[LOCKSUIT]

Supervision is when you tell it how to act or what to know. It's only as good as us.

We have a lot of data about how we do things like avoiding walls and how to write about science or politics or do surgery.

Why can't we do "Supervision" and create a humanoid robot that is as good as us but can't learn? They could be told to do surgeries and identify where to put there knife, then when done slicing, then when to look for the other tool, etc. Assuming they mustn't have to learn anything new and are in a place they know or stationed at the surgery room.

[ranch vermin]

I wouldnt use a knife, thats mostly how you kill people, not help them.

But about what your saying,   it would be hard to control a robot that was learning in a general way.  I know this one.   I dont touch on it so much because im implementing something more basic first.

[LOCKSUIT]

Ok but can anyone provide actual examples of how it begins to fail or "not come together" ?

[keghn]

 Free will. 
 A agi machine need to be created in a way that does every movement possible. Like playing around like the way kids do.

 And keep on doing it into a robot adult hood. But we can bribe them with energy if their batters are low. Or i should say
they will have goals. Fist line of goal is energy. Second line of goal is to act like their parent. If we confined them or keep
a source of energy away until the do what we want. Or we make a deal and show them a better way of achieving their goals,
 
 As kid are peers beat a lesson into us. Like school. No choice.  The strong strong arming the weak.
 What better then beat a lesson into a robot that does not have a mind. The strong strong arming the weak.
 The A+ scientist of today is because of peers beating lesson into them. They do not have a good concept of free will.
 And will never let a robot play. It will be neurotically hard coded in a super control freak manner. This type of approach will not work.
 So the best personality to make a agi machine is a artist.


 

[ranch vermin]

Yeh,  you just put him in a weak body so you can push him around if he decides to gets violent.

[keghn]

neurotic no fee will coding:

https://www.spectrum.ieee.org/automaton/robotics/artificial-intelligence/openai-releases-algorithm-that-helps-robots-learn-from-hindsight

[LOCKSUIT]

Nice article.

"To understand how HER works, imagine that you’re up to bat in a game of baseball. Your goal is to hit a home run. On the first pitch, you hit a ball that goes foul."

"Treating every attempt as a success at something, changing the goal so that the robot can learn a little bit."

"With HER, you’d say, “Oh, well, in that case, great, have a cookie!”"

;3 Great.

[keghn]

 There is a saying "monkey see, monkey do".
 Then on the other hand, if you are not born with instinct, like a animal. Then a kid will try every combination of movements in a tamed
genetic evolutionary way. Which will take for ever for the agi to get off the ground. Because the development will be in a linear way and with
energy management mixed in at every step. And my lead to something not very human like behavior. 

Hayao Miyazaki's thoughts on an artificial intelligence: 
 

 But if a AGI had a internal 3D simulator then it can copy and modle it parent. This way it will not take hundreds or years for the AGI to find
all of  right movements with a genetic evolutionary algorithm on its own.

[ranch vermin]

Theres more rammifications for human movement, id say.   No bunnies were hurt during the experiment.

[Art]

They would have been far better off, and in better taste, had they chosen to use a four-legged entity to simulate locomotion or learning to walk behavior.

Obviously, their use of a twisted human writhing along the floor hit some very real and sour notes with Mr. Miyazaki as he has a friend with a severe disability. While I'm sure the students meant well, a bit of forethought might have saved them the classroom embarrassment.

Just my $.02