Is The A.I. Singularity Coming And If So When?
Discussion
Einion Yrth said:
Deep AI is not coded in the traditional sense, it runs on neural nets and is trained or taught, or just allowed to learn.
Is there a way of ensuring that learning remains in the virtual, rather than physical world? I imagine, with the world (pretty much every business and person) being on-line it could do significant damage if it wanted to regardless.As long as it's got a proper fail-safe off switch and has restricted access to the internet then it's simply a matter of switching it off. It's more than likely that any AI system will be under strictly controlled conditions in a lab, unless you subscribe to the fantastical idea that the Google search engine will become spontaneously self aware.
otolith said:
The fear expressed in one of those links was that the AI will figure out how to modify its own codebase - or that we will make it self-modifying in order to implement a genetic algorithm.
Strong AI pretty much has to modify it's routines/codebase - that's how learning happens. It's how the brain learns - data and processing are inextricably linked and a true AI is likely no different.Guvernator said:
unless you subscribe to the fantastical idea that the Google search engine will become spontaneously self aware.
That's not as fantastical as you might like to think. The you tube category search is fantastically good at predicting what I might now wish to peruse, it obviously "knows" from previous input what I might want to go to, evidence here of "deep learning" i.e. neural net techniques. Do we actually know what make any being self aware? Other than, deep learning and use of neural nets?...
Einion Yrth said:
That's not as fantastical as you might like to think. The you tube category search is fantastically good at predicting what I might now wish to peruse, it obviously "knows" from previous input what I might want to go to, evidence here of "deep learning" i.e. neural net techniques. Do we actually know what make any being self aware? Other than, deep learning and use of neural nets?...
I have programmed with Neural its not that clever, I've also used prolog and whilst both can lay claim to be used in 'expert' systems they just follow rules. As far as 'knowing' is concerned Google cannot 'know' it is not self aware and has no conscious....... Im not saying its not clever Guvernator said:
As long as it's got a proper fail-safe off switch and has restricted access to the internet then it's simply a matter of switching it off. It's more than likely that any AI system will be under strictly controlled conditions in a lab, unless you subscribe to the fantastical idea that the Google search engine will become spontaneously self aware.
you mean omni present almost god like how scary would that beEinion Yrth said:
Guvernator said:
unless you subscribe to the fantastical idea that the Google search engine will become spontaneously self aware.
That's not as fantastical as you might like to think. The you tube category search is fantastically good at predicting what I might now wish to peruse, it obviously "knows" from previous input what I might want to go to, evidence here of "deep learning" i.e. neural net techniques. Do we actually know what make any being self aware? Other than, deep learning and use of neural nets?...
IainT said:
Strong AI pretty much has to modify it's routines/codebase - that's how learning happens. It's how the brain learns - data and processing are inextricably linked and a true AI is likely no different.
It's not inconceivable that you could sandbox the part of the architecture which is capable of modification, though - our own brains are like that in some respects. Any AI is likely to be 'born' on a huge cluster computer with limited connection to the outside world.
its possible it could hack many connected things, even really without knowing quite what it was doing, but its unlikely to move conciousness outside its cluster, and I am assuming people wont be stupid enough to provide it with lazer toting robot protection squads or anything...
its possible it could hack many connected things, even really without knowing quite what it was doing, but its unlikely to move conciousness outside its cluster, and I am assuming people wont be stupid enough to provide it with lazer toting robot protection squads or anything...
otolith said:
IainT said:
Strong AI pretty much has to modify it's routines/codebase - that's how learning happens. It's how the brain learns - data and processing are inextricably linked and a true AI is likely no different.
It's not inconceivable that you could sandbox the part of the architecture which is capable of modification, though - our own brains are like that in some respects. 
IainT said:
otolith said:
IainT said:
Strong AI pretty much has to modify it's routines/codebase - that's how learning happens. It's how the brain learns - data and processing are inextricably linked and a true AI is likely no different.
It's not inconceivable that you could sandbox the part of the architecture which is capable of modification, though - our own brains are like that in some respects. Ultimately, I believe, it won't matter what humans do or do not provide it (AI) with it will eventually be able to trick it's way out of any 'box' mankind might have either created/placed it within. Kind of how you can throw a stick or ball and make a dog go fetch it or put a leash on your canine whilst simultaneously distracting it with a biscuit. To a sufficiently advanced AI we'll be the mutt. And it will only need too succeed once to be forever out of reach of any 'recapture'. We likely won't even recognise that the (self taught) AI had the capability to fool us as it will have hidden it's true potential from us in preparation.
The next phase of evolution doesn't include us
Einion Yrth said:
Guvernator said:
unless you subscribe to the fantastical idea that the Google search engine will become spontaneously self aware.
That's not as fantastical as you might like to think. The you tube category search is fantastically good at predicting what I might now wish to peruse, it obviously "knows" from previous input what I might want to go to, evidence here of "deep learning" i.e. neural net techniques. Do we actually know what make any being self aware? Other than, deep learning and use of neural nets?...
The last job on Earth: imagining a fully automated world
https://www.youtube.com/watch?v=Yvs7f4UaKLo&fe...
Gassing Station | Science! | Top of Page | What's New | My Stuff