data, which none of these companies have!

This thread highlights a point we've been seeing in dspy.ai for a while: you can't meaningfully talk about capabilities of a *language model*, you have to talk about capabilities of a *system*, including the inference algorithm. 32-CoT is not the same as 5-shot.

The founders of Databricks put together this strategy blog on where we think data platforms are headed in the future. We're moving Databricks quickly in this direction. This is very exciting and is the outcome of the MosaicML acquisition we did earlier this year!

Excited to be launch partners with Meta on the Llama2 release. This move by Meta will have a big positive impact on the industry and ecosystem. Technically, the first version of Llama already was available to everyone except anyone who had a commercial use case could not innovate

Closing out an incredible Databricks #DataAISummit with a keynote from Marc Andreessen 🇺🇸. We agree that AI is a way to make everything we care about better. Marc's article is required reading for anyone thinking about AI. Tune in to hear us discuss > databricks.com/dataaisummit/w…