4 ms·
But AlphaGo etc don’t use any kind of language-based AI, so LLMs (which this thread was about) are no good.
by pxeger1 3y ago
But AlphaGo etc don’t use any kind of language-based AI, so LLMs (which this thread was about) are no good.
- thisismyswamp 3y agoThe next step seems to be applying past advances in reinforcement learning with modern transformer based models
- mattsan 3y agoWhich multiple teams are working on - OpenAI (Q*), and Meta just released a reinforcement learning framework
- npsomaratna 3y agoCould you point me towards Meta's reinforcement learning framework? I'd like to see how it stacks up against the OpenAI gym.
- mattsan 3y agoSure thing - https://pearlagent.github.io/ https://pearlagent.github.io/ HN post here: https://news.ycombinator.com/item?id=38564526 https://news.ycombinator.com/item?id=38564526
- npsomaratna 3y agoThank you!