Would be cool to re-launch a kind of SETI@Home where locally running models (and OAI/anthropic accounts) could participate with their agents (acting as sub-agents) to do research during the night. I think there is something already about renting local GPU time out, but for profit.
Lots of security issues etc with this. But conceptually it would be nice to be able to contribute to space research that way.
I said this in another thread recently, but I feel like large-scale generation of open-access synthetic data could be a good way to donate unused capacity.
In the late '90 I participated to SETI@home, donating free CPU cycles over nigh on some servers that were running 24x7 and were not utilized at night. Power saving was not a thing, so all that electricity was wasted, I found a use for it.
These days energy is expensive. It may be better to shut down home computers and let organizations use optimized hardware for such type of processing.
What's happening in the math field at the moment represents on a smaller scale the issue we (probably) will face when AI becomes smarter than us in general. Do we slow the AI down so we can understand what it does and if it's correct (and align with our values for bonus points); do we, humans, adjust the way we work to the new speed or do we give in and let the AI advance while we're not completely sure of what it does and the correctness of it.
The math field is in the position of showing us the way.
I think the bigger issue is that "we" won't be doing any deciding. A handful of oligarchs will. As the technology becomes better than us, I am afraid we will learn that we are (at least seen as) technology as well. And the owner of the better technology will decide what to do with the inferior technology. I don't see anyone trying to design the roads for horse carriages, so I don't understand why would anything change for humans' understanding.
That is, if things go in the current trajectory. I don't see any reason why anything would change though.
Have you thought about making the whole thing "self-similar"? Every time I hear about MoE I think (and I know it's way easier thought than done) "why stay shallow"?
I mean by that: would it be possible to extend/adapt the architecture so that an expert can be a previously trained Mini-AGI model? And recurse like this? Inuitively I would think some form of generalization could happen, as higher level experts (in the recurrence stack) would become sort of the "intuition" layer.
Making model to consists of many small modules is inefficient on GPU, especially as routing adds data dependencies, etc, and especially with pytorch (compared to a custom kernel).
The difference might be smaller on a CPU which has limited parallelism.
But it's basically equivalent to a very deep model which might be problematic for training.
In my experience you can “stop reading the code” if you adhere to prompting the AI aiming for changes that would be reasonable in a PR to review, if you’re using rather opinionated framework(s) to base your work on and explicitly ask for tests (models tend to add them on their own if there are some).
I personally still read the output and start by checking if existing tests were modified (kind of a red flag when this unexpectedly happens imo).
When I read posts about AI generating garbage nowadays it’s either because of a small prompt/big ask combination or a lack of an underlying framework.
Very naive approach but wouldn't training a model on a few specific buckets like "someone is naked", "image is explicit" and "there is a child" in the picture, would do the job without having to train explicitely on CSAM? If the model returns both a high probability of "there is a child" and a high probability of any other bucket than this image is classified as CSAM. It'd be high recall,low precision but one would be on a rather safe side.
If more and more of the web's content is AI generated, AI companies are bound to train on each other's data.
Or, what if I generate content with Claude/ChatGPT/Gemini, warp it in HTML using an open model, put this on my website conveniently dedicated to "Best practices in prompt and AI answers" for example, then train my own model that only scraps my website?
The only ban that should be put in place is on models' general and specific capabilities. If they have open or closed weigths is irrelevant. With the current state of capabilities there is already plenty we can do (in terms of creation and destruction unfortunately) before we as a society need more powerful capabilities.
Once the limit is set, any non- governmental/military entity providing/hosting/using a more capabable model can be prosecuted.
reply