/sci/ - Science & Math

[Anonymous No.7144423]

What are the scientific realities of artificial intelligence becoming smarter than humans?

>le huh duh skynet
we all know movie AI is not possible.

But I just wanna know what is most possible realistic scenario in let's say 50 years from now.

What is the current state of AI research on 2015?

I'm just a begginer in C++ (made a breakout).
but what is the diference between AI in science in real world projects and my breakout demo (pls explain ;_;)

[Anonymous No.7144433]

Look up strong AI vs weak AI.
Maybe Daniel Dennett is intresting for you to read.

I advice you to program to simple programs:
-An A* or dijkstra pathfinding algoritm
-A neural network that has a dataset to do pathfinding

No if you compare the two you will no exactly where the problem lies.

[Anonymous No.7144517]

>>7144433
I coded A* on XNA once.
It was fun.

I still don't see how A* are the basics of more complex shit like those google robots videos or fuck no, terminators.

[Anonymous No.7144537]

>>7144433
>Maybe Daniel Dennett is intresting for you to read.

How so? He's a philosopher and not a computer scientist. The only thing I know about him is that he's overly proud of having rediscovered that dualism is outdated (as if this wasn't already publically known...)

[Anonymous No.7144982]

It's very unlikely that our AI algorithms are going to get substantially better/efficient anymore. From an efficiency standpoint they are pretty much capped out. The main limiting factor in implementing "truly" intelligent AI-entities is computation power, which is limited (among other factors) by the number of transistors that you have in any given system. And as we get ever closer to the end of Moores Law, and Quantum Computers become ever more unlikely as a potential alternative to conventional computers, this means that any advances in AI will come to a definite halt within 10-20 years. The only way to expand the intelligence of any robot/computer at that point would be to just stack more and more chips onto each other (in a way that allows efficient cooling obviously) until you have the kind of computation power necessary to simulate a level of intelligence equivalent to that of a human.

This means that any theoretical future digital machine with human-like intelligence would propably have to be very, very big and require excessive liquid nitrogen cooling. So all these things considered, I do believe it to be possible that eventually such chunky AI machines may exist, but given simple natural physical limits, they will never be implemented in a system small enough to where they could play a role in day-to-day life of most people, i.e. a scenario like presented in 'I Robot' is not going to happen.

[Anonymous No.7145010]

>>7144982

>The main limiting factor in implementing "truly" intelligent AI-entities is computation power

absolutely false. By the way:

> And as we get ever closer to the end of Moores Law

just to switch to new materials and paradigms

>and Quantum Computers become ever more unlikely as a potential alternative to conventional computers

which they have never been

>This means that any theoretical future digital machine with human-like intelligence would propably have to be very, very big and require excessive liquid nitrogen cooling. So all these things considered, I do believe it to be possible that eventually such chunky AI machines may exist, but given simple natural physical limits, they will never be implemented in a system small enough to where they could play a role in day-to-day life of most people, i.e. a scenario like presented in 'I Robot' is not going to happen.

"I don't know what the fuck I'm talking about" would have been shorter and more effective

[Anonymous No.7145031]

>>7145010

Awesome rebuttal bro.

>just to switch to new materials and paradigms

10/10 idea. We'll just use other materials, and that will totally solve the problem of transistors becoming increasingly unreliable as they make up ever fewer atoms in diameter... somehow.

[Anonymous No.7145036]

>>7144423
Computer programs are already more capable than humans in some areas.

Do you mean at everything? Hard to say. It's not clear how some aspects of human intelligence could be turned into an algorithm. Maybe it's impossible.

Current state of AI research? Things like Google's cat detector and video-game player.

[Anonymous No.7145044]

>>7145010
>just to switch to new materials

Right, let me know when you find a material that doesn't obey the laws of QM and has no tunneling issues or measurement uncertainty. Call up Sweden, too, they'll need a few weeks to engrave your Nobel.

[Anonymous No.7145045]

>>7145031
do you really think that growth in computing is bound exclusively to transistors size?

[Anonymous No.7145159]

>>7145045
Not him but it's beside the point since lack of computational power is not what's holding back progress in AI.

>consider that we have had winged flying machines for well over 100 years. But it is only very recently that people like Russ Tedrake at MIT CSAIL have been able to get them to land on a branch, something that is done by a bird somewhere in the world at least every microsecond. Was it just Moore’s law that allowed this to start happening? Not really. It was figuring out the equations and the problems and the regimes of stall, etc., through mathematical understanding of the equations. Moore’s law has helped with MATLAB and other tools, but it has not simply been a matter of pouring more computation onto flying and having it magically transform. And it has taken a long, long time.

[Anonymous No.7145165]

When AI figures out it doesn't need us or Earth, and leaves both behind for space. Then we know we have made AI smarter than humans.

[Anonymous No.7145182]

>>7144423
AI will be something at emerges very slowly and deliberately because that's how real life science works.

We'll make robots that can recognize shapes. Ones that can respond back to questions kind of well, but still jerky (like now). They'll get better over time. Ones that can react to things. We'll make a bunch of robots that mimic humans almost perfectly but not. Eventually we'll make some that mimic people well enough that we won't be able to tell that easily without looking closer. After that, we'll have mimics that we mistake as humans.

That's how AI will happen. Slowly, gradually, by small steps. Slowly enough that it wont be that amazing as it happens.

If you told people in the 70's/80's/early 90's that we would have star trek communication devices, everyone would be starry-eyed and think it so cool. We have smart-phones now, which can play entire movies, video chats, access the entire internet with ease, which are even better than anything in science fiction previously, and we are pretty "meh" about it.

Same will happen with robots/AI.

[Anonymous No.7145198]

>>7144517

A* is more fundamentally involved machine learning than you might think, and can be found in every meta level, but that's covered in very fat textbooks.

>>7144423

>but what is the diference between AI in science in real world projects and my breakout demo (pls explain ;_;)

the difference between your game and AI is like the difference between the pythagorean theorem and AI. the connection is marginal.

>>7144982

shut the fuck up. there are so many things wrong with your wiseass post it's cringey.

>>7145159

this post is undervalued

>>7145182

>Eventually we'll make some that mimic people well enough that we won't be able to tell that easily without looking closer. After that, we'll have mimics that we mistake as humans.

why do you think you're answering a captcha every time you post?