I agree. AI video could have many wonderful uses, but without laws to protect people, especially children, YouTube, TikTok, Instagram, etc will have it their way. Shockingly, they use RL to optimise the length of time users stay on the site and to ensure that they return. They
@nandodf
-
Top Canadian and UK Universities for AI Education
By
–
UBC, Alberta, Montreal, Toronto, Waterloo, McGill, Simon Fraser … so many amazing universities in Canada for anyone wanting to study AI. There’s outstanding places in the UK too: Oxford, Cambridge, Imperial, UCL, … and tons in Europe. Frankly from any of these places you can
-
Red teaming AI models for safety improvement
By
–
Nice work. Good to have people red teaming our models.
-
AI Innovation Empowers Everyone Through Accessibility
By
–
Congratulations @redpony — you and your team put so much work into making this a reality— which I hope the product teams will make easily available to people in need of this. It makes me truly happy to see AI empowering everyone
-
Distributed Algorithm Optimization: Speed Theory vs Empirical Reality
By
–
Not sure, just putting it out there as an option people might not have thought of. It’s faster theoretically, but I’m not sure this is true empirically, specially in distributed settings. I’m thinking about it, but maybe others can do the thinking better and faster here For
-
Formalizing LLM Benchmark: Alternative to MMLU and Arena
By
–
It would be great if you or anyone could please formalise this into a benchmark for LLMs. It reminds me of the Hutter prize https://
en.m.wikipedia.org/wiki/Hutter_Pr
ize
… and it would provide a nice alternative to comparison @lmarena_ai and to multiple choice, eg MMMU and MMLU. -
LLM Agency and Reward Optimization: External vs Internal Motivation
By
–
One more question @balazskegl
: if the LLM was trained to optimise a reward, then it could be argued that it has agency. This is what happens in LLM post training. I guess the counter argument would be that such rewards are external. Yet, the Schrodinger argument is that the -
Do LLMs possess intrinsic intentional stance like humans
By
–
So I guess my question is: Is your position that humans had an intrinsic intentional stance but that LLMs don't?
-
LLM Intentionality and the Herding Method for Neural Networks
By
–
Thanks Balazs. I posted the herding method. It's certainly pretty good for normalised distributions. LLM folks should probably be trying it. But I agree your argument goes beyond this. It is about whether the net has an intentional stance. Daniel Dennett offered some thoughts
-
Deterministic Algorithms for LLM Sampling Beyond Stochasticity
By
–
The LLM parrots need not be stochastic. Long weekend thought motivated by a posting by @balazskegl : A few years ago @yutianc and @wellingmax introduced me to herding. It is a dynamical systems (deterministic) algorithm for producing samples for a multinomial discrete