notacat

Unfortunately not for this first batch. If there is a second batch after this (~10 months from now), then yes.

Yes. You can submit an application as a team. Both of you should be present during calls related to the application. After that, it's ok if only one person can do the reporting/communication, even though it's better if you can both attend. There would be 2 grant contracts (one for each person individually, with the total split between the two)

Yes, please reach out and repost.

I didn't get the feeling that lesswrong is the right place for such a post. Is it? What tag should I use? Anything else I should know when posting there for these this kind of content?

You can send a regular email, I'm happy to chat about it.

I consider it outside the intended scope of the grant because your project (as I understand it) is about applying existing techniques to build a product. What I'm looking for is work that tries to understand or improve on them, or create new ones.

I'm interested in anything that can help us get a better understanding of ML-related topics: generalization, tranformers, optimization... So I have a bias in favor of experimentation and theoretical stuff, and less so about practical applications (eg AlphaFold as you mention). With the caveat that sometimes these applications do lead to discovery.

I want to get involved in the design/prototyping. My skills are Deep Learning, a bit of statistics, finance and frontend dev, and I played a lot of games. I'd say if I spent a few years on it myself, I eventually could make a prototype, but I feel given my position I have more leverage if I just act as an investor + co-designer. I also know this is a very risky investment and I understand that the most likely scenario is that nothing comes out of it (I made other similar investments in the past)

I said noita is the closest I found, but it's still pretty far from what I envision. I also understand that this idea is probably not enough to make a game compelling. It's more a mechanic that could be added to a broader game. There are many possibilities there which should be mostly chosen by an experienced team or studio and not by me.

Yea one plan would be to start with some kind of prototype, which should be possible in the 5-50k range I expect (could be a very small minimalist prototype by a single dev). Then if this looks fun, spec together a bigger project & fund it. I just don't have the right connections to game developers / studios.

I like the idea of finding experienced people. Do you have any pointers to find Discord channels with such people and/or having similar ideas?

One thing is that i'm willing to discuss and change my mind. So I think my best bet is to find a studio or people who are already having similar ideas and iterate from there

AI Research firm | Zug/Zurich area | Full time | On site | 100-200k chf/annum | 1 year research residency program.

I am starting a new AI research team funded by applying ML to crypto algo trading. Over the last three years years, the ML-based trading operation turned 5 digits into 9 digits, trading a volume of around 1-4B per day, with a team that has grown to 7 engineers. The crypto markets are a bit more calm now, so things don’t look as good as before, but still good enough to move on with the plan of using trading profits to fund the research operation.

To be clear, I did start the trading operation with the intention to fund fundamental AI research (capability and safety) since the very beginning. In the meantime, both AI and crypto markets went through boom and bust cycles, and I expect there will be many more to come. We need to build an operation that can go through many such cycles which is why I want to grow the research team slowly and cautiously.

This is the first role for the research team. It includes studying and doing research together with me. The contract will last 1 year by default and end. If the research is promising, and if finances allow, it can be extended into a full-time research role. The compensation for the initial residency program will depend on the person applying and can be higher for exceptional candidates.

The high-level research agenda is whatever might help to get to AGI in the long run, with safety in mind. My own research interests are the following: deep learning, probabilistic programming, world models, self-play, and more recently LLMs. With regards to LLMs, I’m interested in studying their inner workings, for example how optimising for a simple objective (predicting the next token) gives rise to “emergent behavior” such as the ability to count, or theory of mind.

If interested, contact me at not_a_cat at fastmail. co[m].