More Singularity Curmudgeonry from John Horgan
John Horgan goes on the offensive against the Singularity concept on his relatively new blog at SciAm.
My own skepticism is based on simple comparisons of Kurzweil's claims with what is actually happening in science. For example, Kurzweil contends that reverse-engineering the brain isn't that big a deal. "The brain is at least 100 million times simpler than it appears because the design is in the genome," he wrote on the blog Posthumans. "The compressed genome is only about 50 million bytes," which is "a level of complexity we can handle."
I agree with John that this estimate of the difficulty of AI is an oversimplification. It carries the assumption that AI will be a copy of the human brain, which isn't necessarily true. It also ignores the complexity of the process of neurogenesis and continued development. The real brain is much, much more complex than the portion of the genome that codes for it, and it probably won't be until after the Singularity until we understand the details of how the brain is created from the genetic code.
Is it really so far-fetched to believe that we will eventually uncover the principles that make intelligence work and implement them in a machine, just like we have reverse-engineered our own versions of the particularly useful features of natural objects, like horses and spinnerets? News flash: the human brain is a natural object.
I think Kurzweil is wrong and overconfident on a lot of specific points, but I appreciate his overall vision.
June 24th, 2010 - 01:45
Haven’t read any of the material concerned, but how is Kurzweil’s view an oversimplification? Surely the additional ‘matured-brain’ complexity you mention is merely symptomatic of its (genetically designed) reaction methods to being beaten around by the circumstances of existence?
Isn’t he saying, yes, the brain we see functioning today is more complicated – but that complexity is emergent according to the genetic instructions for how the brain should deal with sensory input?
In other words, nail the genetic instructions for neurogenesis and continued development, let them loose on reality in the same way a human’s genetic instructions are let loose, and the very same complexity will eventuate.
I think I just said exactly the same thing 3 times. Don’t know why, please don’t ask.
June 24th, 2010 - 05:52
The problem is that DNA is a physical object acting together with other physical objects. A lot of complexity actually resides in the cell and we do not have a complete handle on this. To get an intelligent being by reading DNA you might actually have to simulate the cell at a molecular level, something obviously silly.
I suspect, however, that Michael is right that one can abstract some degree of function of the top of brains. This is what narrow AI actually does with some success. Narrow AI can feel shallow because it misses a whole bunch of underlying stuff, but that doesn’t change the fact it actually works in some situations. I don’t if the SIAI idea of finding the principles behind intelligence can work and, in fact nobody else does either. It may be that intelligence is just a large bag of fairly simple tricks with no really deep principles behind it.
June 24th, 2010 - 08:27
John Horgan writes:
“But don’t worry! Immortality is right around the corner! Ray Kurzweil says so!
When I debated Kurzweil at the 2008 Singularity Summit, a revival meeting for the faithful…”
John makes the fatal mistake of making personal attacks, losing all credibility in doing so.
Why does John feel the need to be at Kurzweil’s opposite side?
Why can’t he just disagree in a calm manner?
John Horgan doesn’t deserve publicity.
June 24th, 2010 - 09:14
So far, I think Kurzweil is a little out of touch with the human brain interacting with a larger system we live in? I’ve read his book Singularity, I’ve visited the University, and I read the article “Merely Human” by By ASHLEE VANCE Published: June 11, 2010: http://www.nytimes.com/2010/06/13/business/13sing.html?pagewanted=all and then Singularity’s follow-up. I myself wonder about whether there is more than just neurogenesis as well? Doesn’t there still need to be a more systemic cognition issue to look at? Doesn’t the encoding in our ‘matter’ have ties to our ecosystem, wherein Nature of our ecosystem, have various boundaries placed on our own human system controlled by the brain, i.e., all built with a multi array of signaling systems (chemical, magnetic, etc.) for systemic survival (in other words there is a very vast communication system going on throughout us and within us being processed through the brain)? We, humans, have a wide range of built in autonomic and automatic patterns that are hedonically (pain/pleasure) controlled and reflexively driven to reinforce our ability to survive in our planetary transient environment, right? Yet, even here, has the human brain has evolved the capacity to override some of these life sustaining autonomic mechanisms? So, I imagine whatever is built to mimick our brain, etc. would have to be tied well into our ecosystem as well, correct? I’m not an expert, but just wondering? Please correct me if I’m incorrect.
June 24th, 2010 - 10:18
Alas, we don’t know that “the design of the brain is in the genome” – because of the anthropic principle.
The details of why not are in the “The brain and the genome” section of: http://alife.co.uk/essays/how_long_before_superintelligence/
June 24th, 2010 - 11:01
“The problem is that DNA is a physical object acting together with other physical objects. A lot of complexity actually resides in the cell and we do not have a complete handle on this. ”
The complexity of the cell does not seem to carry significant additional design information. Aside from the dna sequence, you might require knowledge of epigenetic information. Spatially distributed RNA is also used in embryonic development, iirc, but that would be about it. We do not need molecular simulations of everything going on in the cell, that is metabolic functions, lipid molecules, proteins, etc. New cells are constructed based practically solely on the information carried in the dna all the time. So while it would be interesting to know all the molecular details, it’s not necessary to understand that this gene codes for a particular protein that does X thing and is transported like this, modulate like that, etc.
Vast simplifications are possible once we know what the rules command, these neurons go here those there, they’re connected this way, they react this way to a certain neurotransmitter, they’re modulated X way, they respond or alter their response in X ways, etc.
” So, I imagine whatever is built to mimick our brain, etc. would have to be tied well into our ecosystem as well, correct? I’m not an expert, but just wondering? Please correct me if I’m incorrect.”
There have been blind deaf fully intelligent individuals able to communicate and showcase the wonders of general intelligence.
I honestly do not think that an individual with high levels of paralysis + blindness, assuming normal human levels of intelligence, would not showcase general intelligence through spoken language. Even if food was provided independent of his actions at precise intervals, he’d likely still be able to learn and communicate. Such an individual would only have mostly verbal communication as a means of interacting with the environment, not that different from being disembodied.
June 24th, 2010 - 11:21
Re “There have been blind deaf fully intelligent individuals able to communicate and showcase the wonders of general intelligence.”
Interesting. I guess there are various concepts on what intelligence is and mimicking the brain?
Complex Adaptive and Complex Evolving?
Some interesting finds (references noted below): In my search for answers I discovered the following: Accordingly to John H. Holland and Murry Gell-Mann and others, complex systems are made up of multiple interconnected elements and adaptive in that they have the capacity to change and learn from experience.
Co-evolution: All systems exist within their own environment and they are also part of that environment. Therefore, as their environment changes they need to change to ensure best fit. But because they are part of their environment, when they change, they change their environment, and as it has changed they need to change again, and so it goes on as a constant process. (Perhaps it should have been Darwin’s “Theory of Co-evolution”. ) [2]
Some people draw a distinction between complex adaptive systems and complex evolving systems. Where the former continuously adapt to the changes around them but do not learn from the process. And where the latter learn and evolve from each change enabling them to influence their environment, better predict likely changes in the future, and prepare for them accordingly.[2]
They still have all the attributes of built in autonomic and automatic patterns that are hedonically (pain/pleasure) controlled and reflexively driven though, you are mentioning only one or two general senses missing? So, isn’t there still more to us, more nerves reacting to signals for complex adaptation to prevail (sensing the environment, etc.)? Again, anything built to mimick the human brain would need this type of complex adaptive system for survival (the urge to strive for more and/or survive). Wouldn’t it?
[1] Rizzolatti, G., & Craighero, L. (2004). The mirror-neuron system. Annual Review in Neuroscience, 27, 169-92.
[2] A brief description of Complex Adaptive Systems and Complexity Theory, by Peter Fryer Also found at: http://www.trojanmice.com/articles/complexadaptivesystems.htm
June 24th, 2010 - 14:20
Does the human brain fit onto a CD-ROM – a rather antiquated, low capacity storage format these days? Probably. But only when you consider it as an emergent effect of interaction with the physical and societal context. So there’s a lot of implicit information out there required to make the whole iguana.
Despite the difficulties I’m fairly optimistic about the prospects of reverse engineering the brain. I think one day there will be a scientific definition for consciousness in terms of particular circuits organised in particular ways. It may take time, and will not be easy, but in principle I can see no major roadblocks.
June 24th, 2010 - 15:39
I do think that simplifications are probably possible. All I was arguing was that the genome is probably not the place to start. After all no genes code directly for neurodevelopment, they code for proteins which interact in complex ways to build a brain. Assuming there are general principles, they can probably be found.
I’m not quite in agreement with the people who say disembodiment is no big deal. After all, children deprived of the ability to experience language don’t learn it. Helen Keller is not an exception as Anne Sullivan managed to expose her to language through much patient and tedious work. Similarly, people who grow up blind and have their eyes fixed have trouble parsing the visual world.
I’ve also had the experience of meeting someone who do to neurological damage had lost the ability to construct a consistent model of her body’s location in the world. While intelligent, she was severely impaired, being unable to tell that she wasn’t in danger of falling onto subway tracks when she was about ten feet from them. I’m not sure a being incapable of learning from the experience of moving about in the world would be intelligent in a human sense.
June 24th, 2010 - 17:34
You said it “deprived of the ability to experience language don’t learn it.”, obviously without exposure to outside information there’s no way it can miraculous obtain it(unless you had infinite computational resources…), I’m not suggesting that. Obviously exposure to a limited environment that can be as simple as verbal communication could very well be enough. I believe that a bedridden blind human with a feeding tube and just one functioning arm, would actually be able to learn and use language. Deafness combined with blindness is a very serious handicap, but provided an individual is able to hear they should more easily be able to master language, likely even if they lacked sight, smell, touch,taste and were bedridden.
With regards to children with restored sight having trouble with it, there are critical development periods after which plasticity is greatly diminished.
And as for 3d environments, virtual worlds can provide 2d and 3d environments, albeit simpler ones.
June 25th, 2010 - 00:33
Why attract attention to what is basically an ad-hominem attack?
June 25th, 2010 - 04:42
Tim Tyler:
“Alas, we don’t know that “the design of the brain is in the genome†– because of the anthropic principle.”
Well said, Tim. Again, Tim Tyler says something of high information content and zero BS content.
Though, I do wonder just how plausible the idea of an anthropically tuned human brain is. It seems more plausible to me that the process of sexual selection and coevolution that produced the brain was anthropically tuned. The number of bits required to tune the humans’ ecosystem to promote intelligence seems a lot smaller than the number of bits required to hard-code it in through some kind of mutational miracle.
One could make this argument rigorous by saying that the neuro-code for intelligence is probably compressible: it is not a random piece of brain wiring, it is brain wiring that satisfies a particularly simple requirement, so in the multiverse, there will be vastly more examples of intelligence that arose as a solution to that requirement than by chance.
June 25th, 2010 - 05:00
Michael:
“The real brain is much, much more complex than the portion of the genome that codes for it”
You don’t know that. It could, in the technical sense of the word “complex” (kolmogorov complexity), be less complex. In fact it probably is, in the sens that a program of < 50 million bytes could be written that reproduces the brain’s input/output behavior to the required degree of accuracy.
June 25th, 2010 - 19:48
Horgan writes: “Believers squabble over how exactly the Singularity will go down. … All the predictions entail superintelligence and immortality.†He also could’ve added nanotechnology.
If this is the case, then why does he treat these ideas like a monolithic and irrational religious cult. If I “squabble†with Kurzweil and estimate that it will take 100 years for biotech to gradually evolve to advanced nanotech (and lead to immortality) and for advanced AI to gradually develop, am I a member of the crazy singularity cult or just a sensible scientific rationalist? If his problem is with Kurzweil’s timing (and I probably agree with Horgan here) then just say so and stop the ad homenin comparisons to religions.
If Horgan believes that fundamental physics forbids a complete understanding and control of human biology, or that it will take more than a century or two, then he is the one being irrational. Catholicism or non-secular Buddhism will be no more true in two centuries than they are now. On the other hand, I can’t imagine anything but an apocalypse stopping science from having a complete understanding and control of biology within two centuries.
June 25th, 2010 - 20:33
The tendency of opponents to fixate on transhumanism as religion annoys me. Too many solid critiques get mislead by the siren’s song of rhetorical impact over intellectual substance.
June 25th, 2010 - 21:01
“Humans are dumber than bags of hair,” declares Entity 296. “Only the most naive scientist would try to develop a technology to understand those smelly lumps of protoplasm,” it states. “The noises they emit from the holes in their heads are ultimately less enlightening than cosmic static.”
June 26th, 2010 - 02:01
A quotation from Glenn Reynolds-
http://www.ideasinactiontv.com/tcs_daily/2006/04/a-rapture-for-the-rest-of-us.html
“I want to focus on a different aspect of Ken MacLeod’s ‘Rapture of the Nerds’ comment, because I actually think it cuts both ways. Yes, it’s possible to draw parallels between the Christian idea of The Rapture — and, even more generally, between religious ideas of transcendence generally — and the notion that, once human technology passes a certain threshold, roughly that described by Vinge and other singularity enthusiasts, human beings will potentially enjoy the kind of powers and pleasures traditionally assigned to gods or beings in heaven: Limitless lifespans, if not immortality, superhuman powers, virtually limitless wealth, fleshly pleasures on demand, etc.
These do sound like the sorts of things that religions have promised their followers throughout human history. That leads some who invoke MacLeod’s comment to contend that because singularity enthusiasts hope for the same kinds of things that religious believers have hoped for, singularity enthusiasts are merely adherents to a new sort of religion, the religion of science.
But as Isaac Asimov has noted, the religion of science is distinguished by one chief characteristic: ‘that it works.’ I express no opinion on whether science will actually deliver on these hopes. But I note that people once looked to supernatural sources for such now-mundane things as cures for baldness or impotence, only to find those desires satisfied, instead, by modern pharmacology. Yet that hardly makes those who place their faith in pharmacology members of a religion — or, if it does, it makes them members of a religion that is distinguishable from those dependent on the supernatural.”
The real meaning of the old fantasies is that they show us what we really want. If science can make it so, it will be so.
June 26th, 2010 - 11:35
Amnon, because Horgan is a well-respected journalist and Scientific American gives him a platform for his articles. Just because someone makes ad hominem attacks doesn’t mean that they lack memetic chutzpah or that they should be ignored. Many people I know are clueless about media and PR because they seem to confuse their personal preferences of argument with those of the majority of the public.
Also, isn’t it obvious that I would want to give publicity to certain critics of H+? In some cases, poor critics can boost an argument more effectively than articulate advocacy. Why do I have to spell this out?
June 26th, 2010 - 12:08
I too have a problem with many “singularitarians”, because I see far too many of them thinking that “Superhuman AI will come and save us poor dumb apes from ourselves.”
Our problems will have to be faced and overcome LONG before we reach superhuman intelligence, artificial or otherwise. And the fact that we spend so much time denying that our own biology and genetically driven instincts underlie the overwhelming majority of those problems is perhaps the BIGGEST hurdle we are going to have to overcome to reach the singularity at all.
So long as we pretend to “not have instincts” we’ll never find solutions to things like poverty, injustice, corruption, massive wealth disparity, and most of the rest of the problems that beset us. AI won’t change that. Superhuman intelligence alone won’t guarantee that we won’t continue to blind ourselves to the inconvenient instincts that we inherited from our savanna days. Friendly AI won’t matter if we program our blindness and biases into it.
So yeah, the more people call Tranhumanism a “Religion” or a “Cult” the more hope I have that those who believe in an “AI Savior” will wake up and stop waiting for salvation, and focus on DOING something to make that better future happen.
And Micheal, for the record that is NOT directed at you, because you are one of those DOERS. It’s directed at far too many others I have met who try and say that man’s problems are “Too Big” and only “Superhuman Intelligence” can hope to solve them.
June 27th, 2010 - 09:20
“Also, isn’t it obvious that I would want to give publicity to certain critics of H+? In some cases, poor critics can boost an argument more effectively than articulate advocacy. Why do I have to spell this out?”
Michael,
You make a good point. A real movement is supposed to have detractors. It makes it more real.
On the other hand, giving publicity to people who make personal attacks, is rewarding bad behavior.
That’s why you have to spell it out.
But in the big picture, it doesn’t even make a difference whether Horgan lets out the occasion brainfart or not.
June 27th, 2010 - 16:24
Transhumanists don’t appear too cash-strapped, apparently a few billionaires amongst them. So what’s holding those super-trancentech developments back?
June 27th, 2010 - 20:23
[blockquote]Transhumanists don’t appear too cash-strapped, apparently a few billionaires amongst them. So what’s holding those super-trancentech developments back?[/blockquote]
Many of us don’t even want them to happen particularly quickly, since that would be unnecessarily dangerous. I hope the biggest developments take a rather long time to happen, since that will allow for maximizing safety and seeing to some other ethical concerns.