#49 - AGI: Could The End Be Nigh? (With Rosie Campbell)
Increments
The Importance of Reinforcement Learning in Reward Hacking
I'm still trying to drill down on how much I disagree with you. How much your examples are meant to display the end of the world versus just like a severe problem? RL algorithms have shown in the past that they can engage in reward hacking. There's that famous video game, but I don't know which one it was,. It spun around in a circle in some way to maximize points. That is a very different problem than doing the same thing and then waking up to this system having taken over the world.
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.
