Dar Mehta
@dar_mehta
Followers
338
Following
248
Media
28
Statuses
309
building @_CausalLabs; prev @cruise, @waymo, brain @GoogleAI, kiyo robotics (YCS20), @AIatMeta, @uwaterloo
sf
Joined January 2015
Thrilled to be revealing a glimpse into what I've been building with Kelsie over the last year!. A key bottleneck on the path to superintelligence is understanding causal relationships. We plan to fix that at Causal Labs, starting with AI physics models to predict and control the.
We’ve raised $6M to build AI physics models to predict and control the weather. Big thanks to @KindredVentures, @refactor, @BoxGroup, @factorialcap, Otherwise, Karman VC, and our wonderful angels. Come help us solve the world’s most difficult problems. We’re hiring!
31
27
116
This team is special, lfg @Marg_Groisman!!.
For too long, data center management and operations tooling have overpromised and undelivered. As data centers have evolved to become AI factories, legacy tooling and limited innovation has meant downtime, hotspots, lack of asset visibility, and more. That ends now for teams
0
1
4
RT @jxmnop: google simply does not get enough credit for the TPU. the amount of conviction & foresight it took to fund and build hardware….
0
252
0
Paying for internal knowledge on other labs’ data strategies & subsequent infra to acquire that data.
Paying $$$ for top AI researchers is the name of the game but Scale seems like a weird choice here. Great business, but not really known as a research lab despite marketing. Fresh start is good but $10B for this versus OpenAI/Anthropic VPs is . not an obvious slam dunk IMO.
0
0
1
Incredible seeing google fumble despite easily having the best models across every category now.
1
0
4
👀
Meta CTO Andrew Bosworth says that they're finding in Meta's robotics efforts that we're hitting the information-theoretic limits of intelligence; all the human media ever produced isn't enough; breakthroughs in common-sense 'world models' will be needed to understand causality.
0
0
1
RT @anandnk24: Today, I’m proud to launch the first MLLM-as-a-Judge. $ pip install patronus.to scale image evals. Here’s why this is game….
0
10
0
RT @stevejang: @dar_mehta "/[physics] rules everything around me/".we're stoked to back you and kelsie!.
0
1
0
RT @_mayankkishore: When Dar explained to me the plan, it became instantly clear that this was THE path to superintelligence and this is TH….
0
2
0
This “Dar” guy has such insightful tweets.
‘Don’t forget that the guy who stuck around your home town and became an electrician just cleared a million this year and gave your old friends jobs so they can put food on the table…. The illusion that you are more than another man because the society has validated you more.
0
0
8
Most people have no proper definition of intelligence beyond OpenAI’s arbitrary definition tied to economic value making all such predictions of “agi” meaningless.
If 1 or 3 or 5 years go by and AGI doesn't emerge from LLMs, there's no backup plan, really. It's back to the drawing board. It's however long it takes to come up with something else. Maybe that's a few months. Maybe it's 20 years. Maybe it's never.
0
0
1
Difference is that AGI fear is often fud, not backed by empirical observations but rather obscured and biased by immense financial incentives.
There is a weird amount of overlap between the people who insist that we take climate experts seriously when they warn of big changes on the horizon, and people who refuse to take AI experts seriously when they say the same thing.
0
0
2
Ridiculous to think high school debates are more coherent than this mess.
0
0
0
how does google sheets still not have gemini-based autofill yet 🤦♂️.
0
0
2
damn this guy names.
"Why call your company Midjourney?" well lots of people in tech act like we have no past, and many regular people act like we have no future, but ive always felt we are very much mid-journey. we have a rich and unbelievable past behind us and a vast and unimaginable future ahead.
0
0
0
What’s more likely is that those domains are just easier. Biology, environmental science, physics, robotics all have to deal with sparsity & partial observability making them much more difficult. It’s interesting that it might be opposite in humans since babies typically learn.
Seems to be not a coincidence that what AI is good at is correlated with the backgrounds of AI researchers. Demis was a chess prodigy; Jakub (Chief Scientist) and Mark (CRO) of OpenAI were competitive programmers; many IMO medalists at OpenAI, x-ai. If our world initialized with.
0
1
3
VCs can get such easy alpha if they just install cameras around south park and log commute hours of all founders in the area.
1
0
3