This is creepy

[Debi]

 

[Duke]

View attachment 76388

I don't get it.

 

[Paulm]

so, we have 180 billion brain cells and it only takes 200 thousand to play video games. i wonder if professional video gamers hate the results of this test.......lol

 

[Debi]

I don't get it.

They grew brain cells on a chip and taught it to play a game that takes problem solving abilities. This is the beginning of "replace a human" or "program a human brain" technology. Welcome to the future.

 

[ozentity]

Elon thinks a lot of bad things are good so this must be bad.

 

[Stevedog]

This is something thats been happening for over the past two decades. Back in the early 2000s there was a group of scientists that ran a similar experiment with a clump of rat brain cells operating a flight simulator. The idea was to use the billions of years of evolution in how neurons function and adaptat to make a better, faster computer processor. A biological computer.

If you've ever seen that video where they attached sensors to a plant and then fed those signals into a robot arm holding a machete; this is the next step up from that. The plant doesn't know it's got control over a machete, it's just doing plant things and is unaware that those signals are being captured. This biological processor is getting feedback and responding.

Data from the game is being translated into a signal and fed into this "brain chip". The chip generates a signal in response and the input signal changes. It doesn't comprehend its own signals, just that generating different signals changes the input. It doesn't know the difference between turning to the left or firing a shot. Overtime it's going to see a specific signal pattern over and over; it will produce different outputs until the signal changes in a positive or negative way. It will then repeat the positive action more often and the negative action less often. It has no idea what those actions equal in game, just the feedback its getting.

This is fundamentally different than an AI playing Doom as the AI approach is to run every combination of input signals and catalog which combinations reached the end of the level. It will then only use these combinations of inputs to play. Again it has no idea what these inputs do only that a specific set of inputs results in the desired outcome.

In the end both achieve the goal of playing the game, one based on statistical probability and the other "learning" through trial and error.

 

[jayNiowa]

View attachment 76388

I read about that...it's interesting to say the least.