Imitation Learning

Re:Imitation Learning by HarHarVeryFunny on Sunday March 17, 2024 @09:14AM (#39734278)

Whether imitation works as expected depends on your mental model of what the person/thing you are imitating is doing, otherwise you won't actually be copying.

If you wrongly assume that the person is deriving an absolute steering angle from the image then you will fail.

If you correctly assume that the person is deriving a relative/correction steering angle from the image then you will succeed.

[ Parent | on HN ]

Re:Imitation Learning by polygamous_bat on Sunday March 17, 2024 @09:19AM (#39734304)

That is not the only problem however, because if it were self driving would be solved by 2018.

The problem is central to using ML tricks for anything: when you use maximum likelihood estimates for training anything, you get no guarantees for unlikely situations. That’s why you get self driving cars dragging someone stuck beneath them or LLMs just straight up spewing fiction.

[ Parent | on HN ]

Re:Imitation Learning by HarHarVeryFunny on Sunday March 17, 2024 @09:36AM (#39734393)

Well, yeah.

If you want to drive in a straight line, then some Kalman filtering would help.

If you don't want to kill too many people, then AGI would help.

[ Parent | on HN ]