He starts by laying out some key motivations or guidelines to thinking about this sort of thing, and he paraphrases David Brin (and this is actually paraphrasing Smart):
We need a pragmatic optimism, a can-do attitude, a balance between innovation and preservation, honest dialogue on persistent problems, ... tolerance of the imperfect solutions we have today, and the ability to avoid both doomsaying and a paralyzing adherence to the status quo. ... Great input leads to great output.So how do new systems supplant the old? They do useful things with less matter, less energy, and less space. They do this until they reach some sort of limit along those axes (a limitation of matter, energy, or space). It turns out that evolutionary processes are great at this sort of thing.
Smart goes on to list three laws of information and communication technology:
- Technology learns faster than you do (on the order of 10 million times faster). At some point, Smart speculates that there will be some sort of persistent Avatar (neural-net prosthesis) that will essentially mimic and predict your actions, and that the "thinking" it will do (pattern recognitions, etc...) will be millions of times faster than what our brain does. He goes on to wonder what we will look like to such an Avatar, and speculates that we'll be sort of like pets, or better yet, plants. We're rooted in matter, energy, and space/time and are limited by those axes, but our Avatars will have a large advantage, just as we have a large advantage over plants in that respect. But we're built on top of plants, just as our Avatars will be built on top of us. This opens up a whole new can of worms regarding exactly what these Avatars are, what is actually possible, and how they will be perceived. Is it possible for the next step in evolution to occur in man-made (or machine-made) objects? (This section is around 16:30 in the audio)
- Human beings are catalysts rather than controllers. We decide which things to accelerate and which to slow down, and this is tremendously important. There are certain changes that are evolutionarily inevitable, but the path we take to reach those ends is not set and can be manipulated. (This section is around 17:50 in the audio)
- Interface is extremely important and the goal should be a natural high-level interface. His example is calculators. First generation calculators simply automate human processes and take away your math skills. Second generation calculators like Mathematica allow you to get a much better look at the way math works, but the interface "sucks." Third generation calculators will have a sort of "deep, fluid, natural interface" that allows a kid to have the understanding of a grad student today. (This section is around 20:00 in the audio)
This about halfway through the speech, and he goes on to list many examples and he explores some more interesting concepts. Here are some bits I found interesting.
- He talks about transportation and energy, and he argues that even though, on a high level we haven't advanced much (still using oil, natural gas - fossil fuels), there has actually been a massive amount of change, but that the change is mostly hidden in abstracted accelerating efficiencies. He mentions that we will probably have zero-emission fossil fuel vehicles 30-40 years from now (which I find hard to believe) and that rather than focusing on hydrogen or solar, we should be trying to squeeze more and more efficiency out of existing systems (i.e. abstracted efficiencies). He also mentions population growth as a variable in the energy debate, something that is rarely done, but if he is correct that population will peak around 2050 (and that population density is increasing in cities), then that changes all projections about energy usage as well. (This section is around 31:50-35 in the audio) He talks about hybrid technologies and also autonomous highways as being integral in accelerating efficiencies of energy use (This section is around 37-38 in the audio) I found this part of the talk fascinating because energy debates are often very myopic and don't consider things outside the box like population growth and density, autonomous solutions, phase shifts of the problem, &c. I'm reminded of this Michael Crichton speech where he says:
Let's think back to people in 1900 in, say, New York. If they worried about people in 2000, what would they worry about? Probably: Where would people get enough horses? And what would they do about all the horseshit? Horse pollution was bad in 1900, think how much worse it would be a century later, with so many more people riding horses?None of which is to say that we shouldn't be pursuing alternative energy technology or that it can't supplant fossil fuels, just that things seem to be trending towards making fossil fuels more efficient. I see hybrid technology becoming the major enabler in this arena, possibly followed by the autonomous highway (that controls cars and can perhaps give an extra electric boost via magnetism). All of which is to say that the future is a strange thing, and these systems are enormously complex and are sometimes driven by seemingly unrelated events.
- He mentions an experiment in genetic algorithms used for process automation. Such evolutionary algorithms are often used in circuit design and routing processes to find the most efficient configuration. He mentions one case where someone made a mistake in at the quantum level of a system, and when they used the genetic algorithm to design the circuit, they found that the imperfection was actually exploited to create a better circuit. These sorts of evolutionary systems are robust because failure actually drives the system. It's amazing. (This section is around 47-48 in the audio)
- He then goes on to speculate as to what new technologies he thinks will represent disruptive change. The first major advance he mentions is the development of a workable LUI - a language-based user interface that utilizes a natural language that is easily understandable by both the average user and the computer (i.e. a language that doesn't require years of study to figure out, a la current programming languages). He thinks this will grow out of current search technologies (perhaps in a scenario similar to EPIC). One thing he mentions is that the internet right now doesn't give an accurate represtenation of the wide range of interests and knowledge that people have, but that this is steadily getting better over time. As more and more individuals, with more and more knowledge, begin interacting on the internet, they begin to become a sort of universal information resource. (This section is around 50-53 in the audio)
- The other major thing he speculates about is the development of personality capture and parallel computing, which sort of integrates with the LUI. This is essentially the Avatar I mentioned earlier which mimics and predicts your actions.