Discussion
ikarl said:
Sorry, not sure I know what you mean when you say 'hokey websites', but I did read an article in The Economist a few months ago that made me think - http://www.economist.com/news/briefing/21650526-ar...
I'm not an expert and don't profess to be one, but looking at the progress of computers now compared to 20 or 30 years ago the bounds are impressive. Pure computing ability has multiplied many times, computers have got smaller, faster and more energy efficient. I also believe there's something called 'Moore's Law'(?) which has been pretty consistant over the years and progress has been pretty much on track with that.
If a computer was tasked with improving itself and had free reign to do so, do you not think it would be possible for it to grow and improve? Could it learn? Could it learn how to understand (like a human brain)?
That economist article is a regurgitation of stuff we have been seeing\reading for years and is surprisingly sensationalist for that paper.I'm not an expert and don't profess to be one, but looking at the progress of computers now compared to 20 or 30 years ago the bounds are impressive. Pure computing ability has multiplied many times, computers have got smaller, faster and more energy efficient. I also believe there's something called 'Moore's Law'(?) which has been pretty consistant over the years and progress has been pretty much on track with that.
If a computer was tasked with improving itself and had free reign to do so, do you not think it would be possible for it to grow and improve? Could it learn? Could it learn how to understand (like a human brain)?
Computers are improving all the time but only because we are improving them. Could a computer learn how to learn or improve itself one day? Sure it could but it would need at least a base level of intelligence in order to be able to do that and more importantly, we'd still need to feed it power, data, storage space and increased processing capacity otherwise how would it grow? Where would it get the spare capacity to grow into beyond a certain point. A computer has physical limitations just like everything else. If you want to improve your computer you have to physically put more ram\storage in it, how will that change in the future? Even a very smart AI can't just magic this stuff up.
Man has the smartest computer yet invented in his skull but even that has limitations as to how much it can improve, remember and learn. Beyond a certain point it can't just make itself faster\smarter out of thin air. The progress of a learning AI would be a long slow process much like how a real human grows, improves and develops. It would take years and it would need external assistance in both how it develops and the physical requirements which we would directly control.
The idea that a computer could go from simple sentience (animals are sentient) to super-being and then get out of control and do a Skynet\try to take over the world, all within a couple of hours of being brought online is a little bit far-fetched.
Guvernator said:
Sorry I think you've been reading too many hokey internet websites (I probably know exactly which one) and watching too many bad sci-fi movies.
Computers don't work like that. A sentient AI might be able to self improve but only within a defined limit and set of parameters. How would it acquire more data or get access to more processing power for a start? More importantly what about the increased power\energy requirements that would require. It will be in a totally controlled and isolated environment.
Even if someone hooked it up to the internet, how would it know how to interpret all that data in a meaningful way? Humans have the most efficient "cpu" ever discovered and it still takes us years to learn how to interpret all the data we receive. We also require external help, other people teach us how to interpret data, how would a sentient AI "teach" itself. Would you be able to carry out brain surgery or build a new computer processor just by reading about it on the internet? If allowed AND given the resources, yes it may develop to a point where it might eventually surpass us but this will take many years and unless you gave it complete control of the physical processes which allowed it to just keep adding more storage, processing power and energy to itself (very doubtful) that learning process would always be restricted by what we allowed it to have access to in terms of physical resources. Getting smarter requires a lot more than just the ability to think faster.
So while I certainly enjoy the books and movies about this stuff, the idea of AI becoming sentient then developing God-like powers or running amok within a few hours is pure sci-fi fantasy.
Not really.Computers don't work like that. A sentient AI might be able to self improve but only within a defined limit and set of parameters. How would it acquire more data or get access to more processing power for a start? More importantly what about the increased power\energy requirements that would require. It will be in a totally controlled and isolated environment.
Even if someone hooked it up to the internet, how would it know how to interpret all that data in a meaningful way? Humans have the most efficient "cpu" ever discovered and it still takes us years to learn how to interpret all the data we receive. We also require external help, other people teach us how to interpret data, how would a sentient AI "teach" itself. Would you be able to carry out brain surgery or build a new computer processor just by reading about it on the internet? If allowed AND given the resources, yes it may develop to a point where it might eventually surpass us but this will take many years and unless you gave it complete control of the physical processes which allowed it to just keep adding more storage, processing power and energy to itself (very doubtful) that learning process would always be restricted by what we allowed it to have access to in terms of physical resources. Getting smarter requires a lot more than just the ability to think faster.
So while I certainly enjoy the books and movies about this stuff, the idea of AI becoming sentient then developing God-like powers or running amok within a few hours is pure sci-fi fantasy.
In fact Ray Kurzweil at NASA and others completely disagree with almost all of that. Read "The Coming Singularity". Whilst not perfect this book got the scientific community thinking. Indeed, Hawking, Gates, Musk etc are all warning about the singularity now. You have no idea what a sentient computer hooked up to the Internet and using it for both knowledge and self transportation/dissemination (initially discreetly ie without your knowledge) could achieve - which is the whole point of creating an AI in the first place - for it to be able to come up with answers and solutions we can't. It's a catch 22.
Whilst "running amok within a few hours" is perhaps overstating it a mite it is now thought that an emergent AI would come from behind us to overtake us like an F1 car at full tilt. It will likely be past us before we've even had time to realise what's happening. This is the genuine fear now being expressed by noted scientists and others.
"How will it get more processing power?" If sentient and conversing it could for example tell you that it requires more processing power in order to crack the problem of (say) Cancer or to double the human lifespan. It may not even tell you, it may fail to answer the problem leading YOU to assume that with more processing power it could get it right.
You also say "allow it to have access to". Once its on the Internet there are 7 billion human beings out there who it can bribe/fool/scam/blackmail to do what you might think were the most innocent of jobs but are part of a plan we couldn't even begin to comprehend the complexity of. Christ some humans are even today still handing over there bank details to Nigerians over the phone. For an AI this will be like taking candy from a baby - and it only needs 1 person to be suckered in with promise of (say) 'the secrets of the stock market' on one day at one point in time for the whole of humanity to fall some short time afterwards.
Do not think that any AI would necessarily be sitting in an isolated box on a desk somewhere either. First of all an emergent AI may come from many smaller programs at many locations hooking up to each other. Although it is more likely to be in one location and to that end the likes of Google and Apple are already allowing their proto AI's internet access - it's the only way to develop and teach an AI about the world around it - which it'll need to solve our problems.
In summary, this thinking that we can control an emergent AI is like the native Americans thinking that they can restrict the technologically superior early European settlers to remaining in certain areas along the coast only. Foolish at the very least.
Edited by mudflaps on Wednesday 8th July 16:14
I've read a lot of that stuff and there are some VERY big if's in a lot of what is put forward, the least of which is why do we immediately assume that any AI will be duplicitous, malevolent and have some nefarious scheme at heart? What would an AI want with someone's bank account details anyway and unless an AI spontaneously becomes self aware (not sure how that would even be possible) it's every state and communications will be being controlled and monitored with a fine tooth comb so even if it had an inkling to (highly doubtful) it's not going to be calling you to get your bank details without it triggering a hundred alarms, it's just code after all. It's going to be communicating down a wire which is traceable, not using woo woo.
I love science fiction, I probably read or watch some at least a couple of times a week but I realise it what for what it is, fiction. I think Kursweil et al while certainly putting forward some interesting premise, need to step back from the sensationalist "machines will rule the earth" mantra that seems to have infected the human psyche. We are putting human labels and emotions on something when we have no idea how it would behave. An AI will probably not even have emotions as it will be born of pure logic, why would it care about trying to take over the world?
To be honest I expect a bit more objectivity from a NASA scientist but then "we achieve singularity and everything works out great" isn't as exciting as "the end of the world is nigh, bow to our robot overlords". I'd like to think that if some clever bods do finally crack true AI, they would be a lot more scientific and objective in their approach.
I love science fiction, I probably read or watch some at least a couple of times a week but I realise it what for what it is, fiction. I think Kursweil et al while certainly putting forward some interesting premise, need to step back from the sensationalist "machines will rule the earth" mantra that seems to have infected the human psyche. We are putting human labels and emotions on something when we have no idea how it would behave. An AI will probably not even have emotions as it will be born of pure logic, why would it care about trying to take over the world?
To be honest I expect a bit more objectivity from a NASA scientist but then "we achieve singularity and everything works out great" isn't as exciting as "the end of the world is nigh, bow to our robot overlords". I'd like to think that if some clever bods do finally crack true AI, they would be a lot more scientific and objective in their approach.
rsv696 said:
I enjoyed it. For those into Nordic Noir this was loosely based on a Swedish drama series from a couple of years back (includes subtitles) : https://vimeo.com/album/2747482
^^^ Swedish version here 
JustinP1 said:
warp9 said:
SpudLink said:
I guess this would be the wrong place to explain my theory that the Internet is already self aware, and biding it's time before turning against its makers.
Go on, I'm curious! 
ikarl said:
JustinP1 said:
warp9 said:
SpudLink said:
I guess this would be the wrong place to explain my theory that the Internet is already self aware, and biding it's time before turning against its makers.
Go on, I'm curious! 
Gassing Station | TV, Film, Video Streaming & Radio | Top of Page | What's New | My Stuff