Written in 1996, slightly edited since then.
It’s hard to grasp how different the far future will be from what we’re familiar with. To start, here is a list of flaws in the human design. These are weaknesses you have to live with if you’re an unimproved human being.
· Humans are slow. Human reaction time to simple events is at least a quarter second, not fast enough to control an aerodynamically unstable aircraft. Even open-loop actions (which are planned in advance and don’t include reaction time) are not fast enough to pump a car’s brakes precisely. A neuron in your brain works in milliseconds; a fast computer chip works in nanoseconds, a million times faster. A skilled human can just barely play chess at one second per move; a chess program will defeat a human every game at that speed.
· Humans communicate slowly. Talking is slow even compared to thinking. When you finish my essay, you won’t have understood all of it the way I meant it; accurate communication is interactive and takes a long time.
· Humans learn slowly. We transfer information to long term memory at an average rate of no more than a few bits per second. My computer, which is a few years old, can transfer to its permanent memory, the hard drive, at millions of bits per second if I plug in a fast enough drive.
Why is it so hard to tell whether a relationship is going to work out? Because you have to learn a lot of information about each other first, and no matter how insightful you are you can only learn it at a few bits per second. The same goes for hiring a new employee, or deciding where to go to college.
This limits the level of expertise you can reach. If a task calls for more knowledge than a human can learn in a lifetime, you can’t do it yourself.
· Humans have poor memory. How many days back can you remember what you had for lunch? I have poor episodic memory and can only remember a few days back. Hardly anyone can remember every day for years back, though you could easily store all the information on one floppy disk.
· Humans make many errors. We need spell checkers even when we know how to spell. I’ve read that the error rate for clerical tasks like operating a cash register bottoms out at one percent and does not go any lower no matter how much you train.
· Humans have trouble with precise reasoning. Computer programming and designing complicated machines are slow and error-prone tasks. A new mathematical proof must be checked painstakingly before people will believe it, even though every step is intended to be simple and clear.
· Humans suffer biased judgment. See “Judgments under uncertainty: Heuristics and biases” by Tversky and Kahneman, Science, volume 195, pp1124-1131. It’s from 1974. A study in 1977 found that when people said they were 100% certain about some fact, they were right about 80% of the time. Politics, anyone?
· Each human has to learn everything from scratch, even things everybody agrees on, like how to talk, how to count, how to drive. It takes years out of your life.
· Humans don’t live long. All your knowledge, learned at a few bits per second for a lifetime, is lost when you die. The stars seem far away because, at the low speeds we can reach, it takes more than a lifetime to get to one of them.
· Humans are fragile. We need a spacesuit for something as simple as working in vacuum, and even then we have to watch out for radiation.
· Humans have hardwired, low-bandwidth senses. The tradeoffs are fixed: vision has high spatial resolution and low depth resolution, so we can’t immediately grasp radar data which has low spatial resolution and high depth resolution. The capabilities are limited: we can build cameras that take pictures in hundreds of wavelengths at once (like the spacecraft Galileo’s Near Infrared Mapping Spectrometer, perhaps 1980 technology, which takes pictures in 408 wavelengths), but we can only look at the pictures three colors at a time. It appears to be physically possible to build a sensor which simultaneously measures, for each incoming photon, its time of arrival, position on the detector, wavelength, and polarization. It would take a human hours to absorb the data that this instrument produced every minute.
· Humans have too much hardwired behavior. Even people who think the world is overpopulated want to have kids. Passengers tend to want to be drivers, even if they can’t drive as well as the current driver.
· Humans have high overhead costs. If you pump gas, a job a robot could do with once-a-month maintenance, you go to school for years and eat three squares a day. If you’re Einstein you might like to think 24 hours a day, but you have to sleep.
My claim is simple. In the far future, all these handicaps will be removed.
Yes. These weaknesses in the human design have high economic costs, so under free enterprise we will work to fix them. Robots will pump gas; there are prototypes in Germany. Einstein will think all night.
I see no technical barriers. The two key technologies are nanotech and artificial intelligence. I’d say that nanotechnology seems unlikely only to people who haven’t checked out the facts, and AI seems unlikely only to mystics. However, apparently most people are mystics who haven’t checked out the facts.
There’s no way to tell. We can see what’s possible, but we don’t know what’s easy and what’s hard.
What will our far future civilization be like, when we’ve fixed all these little bugs? It will be so much more complex than our own that we don’t stand much chance of understanding it yet. But I recommend trying anyway! Think about it.
Check out the Transhuman Web Alliance for some vague ideas. Unfortunately it suffers a high bogosity level.
updated 28 June 2000 and 26 January 2012