#49 - AGI: Could The End Be Nigh? (With Rosie Campbell)
Increments
00:00
The Limitations of Large Language Models
I just don't see a conceivable future where you can train it and it becomes so if, because this is like the first time I've seen a lot of people in the classroom paperclip stuff. The amount of training data is finite and soon it's going to run up against all the text on the web and then the capabilities are going to start to slow. We're already seeing the limitations of a lot of these large language models that will be able to navigate through websites but not disobey their masters. So yeah, I think it's just very risky to rely on that being the path to everything being fine.
Transcript
Play full episode
