Why is AI such an impossible goal?
Discussion
Terminator X said:
Why wish for this as we're fked once it turns up.
TX.
Not necessarily, it depends on many factors, but your 50% right, or 50% way off. If you read the Waitbutwhy piece linked to earlier it talks about the impact of real AI on humanity and talks about it either being our doom, or our saviour. AI is likely to end up in one of two ways. It could be so good and could result in extremely good nano-technology that we basically become immortal the tech will exist to keep us going and repair us at the atomic level.TX.
Or yes, it could form a way of wiping us out completely if it wanted to.
It depends on how it's safeguarded, and also who gets there first and what the "parent" of real AI's motivations are.
I personally don't think we will see true AI until we have cracked quantum computing.
Although technology like this may get us part of the way there.
https://en.wikipedia.org/wiki/TrueNorth
Although technology like this may get us part of the way there.
https://en.wikipedia.org/wiki/TrueNorth
AshVX220 said:
Terminator X said:
Why wish for this as we're fked once it turns up.
TX.
Not necessarily, it depends on many factors, but your 50% right, or 50% way off. If you read the Waitbutwhy piece linked to earlier it talks about the impact of real AI on humanity and talks about it either being our doom, or our saviour. AI is likely to end up in one of two ways. It could be so good and could result in extremely good nano-technology that we basically become immortal the tech will exist to keep us going and repair us at the atomic level.TX.
Or yes, it could form a way of wiping us out completely if it wanted to.
It depends on how it's safeguarded, and also who gets there first and what the "parent" of real AI's motivations are.
TX.
SystemParanoia said:
The amazon Echo/Dot is pretty good at picking up on inference.
Alexa. Who is the Potus ?Alexa said:
Sex was a boutique run by Malcolm McLaren and Viviene Westwood at 430 King's Road, London between 1974 and 1976.
Alexa. Who is the Potus ?Alexa said:
The US President is Barack Obama.
Alexa. What is his wife's name?Alexa said:
Sorry. I don't have the answer to that question.
My AI lecturer told us nearly 15 years ago that we won't see anything more than minor (but often useful) improvements in specific areas of AI without some sort of revolutionary breakthrough, be it quantum computing, or something else and that it may never fulfill its promise.I remember thinking there was enough hype that that seemed an unusual opinion. These days I suspect he's right but I'm more still hopeful for that revolutionary breakthrough than he seemed to be.
otolith said:
Morningside said:
What/who is the latest best AI so far?
We're seeing advances made in single purpose AI for functions like voice recognition or game playing which are edging closer to the possibility of general purpose AI. For instance, the way that the Go challenge was won was to build a neural net, teach it to recognise winning patterns used by human masters and then have it play millions of games against other instances of itself and learn from those. Effectively it was taught to play and then sent away to think about it.
https://www.wired.com/2016/03/sadness-beauty-watch...
That's a very different thing to the algorithmic, mechanical way that chess was won.
Voice recognition is just that it is pattern matching no understanding and no consciousness, and pretty dumb as well. as for Neural Nets, they have their place as does other AI software.
Whilst what is refers to as AI software can be used in constrained environments it is not like a human, another example is Prolog which can underpin expert systems it can complement what and expert does but if your looking at an engineering design or medical application it can need tweaking before the final solution is output.
768 said:
My AI lecturer told us nearly 15 years ago that we won't see anything more than minor (but often useful) improvements in specific areas of AI without some sort of revolutionary breakthrough, be it quantum computing, or something else and that it may never fulfill its promise.
I remember thinking there was enough hype that that seemed an unusual opinion. These days I suspect he's right but I'm more still hopeful for that revolutionary breakthrough than he seemed to be.
Spot onI remember thinking there was enough hype that that seemed an unusual opinion. These days I suspect he's right but I'm more still hopeful for that revolutionary breakthrough than he seemed to be.
Toaster said:
Erm no
Voice recognition is just that it is pattern matching no understanding and no consciousness, and pretty dumb as well.
Hence making the distinction between narrow and general AI, the former of which is widely used, the latter as yet unsolved, and the relevance of the quote earlier about how once it can be done we don't think it's AI at all.Voice recognition is just that it is pattern matching no understanding and no consciousness, and pretty dumb as well.
768 said:
SystemParanoia said:
The amazon Echo/Dot is pretty good at picking up on inference.
Alexa. Who is the Potus ?Alexa said:
Sex was a boutique run by Malcolm McLaren and Viviene Westwood at 430 King's Road, London between 1974 and 1976.
Alexa. Who is the Potus ?Alexa said:
The US President is Barack Obama.
Alexa. What is his wife's name?Alexa said:
Sorry. I don't have the answer to that question.
My AI lecturer told us nearly 15 years ago that we won't see anything more than minor (but often useful) improvements in specific areas of AI without some sort of revolutionary breakthrough, be it quantum computing, or something else and that it may never fulfill its promise.I remember thinking there was enough hype that that seemed an unusual opinion. These days I suspect he's right but I'm more still hopeful for that revolutionary breakthrough than he seemed to be.
that will teach me to take what podcasters say as gospel. nearly purchased one on the back of that podcast! ( coderradio i think )
There do seem to be US/UK differences, I doubt the ability to ask questions based on the previous is one, but maybe.
It's a bit annoying that it keeps saying it couldn't find a device called saving light when I say ceiling light. Only once so far has it not been able to find a semen light!
For £40-50 I'd thoroughly recommend the Dot, especially if you have Hue lights or a Nest. But it won't take you long to find yourself discovering the limitations and acting accordingly.
It's a bit annoying that it keeps saying it couldn't find a device called saving light when I say ceiling light. Only once so far has it not been able to find a semen light!
For £40-50 I'd thoroughly recommend the Dot, especially if you have Hue lights or a Nest. But it won't take you long to find yourself discovering the limitations and acting accordingly.
It must be a bit demoralising being an AI researcher. Take Go for example. Up until very recently it was held up as a milestone in AI intelligence and thought to be a problem that would remain unsolved for a long time due to the enormous move search space. Researchers build an AI which can crush the greatest living player at Go and suddenly it's uninteresting and we're still in an AI dead end.
My perception as someone who admittedly has little deep knowledge about the subject is that we're right on the cusp of a total revolution in capabilities. AI systems are being seriously talked about to replace people in a lot of 'thinking' jobs. That's proper sci fi stuff.
My perception as someone who admittedly has little deep knowledge about the subject is that we're right on the cusp of a total revolution in capabilities. AI systems are being seriously talked about to replace people in a lot of 'thinking' jobs. That's proper sci fi stuff.
We'll never 'crack' Artificial Human Intelligence until we understand what human intelligence is and how it works. There is still much scientific debate as to how our neural structure is affected by external stimuli and the role of glial cells as both storers and processors of information. In short, we know little of how our brains actually work and the chances are it does not act as the 'straight' information processors current AI simulations are based on. Quantum computing may give us the computational capacity to mimic neural activity. But we need a clearer model of the stimuli/feedback mechanics of our brain chemistry on which to base its architecture on.
scorp said:
otolith said:
Yet we have biological machines in our heads which appear to do exactly that.
No one understands how they work though, so how would you begin to model something based on something you don't understand ?otolith said:
We don't know - but intelligence as we understand it is a property which can emerge out of something that works using nothing more than physics. We might not know how to do it, but it's clearly possible.
Could you explain that a bit further - as I understand it, intelligence is something that allows you to acquire knowledge and skills and apply them - a machine can't do that at the moment as none have yet developed the ability to understand the world around them. As yet there are no machines that can take in information and make use of it like a human can.
Monty Python said:
otolith said:
We don't know - but intelligence as we understand it is a property which can emerge out of something that works using nothing more than physics. We might not know how to do it, but it's clearly possible.
Could you explain that a bit further - as I understand it, intelligence is something that allows you to acquire knowledge and skills and apply them - a machine can't do that at the moment as none have yet developed the ability to understand the world around them. As yet there are no machines that can take in information and make use of it like a human can.
otolith said:
We don't know - but intelligence as we understand it is a property which can emerge out of something that works using nothing more than physics. We might not know how to do it, but it's clearly possible.
That sounds like a long way off from becoming a practical reality. We can barely simulate a single protein fold in any reasonable amount of computing time. A neuron simulated down to atomic level would need magnitudes more processing power. Then we'll need a few trillion more than that for a whole brain.Edited by scorp on Wednesday 1st February 10:06
scorp said:
That sounds like a long way off from becoming a practical reality. We can barely simulate a single protein fold in any reasonable amount of computing time. A neuron simulated down to atomic level would need magnitudes more processing power. Then we'll need a few trillion more than that for a whole brain.
I don't think brute force simulation is likely to be the answer - I'm merely pointing out that there is nothing magical or spiritual about how intelligence (and what we call consciousness) emerges from biological machines. Gassing Station | Science! | Top of Page | What's New | My Stuff