I'm one of the many people who Soumith hired to Meta and PyTorch. I had the privilege of working on PyTorch with him and lots of the folks on this post.
As his longtime colleague, the one thing I would want people to know about him and this decision is that Soumith has always viewed PyTorch as a community project. He consistently celebrated the contributions of his co-creators Adam and Sam, and he extended the same view towards the Yangqing and the Caffe2 crew that we merged into PyTorch. At the very beginning, by Soumith's highly intentional design, PyTorch was aimed at being truly developed by and for the AI research community and for many years that was the key way in which we grew the framework, FB PT team, and the wider community. At every single stage of PT's lifecycle, he always ensured that our conception of PT and its community grew to include and celebrate the new people and organizations growing what was possible with PT. He's an incredible talent magnet, and thus more and more smart people kept dedicating their blood, sweat, and tears to making PT bigger and better for more people.
I've worked with some very well known and highly compensated leaders in tech, but *no one* has done the job he has done with ameliorating a bus factor problem with his baby. PT has a unique level of broad support that few other open source technology can reach. In a world of unbounded AI salaries, people who want to move AI research methods forward still freely give their time and attention to PyTorch and its ecosystem. It's the great lever of this era of AI that is moving the world, *due in large part* to the strength of the community he fostered and can now let continue without his direct involvement.
His departure is the end of an era, but it's also operationally a true non-event. PyTorch is going strong and can afford to let one of its creators retire from stewardship. This is precisely what success looks like in open source software.
He deserves our congratulations and our thanks. Enjoy your PT retirement, man.
Also worked with Soumith. The man is a legend, moves mountains and completely changed the course of my career because he liked something I wrote. No arrogance, no politics, just an extremely down to earth and chill guy who elevates everyone around him.
American here who went to a Chinese (grad) school for CS and was admitted to every Chinese school I applied to. This is very much a possible route, if you’re appropriately qualified for the program. The main issue is language: outside of HK, programs in English are rare.
That's extremely impressive that you managed to reach such a high level of fluency. I find written and technical Chinese is extremely tricky and different from spoken Chinese
Not sure what's with the HN tone on this announcement. AI2 are really some of the best people around for creating truly open artifacts for the whole ecosystem. Their work on OLMo and Molmo is some of the most transparent and educational material you can find on model building. This is just great news for everyone.
Maybe because many of us are not from the US. The stated goal is US dominance of the AI field, and sorry if the rest of us don't see that as a good thing nor particularly open.
The Allen Institute for Artificial Intelligence projects so far have been very open. They are open about the trained models, the inference code, the training data sets, and the training code. A research group from any country can pick up where AI2 left off if they want to try a different approach or extension. I want to live in a world where there are many models near the top of leader boards, from many different research groups and countries, and I think that AI2 helps enable that.
The stated "US dominance" goal just pays lip service to what appeals to the funders, kind of like how supercomputing projects traditionally claim that they contribute to curing disease or producing clean energy. (Even if it's something far removed from concrete applications, like high fidelity numerical simulations of aqueous solutions.)
Any country tries to dominate any field if they can do it, it's just human nature. Why is that a bad thing?
That constant competition for superiority between nations is how humanity has evolved from hunter gatherer to having tractors, microwave ovens, airplanes, internet and penicillin.
Yes, competition is good. Monopoly is bad. A more distributed power structure is much better for overall progress, and even for the monopolist in the long run (Ex: Intel).
Nah, I'd say just do more anti-monopoly anti-inequality work. Probably start internally, that's a massive enough task on its own (eg breaking up big tech). Assist other countries eg with aid if (and only if) they are doing the same. This is a big topic, ask your favorite frontier LLM about it.
Unless China does the same that's an unrealistic ask. That would be like doing nuclear disarmament but only you and everyone else gets to keep their nukes.
Your "nukes" are leaking into the water supply. Inequality shows up a million different ways that Americans don't fully understand yet, eg inflation (dominant companies increasing profits), teacher shortages (low wages), student debt (not an issue for the wealthy so why fix it), housing prices (corporate landlords, exclusionary zoning), layoffs (despite record profits) etc etc. This situation (Trump/Musk/Bezos taking most of the gains) is just not long-term stable, and if any other country wants to do the same to themselves let them. The longer it goes the harder it will be to fix.
Again, go have this discussion with the LLM you trust, it's much more informative.
As an American, I obviously can get behind it, but I can easily see how a declared goal of superiority of others would rub those others the wrong way (and possibly prevent their contribution)
>a declared goal of superiority of others would rub those others the wrong way
So what? Does that change anything in how things work in reality? Everyone knows it, so why pussyfoot around it?.
Why are people nowadays so sensitive about saying the truth of how things work? Have people been coddled that much that they've can't handle reality? A good life lesson is that the world does not revolve around your feelings.
It's not my feelings mate, if you don't live outside the US and have not been subjected to their unipolar attitude you will probably never understand and there is literally nothing I'm going to say to convince you of the objective reality the rest of us face.
Of course you can justify this, as people have, but you can't then blame the rest of us non US citizens for not aligning with that goal. The US is only a small portion of the global population and the government itself has a long history of stamping on the rest of us.
Howdy, HN. Authors here. We got tired of text-to-image leaderboards that only focus on aesthetics, so we built our own benchmarks to test what matters for real work: fidelity to complex prompts, safety, bias, and IP infringement.
We analyzed 18 models and found that no single model is good at everything. For example, GPT-4o has the best safety guardrails but also a 98% IP infringement rate on celebrity likenesses. Google's Imagen 4 Ultra actively counters bias (e.g., 90% of its "CEOs" are female) but struggles with generating crowds. X AI's Grok 2 blocks almost nothing.
Lots more detail in the post. We'll be here all day to answer questions.
We're a stealth mode genAI startup working to advance the state-of-the-art in generative visual media. We've just closed our preseed round and are now backed by 3 expert VC funds and a collection of AI leader angels. Our team consists of highly experienced AI leaders from leading companies. And we're working with some of the biggest companies in media.
The role we're hiring for is an engineering leader to drive the evolution of our research platform into our first commercial-grade product. This work centers around the orchestration and scaling of multimodal understanding and image and video generation models. You'll be the primary architect of our platform capabilities and working closely with both cloud and model partners as well as enterprise customers deploying our technology.
This is a unique opportunity to be part of the founding technical cohort of a deeply innovative team with a unique technology that no one else is building.
We're a stealth mode genAI startup working to advance the state-of-the-art in generative visual media. We've just closed our preseed round and are now backed by 3 expert VC funds and a collection of AI leader angels. Our team consists of highly experienced AI leaders from leading companies. And we're working with some of the biggest companies in media.
The role we're hiring for is an engineering leader to drive the evolution of our research platform into our first commercial-grade product. This work centers around the orchestration and scaling of multimodal understanding and image and video generation models. You'll be the primary architect of our platform capabilities and working closely with both cloud and model partners as well as enterprise customers deploying our technology.
This is a unique opportunity to be part of the founding technical cohort of a deeply innovative team with a unique technology that no one else is building.
We're a stealth mode genAI startup working to advance the state-of-the-art in generative visual media. We've just closed our preseed round and are now backed by 3 expert VC funds and a collection of AI leader angels. Our team consists of highly experienced AI leaders from leading companies. And we're working with some of the biggest companies in media.
The role we're hiring for is an engineering leader to drive the evolution of our research platform into our first commercial-grade product. This work centers around the orchestration and scaling of vision and image generation models. You'll be the primary architect of our platform capabilities and working closely with both cloud and model partners as well as enterprise customers deploying our technology.
This is a unique opportunity to be part of the founding technical cohort of a deeply innovative team with a unique technology that no one else is building.
