CEO Unconventional AI. Former CEO MosaicML/Databricks & Nervana/IntelAI. Neuro + CS. I like to build stuff that will eventually learn how to build other stuff.
San Diego, CA
Joined February 2009
- Tweets 4,527
- Following 891
- Followers 33,499
- Likes 13,530
Pinned Tweet
Hello world Unconventional, Inc.
I’ve gotten a new company off the ground. It’s a big swing…rethinking the foundations of a computer to build a new substrate for intelligence that is as efficient as biology. Brain Scale Efficiency without the biological baggage!
We CAN do it. We have to think differently from existing dogma. The computer in our heads proves it’s possible. 20 watt brains are within our reach with the right research and beautiful engineering.
Can you think differently? Do you want to be Unconventional? If so either drop me a DM or email at jobs@unconv.ai
I am deeply honored, humbled, and will have big shoes to fill. @roelofbotha changed my life.
He believed in me, brought me to Sequoia, and taught me how to grow from a builder into an investor.
He’s demanding and deeply caring — the rare kind of friend, mentor, partner who pushes you and stands with you through every storm.
I’m not getting the reason for all the hate. I’m obviously missing context on these conversations, and maybe it’s just personal. But it looks like @Extropic_AI did build something interesting. Existing AI methods are highly overfit to operations that are efficient on GPUs. New hardware that is efficient in different operations might be used to go in new directions. Not sure why there’s such a backlash to that notion.
While big salaries and IPOs get headlines, what makes Silicon Valley special is not the money...it's the mission driven nature of the people. The drive comes from an intense desire to build and change the world. The money is just a consequence.
This is why it has been so hard to replicate anywhere else.
At Unconventional, we’re building the computational substrate for the AI era.
Scientists and SWEs interested in dynamical systems (Diffusion, Neural ODEs, Deep Equilbrium Models, and Energy-based models) DM or email jobs@unconv.ai (subject: dynamics). Things are getting really interesting really fast…
When you log on to the SJC wifi, it literally brags that it's fastest airport wifi in the country. So I tested it. They weren't kidding! It's the little things...
We have some of the world’s experts in this area as part of the founding team: @mcarbin , Sara Achour, and MeeLan Lee. It’s going to be fun 🤩
We have some of the world’s experts on this problem as the founding team. @mcarbin, Sara Achour, and MeeLan Lee are joining me on this journey. It’s going to be fun 🤩
Yup. Been saying it for a while..AI will make things faster and expectations will be greater. The pace of products will be faster. The time to production will be faster. This requires more talented people, not fewer. The job will be different though.
I’m the opposite of a doomer…I don’t understand why anyone would work in this field if you believe it’ll destroy the world. Does that make me a thriver? Optimister?
AI will elevate humanity and allow us to collectively understand our universe in deeper ways. I’m signed up for that future ✅
The reason that AI isn’t going to wipe out jobs in the way that some predict is that we consistently make the mistake of thinking that when we make something more efficient, you need commensurately less supply.
It turns out that in a significant number of fields, better productivity levels actually means more demand for that service. This is the whole point of Jevons paradox.
When the cost of doing work goes down, the demand for it goes up. And usually there’s far more pent up demand than we realize.
When AI drives up the output in these fields, thus lowering costs per unit per output, demand is going to rise of unexpected ways. This is true in healthcare, code, legal services, marketing, and a ton of other spaces.
Today is my last day at @databricks .
~2.5 years ago @alighodsi told me his goal was to build a $100B company. Databricks was at a $38B valuation when MosaicML was acquired in July 2023 and just broke the $100B valuation number. It’s amazing to be part of this growth! And now AI is infused in everything that @databricks does.
While it has only be 2 years, I feel very close to the company. Thanks for the awesome journey in the inner dev loop with Databricks founders @alighodsi @matei_zaharia @pwendell Arsalan Tavakoli-Shiraji @rxin . It’s not often you get to bear witness to a company maturing into a foundational player!
Innovation and investment in the AI era are taking on new forms. We’re spinning out a new effort around foundational challenges in AI and computing and I’m excited to be leading it! I will be staying on as an advisor and look forward to seeing Databricks lead the evolution of enterprise AI.
And I can’t believe it has been 5 years since @jefrankle @mcarbin and @hanlintang and I started talking about how to make neural nets efficient. That was in the middle of COVID and so much has happened in the world! Thanks for the fun ride.
I believe AI will touch the life of every human on the planet and help us understand our world in unprecedented ways. Databricks is extremely well positioned to catalyze enterprise usage. And the work continues to achieve economics that scale. That won’t come easy and we have to build the world we want to create.
Stay super awesome Bricksters!
Congrats @__jamie_b and Rohan! These guys have built something really amazing in a short period of time! A copilot for data science will really enable businesses to make better decisions quickly.
🚀 Thrilled to announce our $9.5M funding round led by @buckymoore at @lightspeedvp, alongside an incredible group of investors from the Valley and New York.
✨ With this announcement, we’re also moving Sphinx Copilot -- the state-of-the-art AI agent for data science -- out of closed beta. It’s now available at sphinx.ai (with a generous free tier!). Our early partners have gone from raw data ➝ commercial insights in minutes instead of days. We can’t wait to see what the data community builds with Sphinx.
🌱 This is just the beginning for Sphinx. We’re redefining how AI works with data, from copilots to fully autonomous researchers and analysts. We're excited to keep building best-in-class machine intelligence for a new generation of data-driven innovation.
sphinx.ai/blog/sphinx-launch…
Wow...well, I don't know what the plan is, but I generally don't see how gov takeover help things. I have some ideas though...
BIG NEWS: The United States of America now owns 10% of Intel, one of our great American technology companies.
This historic agreement strengthens U.S. leadership in semiconductors, which will both grow our economy and help secure America’s technological edge.
Thanks to Intel CEO @LipBuTan1 for striking a deal that’s fair to Intel and fair to the American People.
👏
Databricks just signed a Series K term sheet at >$100B valuation to scale two flagship products:
🔥 Lakebase — serverless Postgres with true compute/storage separation
🧠 Agent Bricks — agentic framework with built-in reasoning guardrails for enterprise data
wsj.com/tech/ai/databricks-r…
This article does a great job laying out some of our core messages. This is why I believe Databricks is so well positioned in the AI era!
As demand for smarter, more nuanced agents grows, enterprises are shifting their focus to evaluation.
Databricks VP of AI @NaveenGRao and Chief AI Scientist @jefrankle explain why focusing on behavioral evaluation over model size is important and how innovations like Test-Time Adaptive Optimization and Agent Bricks are improving the quality of agent outputs in terms of accuracy, efficiency, and alignment with business needs.
Read more in @FastCompany fastcompany.com/91384747/dat…
Some really cool work on making agents more reliable
Seems like this hat is now pretty relevant lol! @jlarosenthal made this in 2017 when I was starting the AI group there. We wanted to change the company from within to be an AI focused company!
🤔 so you’re tellin’ me they didn’t have AGI figured out all along?!
Google search vs LLM (ChatGPT)
This is really "find something similar" vs "synthesize answer based on facts". Really specific queries are much better with LLMs since there is no perfect match in a retrieved article.
"How many times has a driver with a missing thumb won the Monaco GP" . LLMs give a comprehensive answer, but search gives related but irrelevant results.
@Google has such an opportunity here...using agents they could produce dynamic web content that specifically addresses the question. Make it rich...words, pictures, videos. It's like an expert in that field with access to all knowledge that uses experience to write an article that addresses the specific query.
Why are they not doing this?!?!