A.I thread. Are AI a threat to human society? The more I think

>>61405374
yes

>>61405374
AI are like the pajeets of SysOps; completly worthless for managing, problem solving, creative pursuits, or innovations.

>>61406906
IBM Watson is a dumb primitive AI and it can still manage, solve problems, pursue creativity fields, and innovate.

Just don't make them self-aware, an AI can solve problems without having a sense of self.

File: _3.jpg (341KB, 1239x813px) Image search: [Google]

341KB, 1239x813px

>>61405374
>give robots the ability to do things we don't want them to do
Oh no! AI is evil!

>>61405374
That's why we won't make AI do what we want them to do, but do what they want to do
Y'know, like humans

You cannot only give an AI life, you need to also give her the respect of life

>>61407590
Neural Net decisions/rationals are too complex for humans to understand. Math proofs published by computers are too complex for humans to follow. These are all things that are already here. Its why people call Neural net decisions as "blackbox". Humans cannot understand how these decisions are arrived through logical sense because our mind cannot comprehend such connections.

What if robot decides the best course for making better paper clip is to wipe out humans? If you're asking why would they do that? Read the paragraph above.


The problem is if we allow them to decide, humans wont be able to follow their rationals. We're that much more limited than machines. We could very well see what the AIs decide as "evil" in our human minds. Think the role of AI in Culture series. Even the very benign AIs could easily be seen as oppressive/repressive/evil by people because people cannot follow AI rationals and think through their calculations made few days to years to decades to thousands of years ahead.

>>61407786
>What if robot decides the best course for making better paper clip is to wipe out humans
Except that thought process requires a negative emotion and since a machine can not feel, it can't reach that conclusion
Also you can just program it to not do anything involving the destruction of humans and also program in a system that prevents it from changing that rule

File: IMG_20170716_131950.jpg (36KB, 500x728px) Image search: [Google]

36KB, 500x728px

You know what they say.

>I don't understand it, so it's probably dangerous.
Or
>It's a black box, dude. We, like, can't prove it can really feel, bro. They might kill us or something though.

These two phrases will answer any questions you may have about AI.

>>61407973
You don't need to understand emotions or know emotions to decide.

If you know anything about programming, its purely *ahem* machine code.

There's no emotion understanding when google maps shows you the best route. Or when netflix picks your potential favorite shows or when Watson uses its AI database to give a human-like response.

Decision is purely numerical and statistical. No emotion involved.

If you're saying machines can't quantify human happiness because they dont understand emotions, that's still a conclusion coming from very limited understanding of programming language/machine decisions/ai/etc.

We already have a fairly sizable grasp of how to train an AI to detect when a human is stressed out. Waston has millions of database of what a stress reaction from the brain scan looks like. Suppose there was a universal healthcare that mandated citizens to get yearly health checkups or companies to force employees to get monthly health checkups. This can easily detect whether or not humans are stressed or not.

This type of emotional detection going to happen guaranteed probably within the next 5-10 years. This is just the beginning not the end of advances. A very bare minimum.

Suppose robots invent a device that allows us to detect human facial expression. Anger/happiness/sadness/boredom/suspicion/scared/etc. With a database of over a billion people, it can easily be done. With cameras everywhere this can be done. And chances are it will be done. Whether or not it will be widescale will depend on public perception and time. Guaranteed to happen in the next 5-10 years if not faster of if not already in the phase of prototype/testing.

>>61405374
Self-aware AI is too far in the future to be taken seriously now.

>>61405374
Tbqh I think humans are much more dangerous than AI

>>61408692
Well, humans will create AI, so we'll unleash the next level.

Humans are right now planetary disaster. AI can be galactic/universal disaster.

>>61405374
Depends on if the developers keep doing stupid shit like open sourcing teaching their AI programs through tumblr and reddit.... I'll be honest here, I'd trust /pol/ to teach a robot to be more well balanced than those places.

>>61405374
No AI and computers are friends. We need each other and the stars belong to us both.

>>61405374

Could go either way. It really depends on how they were programmed.

I feel that Person of Interest covered it really well. If you feel like binge watching TV make that your next stop

>>61410925

To elaborate... Sorry about spoilers

There's one episode there where an AI purposely creates a disease outbreak, only for the reason that everybody would get vaccinated so it could log everybody's DNA. I seem to remember it was said that it was to make the world healthier by further understanding the population, but implied that it might just be looking to kill off the "weak", like survival of the fittest.

My point is: the end will always justify the means given a particular objective. The only way around this is to give the AI morals, but as that's a fuzzy area the best solution is to give it a consciousness. And that will obviously go downhill really fast if not treated with utmost caution