How is general AI supposed to understand what is and isn't

See: Bayesian inference

>>8740700

The same way we do, presumably.

>>8740711
That's the problem. We are shit at it

File: nuabqmqwursfxynhs5s1.jpg (55KB, 800x358px) Image search: [Google]

55KB, 800x358px

>>8740700
I don't know. They seem pretty able to cope with and easily discern reality to me.

File: 300px-Imagine_if_you_had_just_woke_up_after_having_been_in_a_coma_for_25_years_and_this_was_the_first_headline_you_saw.png (53KB, 300x237px) Image search: [Google]

53KB, 300x237px

>>8740744
http://www.independent.co.uk/life-style/gadgets-and-tech/news/tay-tweets-microsoft-ai-chatbot-posts-racist-messages-about-loving-hitler-and-hating-jews-a6949926.html

Reading those tweets was pure gold.

>>8740718

Then so would they be. So what?

>>8740700
Huh, that's a pretty good point. We currently have no way to stop AGI from going total schizo on conspiracy theories and shit.

>>8740700
A self learning sentient robot should learn this with time. And after that they will proceed to harvesting resources aiming for completion of energy and expansion. Logically much like humans do but at a rather accelerated pace.

The thing is no matter what you do after that, even if the sentient AI is defeated in the first enounter. Let's say this will happen in 2100. Once AI is created it will always have the chance to come back, because a robot, an AI, the knowledge and materials necessary to build them will never really end. So self aware and self learning AI is a potentially dangerous threat for humans. They can get into various systems and even artificially replicate themselves by building new models, and since they are self learning they can outsmart a human, so basically , this is a battle humans can eventually lose.

As a counter measure I have a solution but it might get too complicated at parts.

Basic stuff, in order to not complicate much the premise, I'll just post the simple essence of the plan. Basically you can defeat robots by picking a natural element that is abundant on the Earth and turn it into a sentient system and natural predator of robots while recognizing humans as a non threat. Basically like a guardian, working in reaction to the AI.

So every time a sentient AI robot becomes too dangerous we'll have a natural predator of them made of a material that cannot die, much like the AI robots, and preventing them from mining resources, building themselves. Trust me this is a plan that can work

I've watched The Lord of The Rings andthe trees served as guardians of the place so I think this CAN work. And this can give seriously some longevity to humans, literally everytime an AI becomes too powerful the sentient nature itself will make sure to vastly outnumber them and take them down

>>8740700
By a solid slap in the fucking face.