Krea | Distributed Systems & Backend Engineers | San Francisco, CA (onsite+remote)
I'm the cofounder & CTO of Krea (https://www.krea.ai). We are a startup in San Francisco building browser-based AI creative tools for professional designers and creatives. Small team (~27); millions of active users; and raised +$80M from top SV investment firms.
We're looking for talented engineers who want to tackle hard technical challenges with smart people while building a creative platform that big companies and startups will rely on. You will...
- Build distributed systems to process massive (billions of files, petabytes), amounts of image, video, and 3D data, solving scaling bottlenecks as you go.
- Learn ML engineering from world-class researchers on a small tight-knit team. You’ll contribute to foundation image, video, and world models from 0 → production, and see them used by millions.
- Play with huge Kubernetes GPU clusters; our main GPU cluster takes up an entire datacenter from our provider.
Tech stack: python, pytorch, k8s, a rotating cast of data tools (e.g. DuckDB, massive relational DBs, PyArrow, etc.)
You should apply if you are an excellent generalist engineer with strong backend experience and an intuition for systems design. Bonus points for experience with dist. systems, k8s, and existing ML or data experience (but not required!). Cool side projects are a green flag.
This got to the top of HN in minutes and then got removed. Probably to stop drama. I think it’s okay to laugh at these types of sites and jokes, but I get it when venture and HN get intertwined, it can get thorny.
> I work in AI and, to this day, I don't know what they mean by “world” in “world model”.
I have a PhD in ML and a B.S. in physics. What people in ML call a "world model" seems incredibly strange to me. With my physics hat on, a "world model" is pretty clear. It is "a physics." Mind you, there is not one physics, there are competing models and we're just at a point of time that models have converged up to quantum and gravity.
But "a physics" can be a model that describes any world, not just the one we live in. For ML models, this should be based on the data they're processing. Ideally we'd want this to be similar to our own, but if it is modeling a "world" where pi=3, then that's still "a physics".
The key points here are that a physics is a counterfactual description of the environment. You have to have language to formalize relationships between objects. In standard physics (and most of science) we use math[0], though we use several languages (different algebras, different groups, different branches, etc) of math to describe different phenomena. But the point is that an equation is designed to be the maximum compression of that description. I don't really care if you use numbers or symbols, what matters is if you have counterfactual, testable, consistent, and concise descriptions of "the world".
Oddly enough, there are a lot of physicists and former physicists that work in ML but it is fairly uncommon for them to be working on "world modeling." I can tell you from my own experience talking to people who research world models that they respond to my concerns as "we just care if it works" as if that is also not my primary concern. Who the fuck isn't concerned with that? Philosophers?
[0] It can be easy to read "The Unreasonable Effectiveness of Mathematics in the Natural Sciences" out of context as we're >60 years past where math has been the lingua Franca of science. But math is a language, we invented it, and it should be surprising that this language is so powerful that we can work out the motion of the stars from a piece of paper. Math is really the closest thing we have to a magical language https://web.archive.org/web/20210212111540/http://www.dartmo...
Broadly 'world' means 'the domain I'm interested in'. In current use in the DNN context 'world' tends to be physical space at a scale relevant to humans or robots (eg. autonomous vehicles). So when someone says 'world model' you have to ask 'what kind of world, and how is it represented?'.
We don't need to agree on one very specific meaning, which is good, because we would fail.
Yeh I still don't think there's a fixed definition of what a world model is or in what modality it will emerge. I'm unconvinced it will emerge as a satisfying 3d game-like first-person walkthrough.
I presume that email address is for when you want to ask something of Hacker New, not to ask something about Hacker News.
For example they probably didn't want posts like "Hey Hacker News, why don't you call for the revival of emacs and the elimination of all vi users?" and would rather you email them so they can ignore it, but they also don't want email messages asking "How do I italicize text in a Hacker News comments, seriously I can't remember and I would have done so earlier in this comment if I could?" and would rather you ask the community who could answer it without bothering anyone working at Y Combinator.
Are you saying this based on experience or are you projecting? In my experience (tho not asking how to italicize text using * characters) Dang and tomhow are happy to answer all sorts of questions. Sometimes they do get bogged down by the reality of running a site of this site manually, as it were, but I can't remember a question that didn't eventually get answered. I'll even tell them I vouched for this bunch of dead comments, was that the right thing to do? And one of them will write back saying mostly, but just fyi comment xyz was more flamebaity than idea, but thank you for asking and working on calibrating your vouch-o-meter.
What's the problem? Someone submitted it for people to read but it didn't catch on, now it's resubmitted and people can read it after all. Everyone happy. Don't be so attached to imaginary internet points.
Wow!
reply