GPT-4 with Matt Browne
Embrace The Void
The Limits of Large Language Models to Embodiment
The authors of this paper sort of concluded that there was some fundamental limitation to large language models to do with embodiment. That's not true right because with four. They did nothing but really scale up the size of the model. And we have these sort of emergent capabilities that weren't there in 3.5. Now you still found it had some limits on the embodiment stuff right that it was giving like unnecessarily complex answers. Or did that sort of not pan out when you were digging into it more? Aaron: I did see some issues with the embodiment because I really wanted to find out whether there were some limitations there so one I managed to stump 4.0 with a couple of embodied questions
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
