3 ms·
Ever heard of reinforcement learning, AlphaGo, AlphaStar? AIs can already make that trade-off and rather well. AlphaGo was actually better than humans to play r
by js8 4y ago
Ever heard of reinforcement learning, AlphaGo, AlphaStar? AIs can already make that trade-off and rather well. AlphaGo was actually better than humans to play robust moves that cement its victory than riskier moves that increase the winning margin.
- Vecr 4y agoRight, but then you go back to the whole "AI alignment" thing and you need to make sure you reinforce the right things during training.