Ken Goldberg: Creative robots, Kurzweil fallacy, and what it means to be human
nautil.us
nautil.us
You know this thing called reading? It's about ten times faster than video for sharing verbal content.
How is that an argument against Kurzweil? So far, we have no real ideas on consciousness. So it seems that until we've at least figured out that, no should should be making flat prohibitions on what we can achieve. (Kurzweil could be wildly off on timelines but that's the most obvious and only real criticism there is right?)
Not to mention, as far as we can tell, we are machines. So it takes an extra something to say we can't ever build ourselves. (Let alone figure ourselves out and improve on the design.)
It's the fact that all of science deals with quantitative expressions and measurements, whereas consciousness is qualitative. We don't know what it like to be a bat, me as a man I don't know what is it like to be a woman, nor does my wife know what it is like to be a man, or someone else for that matter. Only for the simple reason that qualitative experiences, in themselves, are subjective and bound to the subject experiencing them.
How do we know the electrons don't feel? We don't. Some, like me, presume they don't. Others, in panpsychism, presume they do. At little tiny bit, at least.
>> Not to mention, as far as we can tell, we are machines. That is a very strong statement.
As far as Kurzweil goes as a prophet: how is his track record? Pretty bad as far as I can tell.
Actually, it's pretty good, if you consider his predictions from 15 years or so ago and take some liberty in interpretation. I forgot where I read the list that enumerated it, though.
>So far, we have no real ideas on consciousness.
Oh really? How long a literature search did you do? Which authors have you read on the subject?
Or do you just mean, "We don't have clear-enough ideas about consciousness that we teach them to undergraduate students and give TED Talks about them, because we're still actively researching the matter"?
So, sure, there's plenty being said and perhaps some not-totally-worthless ideas. But nothing even remotely close to explaining consciousness. Sure, perhaps it's really, really, hard, but lets not pretend we've made any useful progress.
The key word there is apparently. No, you are not made of magic.
>So, sure, there's plenty being said and perhaps some not-totally-worthless ideas. But nothing even remotely close to explaining consciousness. Sure, perhaps it's really, really, hard, but lets not pretend we've made any useful progress.
Or in other words, you haven't actually looked in any neuroscience or cognitive science journals.
Uh, I don't doubt that this kind of sentiment arises when one has worked with robots for a while and done some impressive stuff but had them consistently fail to inspire a feeling of interestingness in one.
But the statement itself is, ironically, neither new nor interesting since beyond interestingness being something people sense, the statement gives one not a clue what the criteria is for determining its presence. And a remarkable number of critiques of AI by very insightful people come down to a kind of vitalism - come down to a long, round-about way of saying "robots just don't that special something".
And, of course, humans so far have failed in producing a wide range of human-like behaviors in robots and programs. There is something of a psychological tendency to take everything a computer can do and deprecate relative to everything a computer can't do - playing chess was once something people imagined would involve "intelligence" but now no longer do, etc.
All that said, I'd like to shout-out for Douglas Hofstadters' Fluid Analogies [1] approach, which actually makes an effort to describe and define some aspects of creativity.
[1] https://en.wikipedia.org/wiki/Fluid_Concepts_and_Creative_An...
The vast majority of humans aren't particularly creative either. People who can do (e.g.) music a bit are maybe 25% of the population.
Of those, most are imitators with enough technique and expression to successfully reproduce familiar tropes, but not to be inventive.
Innovators are maybe 0.1% of the population - and that number is a complete guess, because so far as I know there's been absolutely no research into creativity in populations.
So "that special something" is rare already. The problem with computational creativity is that success is partly social and psychological - new tropes have to capture the imagination in some way, and it's very difficult for a machine with no model of human emotional and psychological responses to calibrate its output to meet that goal.
Everything after that was one attempt after another to invent new visual languages for painting.
The cultural value of those attempts is completely different to the resale auction value of a Picasso, which is driven entirely by market speculation and the fact that highly rated artworks are treated as a convenient store of value - rather like gold, but more compact and lighter.
Cultural value is probably best measured by influence. On that basis Picasso scores very high, as one of the creators of abstracted figurative art. That trope remains mysterious to people who believe painting should be about photo-accurate or easy-to-read figurative work, but it's been a huge driver of fine art and all kinds of commercial illustration.
In my opinion KG's view here is closer to the truth of it.
It seems to me that the human organism is the singularity, and like many of our technologies, including metallurgy, cities, nuclear weapons, and bioengineering, computer tech and automation has the ability to wipe us out.
But it won't be because it becomes conscious and turns on us; it will be because we cause some chain reaction or blunder, by accident or on purpose: like wiping out an invaluable resource like air or water, or by otherwise destabilizing the global biosphere, or through some code typo that causes automated manufacture machinery to eat us for breakfast.
But the fear of machines gaining consciousness is, at least in our age, pure fantasy. If consciousness were a directly related to complexity, then the moon and sun would be awake and singing -- both are stuffed to the gills with highly organized information! Computers that can provide complex answers, or walk and talk as if they were conscious are no more awake than the moon or sun, and to say otherwise is to be a carnival salesman.
Kurzweil and his ilk are seeing the human consciousness in the mirror of the machine and it scares them, as it should. But it's a mistake to think that the machine will awaken accidentally or casually, as result of simply adding too many circuits, or feeding the wrong code into the wrong compiler. That idea applied to computers is a philosophical misunderstanding of the difference between intelligence and consciousness.
A conscious machine is not something that can be made without extraordinary technical ability and intention-ality - it's on par with the ability to colonize planets - way beyond us, and not something likely to happen by accident, except of course in the traditional sense of already having the ability to do so :}