Keith Norbury on Ray Kurzweil Response Tuesday, Jan 19 2010
Here’s a comment from Keith Norbury on the Kurzweil response post that I agree with:
It looks as though Kurzweil and Anissimov are both quibbling. I had similar thoughts as Anissimov did when I scrolled through the predictions in The Age of Spiritual Machines. But I also thought, well, Kurzweil is just a little hasty in his enthusiasm. Yes, there’s a danger in setting firm dates for predictions of technological progress. However, because he makes them, Kurzweil gets people’s attention. Even when he is wrong on the exact date, he is still able to point to a trend that indicates he will be right soon enough (in most cases). So far, though, the dates have passed for the easier predictions. It gets harder going ahead.
Kurzweil’s main point is that technology is improving exponentially not linearly. That’s a difficult point to grasp. However, we still don’t know if even exponential growth is enough to tackle some sticky problems, such as simulating human intelligence. Nobody knows where the goal posts are yet. Nor do we understand yet the principles involved in uploading a human mind to computer, never mind the engineering it would require. The answers might be just around the corner, or they might be a long way away. Time travel, for example, is possible under the laws of physics. However, the huge energies required pose a giant obstacle to making it a reality.
I’m now reading Kim Stanley Robinson’s excellent Red Mars, which points out the difficulties in making predictions. It’s speculative fiction but also hard science fiction. The trouble is, though, that the hard science in Red Mars is the science of 1993 when it was written. In the book, the voyage to Mars took nine months, as predicted using the technology that was proven in 1993. Since then, an ion propulsion system is well along the road to development that promises to shorten the trip to about 40 days — when it happens. That certainly doesn’t look like it will be by 2026, as in Red Mars. One could argue that Robinson wasn’t being a futurist when he wrote Red Mars. However, at the time he was striving to imagine as accurately as he could, based on the knowledge available, what that future mission would look like. Unfortunately, he didn’t imagine that humans would develop a better technology for getting to Mars, even though the principles of ion propulsion were already well known back in the 1990s.
My guess is that Robinson, in writing Red Mars, was thinking too linearly about technological progress and not in the exponential way that Kurzweil does. That’s what sets Kurzweil apart from other intelligent people who speculate about the future.
I agree with Kurzweil that many important technological metrics are improving exponentially, and that his linear-thinking critics are incorrect. I have always argued that major change is likely in the relatively near future. I regard a Singularity at 2029 or earlier as definitely within the realm of possibility. I am a “Singularitarian” of the type that Kurzweil describes in his book. Much of my life is focused around the idea of a Singularity, similar but not the same as Kurzweil’s idea of the Singularity. I object to Kurzweil’s statements that MNT and nanorobots will certainly be a reality in the 2020s. I object to a lot of other things. I agree on the broad outlines of exponential change. I do not think Kurzweil is an “idiot”, as Singularity Hub misleadingly claimed recently. I think Kurzweil is a genius and I applaud him for making predictions at all.
It is much easier to criticize than to make predictions, I admit that. I believe that Kurzweil’s model is a good framework, and my model of the future is extremely similar to his relative to the mainstream. Still, the fine points are worth arguing. My main focus is on the points themselves. Perhaps I should have just listed the items and not even called them Kurzweil’s predictions, so I could criticize them at will without in any way threatening his reputation. In any case, I don’t think that Kurzweil’s reputation is at risk here. As he pointed out, I just poked at 7 out of 108 of his predictions in the book. I apologize for the sensationalist title of my original post — I didn’t mean that ALL of Kurzweil’s predictions for 2009 had failed, just “Here’s a few failed predictions that I found on this specific web page and I agree with”.
I’m sure that everyone is interested in seeing Kurzweil’s point-by-point analysis of his predictions in The Age of Spiritual Machines. Considering the concerns raised by those seven predictions I mentioned, I think a thorough review of the book is in order, and I’m pleased that Kurzweil himself has taken up the task. I gave the original post a provocative title because I strongly believed that investigation would benefit the entire futurist community, and I hoped to start a conservation on it. In that respect, it appears to have succeeded.
Michael:
You and Kurzweil are both very bright guys. I wouldn’t worry about the tone of your post. You are extremely cordial and fair in your blog and we all get enthusiastic in the heat of argument.
It is possible that both the Kurzweil and Anisimov-Yudkowsky schools of thought are separately correct. Progress continues to accelerate as Kurzweil claims and, when genuine AI develops, it may super-accelerate as you claim. (What could we call that? Calc 101 says acceleration is the 1st derivative of velocity and “jerk” is the 2nd derivative of velocity (1st derivative of acceleration); somehow the “jerk era” doesn’t sound right)
Even under a slow ascent scenario on the order years, the AI(s) would likely lead to super-acceleration like change.
Interesting times are ahead.
“However, we still don’t know if even exponential growth is enough to tackle some sticky problems, such as simulating human intelligence. Nobody knows where the goal posts are yet. Nor do we understand yet the principles involved in uploading a human mind to computer, never mind the engineering it would require.”
Actually, we largely do. We don’t know everything, of course, but we have a reasonably clear idea of what milestones we need to reach and how we can reach them. See http://www.philosophy.ox.ac.uk/__data/assets/pdf_file/0019/3853/brain-emulation-roadmap-report.pdf.
“Since then, an ion propulsion system is well along the road to development that promises to shorten the trip to about 40 days — when it happens.”
Although this is certainly possible, there is little point to doing so; it requires expending a HUGE amount of energy and developing a ton of new technology for relatively little gain. When we have bases on Mars, there will be a reason for *some* spacecraft to travel this fast, but certainly not all. People still use trains and buses even though we have jets and helicopters.
I was a bit outraged when I read your first critique for reasons I took to be obvious. I was frankly surprised when it was answered. I sympathized with Kurzweil’s comeback almost line for line. I think the tone here is more appropriate, and I agree with this almost line for line.
Something happens to us when we try to get people’s attention. Sadly the things that happen usually aren’t great. The same could be said of Kurzweil. There is so much you guys are saying that is profound and historic, I wish you’d just be a bit more patient about the attention.
Personally, I don’t need to be titillated by the hype. The content is almost so disturbing as to cause one to lose sleep over it.
The Uber-Prodigy challenges the Master. Classic Hero’s Journey archtypes!
What must be kept in mind is that Kurzweil has a well-established history of making ideas and theories work in the real world. He has a much better idea than any theoretician of what will and will not work, and he has a proven history of manufacturing beneficial products from those ideas and theories.
Ray Kurzweil has thus far handled this little conflict with grace and panache.
Had he handled it any other way, the social and political damage to any and all futurist theories and agendas, even to product placement of new technologies, may have been incalculable.
“I regard a Singularity at 2029 or earlier as definitely within the realm of possibility.”
Kurzweil predict the singularity for 2045.
2029 is when according to his prediction AI pass the Turing test.
“What must be kept in mind is that Kurzweil has a well-established history of making ideas and theories work in the real world. He has a much better idea than any theoretician of what will and will not work, and he has a proven history of manufacturing beneficial products from those ideas and theories.”
That’s true. However we have also to keep in mind two things. First: his most bold predicitons have nothing to do with work he is actually doing. So it is easy for him to be correct in predicting small advances in software or hardware-it comes naturall to him as somebody working in IT field. And nobody is arguing with him wheather translators are possible or not. So those predicitons are safe predicitons and he is rightfull to do them. No big glory if he is correct, though (of course he thinks differently, just look how he stresses the fact how computers looked like and how they are looking now-he is proud of accomplishments. So simply speaking-he is biased). Second: as for linear vs exponential growth I think that almost everyone agrees, that Kurzweil’s view is superrior. But he constantly fails at understanding that it is not exponential growth (derivative of mass society and capitalist economy I would argue, and not some law of nature which forces life to grow in complexity) per se that matters, but rather how we use its fruits and how important those fruits for the system in general are. And this is still linear and will probably remain linear (because, for instance, we are thinking in such way). That is why Iphone is not important, though it fits singularity chart together with CPU speeds and so on. It is misguiding to present logical small steps in development in such a way that they grow into big predictions. And this Kurzweil does all the time. He’s got 70% predictions ocrrect? OK, but they weren’t big. They seemed big to him, because they were challenges for him and his collegues. But their impact on society and world is insignificant (tell to starving guy in India that he is better off because we got 4 cores in processors, just tell him). Computers in clothes, 20 petaflop computer-it’s really not that important. Important is however how those things will be utilized, next to how common they will be. Up to this point Kurzweil predictions were of easy kind. And honestly – he was only one of many who made correct guesses (but only he wrote a book). Really important are “hard” Ray’s predicitons, such as AI or mind uploading. And those he gots mostly wrong in my (and otherts too) opinion. It’s not that he is complately wrong, he just understand those terms in incorrect way. Sorry for all of you believing in functionalism, but this method is going to give you software mimicing intelligence and nothing more. Passing Turing’s test is just a matter of writing a good software. Hell, there are programs that can fool 30% of people now. With correct algorithms you can make software that will produce outputs of an inteligent and counscious being. As far as results are concerned it does not matter that much, but it changes the way how singularity will look like in most dramatic way-no uploading a’la Charles Stross and no Ramona for Kurzweil (he can make her still, though. But she won’t be a “true” AI, that’s all). We still will be richer, better off and endangered by our intelligent systems-just like Anisimov and others are predicting. But those systems will be mindless geniuses-smarter and better, yet dumb as brick. What does it mean? The same as always-world is much weirder place than we thought, that’s all.
But I admit that I’m not a psychic, so tommorow I can read about some discovery that will make me look as a classical example of an idiot (prooving for instance that functionalism is correct and algorithms can actually think, not just look like they are thinking so they can cheat our hunter-gatherers minds). However there is one thing that I’m absolutely sure about: the world is always much more complex than in our simplified models and no single mind was ever able to understand it fully. The same goes to Kurzweil. And that is why the more we go into XXI century, the more mistaken he will be by definition. If not, I will start to suspect that we are already in some kind of machine and Kurzwiel is its administrator, playing some sadistic game.
Just one final remark: Robinson did not used ion drive, because it was not in development at that time, it was something like orion drive-theoretically possible, but unrealistic beacuase of non-scientific factors. That is why he, being responsible writer, decided not to be overoptimistic and didn’t include ion drive in the story. Besides, we are at minimum 10 years from buidling ion drive which could take us to Mars in just 40 days (so realistically speaking much more, because “10 years” goal assumes that we will become space-hyped just like during the race to the Moon).
I’d like to mention the fact that Kurzweil himself recently shifted
some predictions from 2009 to 2020: devices inegrated in clothing, images projected on retina, and self driving vehicles.
This should be taken as a quiet admission that these predictions were too optimistic , as far as i am concerned.
Everyone can make predictions! Keith Norbury, you are wrong, Kurzweil is not a genius, he is an idiot! The claim that we could reverse engineer the human brain in 10 years is far stretched! Kurzy doesn’t do his homework at all! i bet he will miss a lot of his predictions, don’t look at him like a genius if he got lucky and got a few right!