Ben Goertzel Scolds Alex Knapp for Calling People Who Advocate Approaches to AGI other than Brain Emulation “Magical Thinkers” Thursday, Jun 23 2011
AI and singularity 2:35 pm
In the comments thread, Ben Goertzel scolds Alex Knapp:
Heh… thanks for this post Alex, it helped me understand your world-view a lot better.
You previously wrote an article accusing one of my H+ Magazine articles of “magical thinking” — and in this article, you use the same phrase to jab at Michael Anissimov…
Calling those of us working on strong AI from approaches other than brain emulation “magical thinkers” is incorrect, obnoxious, and (sorry to be harsh, but…) poor journalism. It’s the kind of opinionated, non-fact-based journalism more appropriate for the politics or arts page than the science page, IMO.
It’s certainly your right to be skeptical of the possibility of creating human-level AI via non-brain-emulative methods. Many serious scientists are. However, there are also some serious scientific arguments as to why non-brain-emulative human-level AI **may** be possible within a few decades of work.
If you are not familiar with these arguments, so be it. You’re just one man, and you write about an impressive variety of topics. But that doesn’t make it right for you to insult the rationality, scientific-ality or intellectual honesty of those of us scientists and engineers pursuing non-brain-emulative AGI.
I might be wrong about the possibility of non-brain-emulative human-level AI, but if so, it’s not because of engaging in “magical thinking.” And nor is Michael Anissimov (who I know fairly well) engaging in thinking of that nature.
I’m disappointed. You seem an intelligent person, and you share a lot of interests with those of us in the Singularitarian world. And you’re a good writer, with the ability to turn out an amazing diversity of sci-tech articles each day. But then you sink to the level of ad hominem attacks against the thought processes of individuals whose views differ from your own! Tsk, tsk, tsk…
Alex, I can’t wait for your next post on brain emulation! Not as good as Dale Carrico’s writing, I must admit, but still entertaining in the same vein.
This is correct you are not a magical thinker, you are however a non-expert and a bit of an ignoramus.
I suggest you leave the topic be since you have nothing to teach us and almost everything to learn. If someday you attempt to make AI great then perhaps you will have some valid opinions. As it stands you are coming down on a side of an argument which you don’t understand and cannot defend.
Tweeted and FBed: What the fuck is wrong with magical thinking? It is what got Columbus to America.
You are right, Carrico is more fun and entertaining.
You guys should watch this video from Scott Aaronson – it really codifies the idea of computability:
http://www.youtube.com/watch?v=8bLXHvH9s1A
Anyway, as other people have pointed out, its all an efficiency problem, not a possibility problem. The brain is a physical system, and thus can be simulated by a Turing Machine. The question is if it can be done efficiently (i.e. quickly).
This is why hardware is a big deal: say we had the right algorithms, etc, but we just had a computer that was 100 times too slow. To almost anyone, this AI would seem like a complete moron, taking 100 years just to gain the intelligence of a 1-year old baby. But this would of course be very close to human level AI, as we’ll make that kind of hardware gains in less than 10 years!
For this reason, I don’t think any AI criticism is valid until we are pretty sure we have the necessary computational power, but still don’t see human level AGI. Until then, we could be close, but seem very far away…
SAI might take 20,000 years to create…..
at the rate of progress of the year 2000
just sayin’
There are many paths to emergent hyper intelligence.
My feeling is that we will achieve it through exploring all of the paths. Literally, the act of exploring every possible path, including the blind alleys, is what will get us there.
There are constant arguments amongst transhumanists over which path to take.
I say, take them all. There’s enough people to explore all options. Of course, these arguments are driving everyone to do just that.
It’s like religion. Infinite discursiveness. Fracture and splinter and form opposing groups. Like a fractal. Fill in every crack to illuminate the form it will take in the future.
I have more faith in IA because I think it will win the race.
AI seems to have this problem of re-inventing the wheel, while IA takes what we already have and makes it better. For that reason, my biggest concern is not Skynet or the Matrix, but Lawnmower Man.
A million augmented humans will become SAI overnight (overweekend?) and the sound of their awakening will be every phone on the planet ringing at the same time.
We are creating these Lawnmower Men right now. Guess who, currently, is most likely to receive BCI implants? People with neurological disorders and mental illnesses.
Assuming that this remains true for at least the next ten years, and that BCI continues to improve exponentially, we can expect those BCI’s to extend further than just mitigating the symptoms of those neurological disorders, and actually begin to enhance intelligence. And we haven’t studied how the brains of these people will also co-evolve in the presence of these BCI.
In any case, my point is that it’s already here. We already have BCI enhanced humans.
So, it’s a race, and my money is on the Lawnmower Men. Of course, AI has the advantage that as soon as it’s turned on, the race is over. SAI could be sitting at the start line, while BCI is a hairs breadth away from the finish line, and the second someone throws the switch, SAI wins just like that, so I’m not claiming the race is over, just that BCI is already in the lead, and has been for awhile.
Bring on the Singularity.
My comments: http://giulioprisco.blogspot.com/2011/06/magical-thinking.html
I think transhumanism is magical thinking in its purest and noblest form, the kind of magical thinking that has taken us from caves to the Moon, and will take us to the stars and beyond.
Absolutely agree with Prisco. Scientists should wake up every day with this idea. We trust existing science way too much. Yes, use the method, yes, trust the measurements, but… don’t trust the Science that says these are the only methods and that the only way you can interpret the measurements is what you already know.
Reinterpretation of what you see with a new, often radically new, conceptual framework (of the kind “there is no… it’s actually…”, “up is down”) often precedes advancements. But it’s really difficult to entertain an unproven hypothesis, and it may take years to accept it yourself, and even more for the scientific establishment until the hard data is in.
Knapp: “Silicon chemistry is more limited than carbon chemistry by its very nature. Electric circuits are more limited than electrochemical circuits by their physical properties.”
essentially says that there’s some (arbitrary, low) limit to digitally representing things. This makes no sense at all. This is beyond dumb. This is not even wrong.
Proponents of the “can’t be done” side are a waste of time. They never got and never get anything done, by definition. Why would you bother discussing with them since they think things can’t be done? It’s not as if they’re going to be thinking about solutions to the problems.
They’re nothing but an example of humanity shooting itself in the foot by telling itself what’s possible and what’s not. Such guys will be proven wrong. Again and again, but the damage they can do by persuading people from thinking about “crazy,” “impossible,” “too complex” solutions can be great. Humanity couldn’t have done what it has if everyone believed all that is in the text books.
Science, particularly in engineering, that says “Don’t do this, don’t bother trying this, this doesn’t work, it’s been tried and there’s nothing there” only means that this particular scientist (or the whole community) was unable to make it work, or failed to get any interesting results, or interpret them in a way that gets you something new.
When science does that, it prevents people from looking closer into things, which may, in the end prove to be exactly what you should have been doing all along!
The lock-in that scientific paradigms suffer from is a major hindrance to advancement because getting funded is hard if you don’t work within the paradigm.
The current paradigm is always the most solid-seeming one: everything is backed up by hard data and equations …all built upon a misinterpretation, a fundamental flaw in the premises of the conceptual framework.
When Moore’s law fails and it’s no longer mooooore every year, only the rich can afford faster computing. This will make uploading and all sorts of interesting things a rich-only option. The rich of course will be pleased to have another toy to show off. But immortality through uploading would become a really good incentive to get rich. Now, it’s mostly just pleasures which are overrated and become boring soon (if you’re intelligent enough – some people never tire of “the lifestyle”, and they’re not the brightest bulbs).
By the way nobody rich except perhaps Stephen Wolfram is running their own supercomputers…. I wonder why? Perhaps there’s just nothing interesting to compute today.
A paradigm is a paradigm only because everything else that has been tried has failed or hasn’t been yet attempted at all.
I see most of the arguments against transhumanist technologies as simply repeating that the paradigms of today are the last paradigms there ever will be, that the future will have to deal with today’s limitations, and will have to work within its established framework.
But paradigms are brittle. The change has been accelerating. The future IS accelerating. Instead of thousands or hundreds of years, they last only a few decades or even years.
You can be sure that some of the current paradigms will look laughable, obviously untrue in the not-so-distant future.
When something else works, the paradigm is dead.
All we must do is something else that works. To me, that is the essence of all progress, and transhumanism is its most highly concentrated expression; that is why I’m a transhumanist.
Our propensity to discuss and argue about what’s possible and what isn’t is counterproductive. All that matters in the end is turning dumb matter into smart matter. Get to work and ignore the yapping or dreamless pessimists, the self-appointed only true “realists”.
I’ve done my share of improving the world and I can tell you I’ve heard 10-100 times more pessimism from people than genuine optimism (they may say they support you but if you ask for funding …not so much). This is a disappointing and boring world for those who can’t do anything with their minds, and they keep on telling everyone else no one else can’t either.
Really? Because, you know, this is basic computer science. Digital representation is limited by its fundamental physical properties, and the design of transistors limits the way processors work in ways that neurons don’t.
I’m all for breaking paradigms, but until you provide evidence that current theories of physics and chemistry are wrong, you’re not doing science.
I don’t quite understand your reply Alex. The Church-Turing Thesis says that any physical process can be implemented on a Turing Machine…
http://www.youtube.com/watch?v=8bLXHvH9s1A