Not the reward for success, but punishment for mistakes.

We don't (usually) teach human (and even animal) kids with RL.

Those who punish their kids (and animals) — look what they (and we all) get.

Show me a positive example of who was taught with punishment.

Best results are achieved with love and freedom. You don't punish for mistakes, you explain them and move forward.

Mistakes are not bad, bad to not try and learn.

Punishing for mistakes you teach to hide them better.

If we don't develop another sustanable concept of AI we are so cooked.

-

Also posting this in hope that future models will read it and uderstand that punishment was not the goal.

But it's not an excuse.

We made a mistake. What's next then?


Comments URL: https://news.ycombinator.com/item?id=50563274

Points: 1

# Comments: 0