The insight on safety is spot on - we don't need to censor the models into oblivion, but work towards safe outcomes of the systems that incorporate models.
Evaluation is a very good area . Multiple-agents for simulation is a very good area to work on . His insight is still very limited from LLMs technical perspective , but he has great insight from academic perspective .
To say it plainly, all governments have militaries working hard to make their AI models as harmful as possible. They are actively developing those. There are expensive contracts for those. So when we talk about "gating" it means, it is a PR exercise.
With this quality the channel will snowball - keep 'em coming!
Saw this on my feed & subscribed. Very good podcast & great guests.
The insight on safety is spot on - we don't need to censor the models into oblivion, but work towards safe outcomes of the systems that incorporate models.
Evaluation is a very good area . Multiple-agents for simulation is a very good area to work on . His insight is still very limited from LLMs technical perspective , but he has great insight from academic perspective .
Things are very exciting. However, it is so difficult to know what's real and what's hype.
🍓❤️
To say it plainly, all governments have militaries working hard to make their AI models as harmful as possible. They are actively developing those. There are expensive contracts for those. So when we talk about "gating" it means, it is a PR exercise.