Google's AI taught itself to walk

Status
Not open for further replies.
I wonder how they 'incentivized' it to go from point A to B...

The climactic music they used elevates existential feelings of anxiety. Just look at the comments :dead:

incentivization in this case is the result of the machine learning algorithms trying to minimize a certain function

in this case, the function is a distance function. the ai tries to make the distance between the player and the end point equal to zero. when it fails at point X, it tries to update its strategy in certain ways so that next time it fails at a point Y which is further than X

basically how it works

also, this explanation might sound very technical and verbose but it is how all artificially intelligent systems work. it's how skynet will one day be able to predict what you're going to buy on January 19th, 2031, for example

creepy stuff
 
But this leads to another question, what will happen when AI become powerful enough to make decisions faster than humans with a greater accuracy?

Will we limit the AI capacity to the extent of the human levels or will we let it surpass us and we get no final say im decision making? If so then say good bye to almost all of the industrial and Tech jobs..

Most AI have algorithms programmed to improve upon failure. When one strategy to reach a certain point fails the AI comes up with a different strategy which surpasses the previous one and achieves the goal. This constant improvement will one day surpass us humans its just a matter of when...
 
Status
Not open for further replies.
Top