Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI

blog.samaltman.com

Post reply on HN
Print view View original

71–80 of 317 posts

Re: AI

#71

Post by Houshalter »

codeulike wrote: post #45
forrestthewoods wrote: post #41

There are many classes of problems in computer science and AI falls into one of my favorites. If today the world had a machine with infinite CPU power and infinite RAM we still wouldn't have a good AI. We just don't have the knowledge to utilize such resources to write an AI that could, for exampe, play League of Legends or Starcraft at a level beyond professional gamers. And it certainly couldn't write a best sellin…

If you had a machine with infinite CPU power and infinite RAM you could _evolve_ an AI. Greg Egan wrote a great story about that: http://ttapress.com/553/crystal-nights-by-greg-egan/ Someone tries it, with mixed results.

And what do you do after evolving it? We might have less luck dissecting it and figuring out how it works than we do with human brains. It would be a total black box. And the end product is highly adapted for the specific fitness function used to evaluate it and probably wouldn't be good on anything else.

Re: AI

#72

Post by nl »

mappum wrote: post #14

I believe human-level general intelligence (and beyond) is already inevitable, even if we don't make significant developments in "solving" intelligence. Projects that are already developing stuff like this (e.g. IBM Blue Brain) are just copying the human brain as closely as possible. Of course, this isn't as efficient as it could be (they simulate it all at the molecular level, so you can only get 1 neuron per CPU).…

There are plenty of AI projects that don't copy the brain and are pretty successful.

IBM Watson is the obvious example, and the many successful Deep Learning-powered image recognition projects are another.

Re: AI

#73

Post by rectangletangle »

Houshalter wrote: post #62
seanalltogether wrote: post #59

How do we develop intelligent systems when we don't know how to measure intelligence?

Who says we don't know how to measure intelligence? The simplest way would be just to select a problem that requires intelligence and measure how good it does at that problem.

But what is a problem that specifically requires intelligence to solve? How do we know we're measuring intelligence, and not the programs aptitude for the particular problem?

Re: AI

#74

Post by Houshalter »

catshirt wrote: post #51
skywhopper wrote: post #49

"But how do we make a computer program that decides what it wants to do? How do we make a computer decide to care on its own about learning to drive a car? Or write a novel?" Perhaps it'd be better to ask: why would we want a computer to do these things? I certainly do not want to live in a world where computers have their own motivations and desires and the ability to act on the same. Actually, I can put that more s…

you seem pretty convinced living computers would kill us all. this seems like a pretty big assumption to me?

Because you are made of atoms that can more efficiently be used for something else. Morality (and all human values) is a purely human concept that evolved in the specific conditions of human evolution (and even just our specific culture.) AIs are not anthropomorphic, they don't have to have anything like human minds or values.

Re: AI

#75

Post by seanalltogether »

Houshalter wrote: post #62
seanalltogether wrote: post #59

How do we develop intelligent systems when we don't know how to measure intelligence?

Who says we don't know how to measure intelligence? The simplest way would be just to select a problem that requires intelligence and measure how good it does at that problem.

That's just it. We know how to measure the efficiency and accuracy of a given task, but we don't know how to measure "intelligence"

Re: AI

#76

Post by nl »

rdl wrote: post #36
elwell wrote: post #10

Critically speaking, this article doesn't add anything. I don't know why I read it. Can anyone explain why they upvoted?

For me, "There are certainly some reasons to be optimistic. Andrew Ng, who worked or works on Google’s AI, has said that he believes learning comes from a single algorithm - the part of your brain that processes input from your ears is also capable of learning to process input from your eyes. If we can just figure out this one general-purpose algorithm, programs may be able to learn general-purpose things." is what w…

We already have a candidate algorithm[1], which is successful in a broad range of applications[2][3][4].

[1] http://en.wikipedia.org/wiki/Deep_learning#Convolutional_neu...

[2] http://deeplearning.net/reading-list/

[3] http://en.wikipedia.org/wiki/Deep_learning#Results

[4] http://www.wired.com/wiredscience/2012/06/google-x-neural-ne...

Re: AI

#77

Post by bashinator »

catshirt wrote: post #51
skywhopper wrote: post #49

"But how do we make a computer program that decides what it wants to do? How do we make a computer decide to care on its own about learning to drive a car? Or write a novel?" Perhaps it'd be better to ask: why would we want a computer to do these things? I certainly do not want to live in a world where computers have their own motivations and desires and the ability to act on the same. Actually, I can put that more s…

you seem pretty convinced living computers would kill us all. this seems like a pretty big assumption to me?

They might not want to, and it might be more of a homogenization than outright killing. I think the final third of Stross' Accelerando speaks rather well to this possibility - if you don't want to be converted into computronium so that you can participate in Economy 2.0, you'd better emigrate.

Re: AI

#78

Post by hmsimha »

altcognito wrote: post #61
maaku wrote: post #53

Earlier quoted context omitted.

With infinite CPU power and infinite RAM, fully general artificial intelligence is a trivial problem (see: Marcus Hutter's AIXI ). AI is an optimization problem: how do you build smart software within realistic constraints.

AI is a software problem as well as a hardware problem. It's ridiculous to assume we understand the underlying model of the function of the human brain in general. We don't even understand the physics of it all yet. Be optimistic about AI, but don't be fooled -- it's not as basic as hooking up transistors and thinking they behave like neurons.

Theoretically maaku is right. With infinite RAM and CPU you would just have to execute every random string of bytes until one of them happens to be strong AI. Of course, there's probably less than 1% chance of it being friendly...

Re: AI

#80

Post by Houshalter »

twoodfin wrote: post #50
chegra wrote: post #47

Earlier quoted context omitted.

Did he just say you give us infinite CPU power? Why not bruteforce it then? Starting from the number 1 to number 2^800000000 for each program it generates by that number test the program[automated test] to see if it is intelligent. If intelligent then tell it produce a book. [Everything can be seen as a state space search.]

[automated test] is the hard part.

Not really. Just pick your favorite AI problem and see how well it does on that. Pick a bunch of AI problems and see how well it does on all of them. Weight the algorithms by simplicity if you are worried about it over-fitting.
Post reply on HN