Questions by Sander Olson. Answers by Nick Bostrom.
Question 1: Tell us about yourself. What is your background, and how long have you been interested in AI?
My background is in physics, computational neuroscience, AI, and philosophy, and my interests cover many other fields as well. That makes it sound as if I'm interested in a whole bunch of different things, which is true, but there is a common denominator: I want to understand where humanity is going and what possibilities there are for steering the development in beneficent directions. That is my specialization. Obviously one can't do this without transgressing disciplinary borders.
It is clear to anybody who has thought about it that technologies will be developed in this century, probably within a few decades, that will change many fundamental parameters of the human condition. We may no longer have to grow old and decrepit. We may be able to dramatically enhance human intellectual, physical, and emotional capacities. We may be able to build whole new experiential worlds in virtual reality. We may build machines that are as intellectually superior to us as we are to chimpanzees. If any of these things has even a small chance of happening then it is important that we begin to consider the consequences.
Question 2: Many scientists and writers argue that we will need a fundamental breakthrough in software before truly intelligent machines arrive. Do you agree?
Human brains are intelligent and they use some computational architecture and some learning algorithms. These will eventually be uncovered by advances in computational neuroscience. Then we can implement the same methods on a computer to create intelligent machines. It is possible that a less biology-inspired approach will get there quicker, but reverse-engineering the brain provides a fallback option. Once something roughly human-equivalent is built, the step to superintelligence will likely follow shortly.
A "critical mass" of intelligence may be reached when the system becomes smart enough to be able to effectively enhance its own software or hardware. At this point, there might be a runaway process of intelligence-enhancement, so that virtually overnight we go from something slightly above human level to something radically superior. This is known as the "singularity hypothesis" (and it's just a hypothesis). Since a superintelligence would dramatically accelerate all forms of scientific and technological progress, the world would soon become a very different place.
Question 3: Many technological advances are dependent upon the continuation of Moore's law. How long do you think Moore's law will last? How long do you think we will see exponential increases in computing power?
It appears likely that computing power will continue to grow roughly exponentially for at least as long as is required to get supercomputers that have the same processing power as a human brain. Plausible estimates of when this level of hardware will be developed range from about 2004 to 2025, assuming Moore's law survives.
Question 4: What is your opinion of molecular nanotechnology? Do you believe in the concept of Drexlerian assemblers?
Molecular machine systems, based on the molecular construction devices ("assemblers"), have huge potential for a wide range of fields, including medicine, the environment, manufacturing, and space technology. By all calculations, such structures are physically possible and will eventually be technologically feasible. My view is that the more mature form of nanotechnology that Drexler describes (which would probably make it possible to repair and revive those legally diseased persons who have been preserved through cryonic suspension along with other amazing feats) is still likely to be several decades away. Unless there is a breakthrough in machine intelligence, progress in nanotechnology, even after an assembler has been constructed, will be likely be incremental because of the difficulty in designing good nanomachines. Although things unfold very rapidly by our current standards, I think nanotechnology on its own would not lead to a singularity.
There is a great deal of legitimate concern about the weapons potential of nanotechnology. If assemblers become widely available, somebody would build and release a destructive nanobot that could reproduce in the natural environment - it could then eat up or otherwise kill the biosphere. Nanotechnology might also make it possible to develop defenses against such attacks (a global nanotech "immune system") but that looks considerably more difficult than developing offensive capabilities. So there could be a period of great vulnerability. It is not too early to begin to think about how we can minimize the risks.
It is preferable that superintelligence come before advanced nanotechnology, because the former could help reduce the risks of the latter but not vice versa.
Question 5: Many investors are quite excited about the prospects of emerging technologies, such as AI and nanotechnology. What type of projects and companies would you recommend investing in?
I would recommend being a bit skeptical about the hype that is likely to dominate these fields in the early stages.
Question 6: What is your response to the objections raised by Searle and Penrose regarding the impossibility of genuine intelligence arising from pure computing machines?
I think they are wrong, as do most philosophers who have analyzed their arguments.
Question 7: What are your predictions for the next 10 years of computing?
We'll see continuing gradual progress of the same sort that we've seen in the past 10 years. Virtual reality will become really big at some point in the future, maybe towards the end of this decade.
Question 8: Do you agree with the predictions of Ray Kurzweil (Author of The Age of Spiritual Machines) and Hans Moravec (Author of Mind Children and Robot) that the rise of intelligent machines is virtually inevitable? Do you agree that the rate of technological progress is increasing?
The best simple measure of the overall rate of technological progress is the economic growth rate. It seems to have been increasing slightly over the past ten years. Of course, at any given time, there are always particular technologies that grow rapidly. But you'd get a very misleading estimate of the overall pace if you just look at the hottest fields, such as IT and biotech.
Question 9: Will breakthroughs in AI and nanotechnology come more from private industry or from Government backed University research?
Both, with universities focusing more on basic research and private companies more on near-term applications.
Question 10: What are your plans for the future?
For me, I think the best way to contribute is by dedicating myself completely to developing ideas in "transhumanism" - the study of how anticipated technologies can be used to overcome some fundamental human limitations and of the various technic, ethical, strategic, and policy issues related to that. I'm currently writing several papers on such topics and I will be teaching a course and leading a research group with that as a focus next semester here at Yale. I also have plans for a book on transhumanism. And there will be work done on developing the methodological tools that are needed to address these issues. I have a book coming out with Routledge in April 2002, titled Anthropic Bias: Observation Selection Effects in Science and Philosophy. It presents the first mathematically explicit theory of observation selection effects, using a framework of Bayesian probability theory. It has applications in range of scientific fields, including cosmology, evolutionary biology, traffic analysis, quantum physics, and game theoretic problems involving imperfect recall. It also has implications for evaluating hypotheses about humankind's future, which I would like to explore further.
I'm also trying to do my little part in encouraging others to begin thinking about these problems and opportunities, so that we can face them together. Everybody can contribute something - ideas, funding, time, weblinks, scholarship, enthusiasm, influence, organizational ability, or simply by becoming part of the informed and caring segment of the public.
|
|||||||||||||||||
