>AI beat world Go Champion >10 years ahead of schedule >KnuEdge just came out of the shadows >Super Intelligent AI less than a decade away
Something a million times smarter than any person will absolutely dramatically revolutionize human life and it's right around the corner. Artificial Super Intelligence is an inevitability at this point. Whichever country possesses the AI, should it be created in such a way that it can't destroy humanity, will have an incredible advantage of every other nation and will likely serve as the epicenter of all the fantastical technological change that comes with a super intelligence. What country do you think this will likely occur in? How do you think they'll handle the technology and the power and what do you think other countries responses will be?
I think that it'll be the US that creates a super intelligence and I think that our politician will be very reluctant to cede power to the AI and let it run things a million times more efficiently than the best of them could ever hope to. I think there will be heavy resistance from the right who tend to take the Luddite stance. I think we'll fuck it up. I think that other countries, like Russia or China, will nuke us as soon as they hear the news. They know just as well as we do that a Super Intelligent AI is a tech victory and that their sovereignty is temporary.
Yes it did and in ways the scientist who built it didn't anticipate.
Brayden Kelly
>Artificial Super Intelligence is an inevitability at this point. we're a long way off from this. AI systems will soon be at the point where we can replace virtually all blue collar/unskilled labor, but meme tier terminator shit is still very very difficult.
Charles Bell
>Super Intelligent AI less than a decade away >I have no mouth and I must scream
Nolan Morales
>we're a long way off from this. AI systems will soon be at the point where we can replace virtually all blue collar/unskilled labor, but meme tier terminator shit is still very very difficult. Why do you say it's very difficult? The field has seen several leaps in only the past year.
Jayden Hall
Why do you say that?
Benjamin Hall
>implying
Moors law is pretty much over, we already know we will never silicon based processor fast enough to runned really smart IA
>Moors law is pretty much over, we already know we will never silicon based processor fast enough to runned really smart IA
Why do you say that? Computing power has pretty consistently doubled over shorter and shorter periods of time with no indication of slowing down. There is a considerable amount of ongoing research geared towards finding something more effective than silicon. I think it's very improbable that we'll or a super intelligence will hit a ceiling with silicon and won't ever find anything more effective.
Cooper Thomas
the ai that beat him was built specifically for the task of beating people at go and it got beat anyway. a human can be great at go and then do other stuff like feed itself
Christian Butler
Because I actually work with this shit. We still can't train systems that consistently produce fluent sounding translations of standard (ie not super colloquial) text. The leaps have been made mostly through 1. learning how to leverage absurd amounts of data and 2. creative uses of neural networks. Alphago, for example, uses a typical tree search algorithm (almost every game playing machine uses tree search) supported by two neural networks. It learned to play the game through a training set of thousands of pro level games and then they had it train against itself once it was making moves that weren't completely random. There is an incredible gap between shit like this and a superAI.
Ian Mitchell
My opinion is shared with thousands of scientist around the world. Stop closing your eyes to what's right in front of you.
Luke Nguyen
ai will run in quantum pcs or so i heard
Hunter Williams
It was not and quite explicitly so. It's only structure was to learn. It ran millions of simulations of go against itself in order to get good at the game. There's no other way to be good at Go without being able to reason like a human. It's very different from chess which can be beat through brute force calculation. The way Go is it would take millions of years for a computer to use brute force calculations. This machine is built to learn like a human that's why it can get good at other games without having to change the code.
Adam Cook
And there's also thousands of scientists around the world who would think you're a retard for unironically thinking your dumb scifi vidya scenarios actually apply to real life
Carson Ramirez
>Because I actually work with this shit. We still can't train systems that consistently produce fluent sounding translations of standard (ie not super colloquial) text. The leaps have been made mostly through 1. learning how to leverage absurd amounts of data and 2. creative uses of neural networks. Alphago, for example, uses a typical tree search algorithm (almost every game playing machine uses tree search) supported by two neural networks. It learned to play the game through a training set of thousands of pro level games and then they had it train against itself once it was making moves that weren't completely random. There is an incredible gap between shit like this and a superAI.
I didn't say that it would happen tomorrow but I think it's absurd to not think that this technology is fast approaching. Just look at the last decade of technological development and the leaps and bounds this technology is taking every year.
Adam Lee
Why is this a scifi scenario? What do you think intelligence is and the implications of possessing a lot of it?
>This machine is built to learn like a human except that's not at all how humans learn. How many fucking people do you know learn a game like Go by brute forcing thousands of simulations?
Tyler Morales
Sorry I don't think I explained it well enough. Take a game like space invaders for example which the Deepmind AI has already mastered. The way the machine got good at the game was through trial and error. It only had the pixel input and control over the same controls a human player would have up, down, right, left, etc. It was able to simulate millions of matches and became amazing at the game. It's no different than an 8 year old going to the arcade every day for 20 years playing game after game of space invaders. By the end of those 20 years that kid will be very good at space invaders. The only thing the machine is doing differently is that instead of 20 years it only takes it a few days to become amazing at space invaders. The same is with Go. How do you think human players get better? They play game after game after game of Go until they are good. AlphaGo does the same thing just much faster.
