Here is the Robin Hanson interview by Sander Olson. Dr. Hanson is a Professor of Economics at George Mason University. Before becoming an economist Dr. Hanson worked on AI at NASA. Dr. Hanson has done pioneering work on prediction markets. Dr. Hanson also believes that whole brain emulation may be achieved within the next several decades, and that this could lead to true AI and soaring economic growth rates.
Robin Hanson blogs at Overcoming Bias.
Question: You started out your career working on AI. Why did you switch fields?
Answer: Back in the 1980s, I worked for the aerospace industry on AI. But the insights that I was getting in AI were running low, and I became increasingly interested in institution design. So I switched to that field, and I am now a professor of economics at George Mason.
Question: What is your current assessment of the AGI field?
Answer: It is making slow but real progress. Many of the supposedly new concepts in the AGI field have actually been around for decades, but now have new hardware. But mature AGI is a long way off. At current growth rates, I don’t see hand-coded AGI arriving for another century or more. I believe that we may see machine intelligence emerge in the next few decades, but it won’t come from hand-coded AI.
Question: So you think it will come from reverse-engineering the brain?
Answer: This isn’t reverse-engineering the brain per se, but rather whole-brain simulation. I think that whole-brain emulation is the most likely scenario for creating true artificial intelligence. It is a difficult problem, but we know roughly how difficult it is, and one can discern trendlines for the enabling technologies. The beauty of whole-brain simulation is that you don’t need to understand how the process works in great detail.
Question: What technologies would be required for high granularity emulation?
Answer: The three prerequisite technologies are sufficiently fast computers, fast and inexpensive scanning, and models for each of the brain regions. The next couple of decades may well be sufficient time for the necessary computer and sensors, so if enough progress is made on modeling then we could have it within the next thirty years or so.
Question: You have written about how a major AI advance could bring about a step change in economic growth. Could another breakthrough, such as fusion power or molecular manufacturing, lead to similar growth?
Answer: We actually spend less than 10% of the economy on energy, so it is difficult to see how an energy advance by itself could bring about exponential changes in economic growth. Molecular manufacturing is more plausible, but it is still a long shot. We only spend about 15% of the economy on manufacturing, though advanced nanotech could facilitate a cascade of changes. So artificial intelligence is the most plausible way towards much faster economic growth rates in a short timeframe.
Question: To what extent is technological change mirrored in economic growth rates?
Answer: An analysis of economic growth rates shows steady growth rates throughout human history, punctuated by a few sudden step changes. The two step changes that can be most clearly discerned happened when humans switched to settled agriculture, and during the industrial revolution. Despite numerous individual technologies being introduced, the economic growth rate has for most of human history remained at a steady pace and hasn’t wavered much.
Question: But you believe that a third step change could occur in the next few decades with artificial general intelligence?
Answer: AGI based on models of the human brain may well be constructed within the next three decades or so. If so, it could usher in far faster growth. In such a scenario the world economy could double within a matter of months.
Question: You have written about the potential of prediction markets. Do these markets work by leveraging the wisdom of crowds?
Answer: Actually they are designed to leverage the wisdom of whoever actually know best, be that experts, amateurs, or whomever. So for instance, if I wanted to know how much longer Moore’s law would last, I could subsidize a prediction market, and that subsidy would go on average to folks who make the market estimate more accurate. The more money that I was willing to offer, the more accurate the answer should get.
Question: So these prediction markets would be able to tell you near term events with a high degree of accuracy?
Answer: No, but they could be nearly as accurate as possible, and the could tell you their accuracy. So I could create the prediction markets for the question “How long will Moore’s law last? and get a full probability distribution about when that future date will occur.
Question: If you had a million dollars to establish a futures market, what question would you want answered?
Answer: I would probably establish a prediction market for betting on when whole brain emulation and other key future developments were going to occur. I would be very interested to see the statistical distribution of bets for that question.
Question: What if your goal was to create a prediction market that would demonstrate the potential and accuracy of prediction markets?
Answer: In that case I would create markets on firing the CEOs of every fortune 500 company. I would create 1,000 markets, two for each company. One market would estimate the corporations stock price if the CEO is fired in the next quarter, and the other one would estimate the stock price if the CEO is not fired. Then one could compare the two prices.
Question: So if the prediction market indicated that the stock price for company x would go down if the CEO were retained but would rise if the CEO were replaced, that would provide valuable information for the board.
Answer: Yes. The board could disregard the information, but it would be fairly easy to track the records of which companies follow the advice and which ones don’t. Within a few years it would hopefully become apparent that following the advice of the prediction markets was a wise strategy.
Question: What is your assessment of prizes, such as the x-prize, to stimulate research in specific areas?
Answer: Prizes can be a highly effective mechanism for inducing specific achievements. However, many people who allocate money in the name of research aren’t primarily concerned with achieving identifiable goals. These individuals care about other things, which they sometimes get less of if they fund the prizes.
Question: You are an Alcor member and have enrolled to have your head frozen when you die. How do you anticipate achieving consciousness in such a scheme?
Answer: The plan is to freeze my brain and cryogenically preserve it until science has succeeded in creating whole brain emulation. They could at that point download my brain patterns into a computer. This computer would contain all of my memories, my personality and beliefs, and my consciousness.
Question: So you foresee a time in the future when downloading brains into computers will be common?
Answer: Yes, but most such downloads will be from a relatively small number of humans. The “best and the brightest” will have myriad copies of their minds downloaded to computers, since these minds will be the most valuable. So these few minds will be massively reproduced, and will form the bulk of our descendants.
If you liked this article, please give it a quick review on Reddit, or StumbleUpon. Thanks