Liveblogging the Cold War: September 20, 1946: Eleanor Roosevelt
Ed Kilgore Smackdown of Michael Barone: The Biggest Obama-Hating Hack in the World: Hoisted from a Year Ago/Late Monday Smackdown:

Comment of the Day: Tracy Lightcap: AI:

Oh, horse manure...

It is true that by programming AI machines probabilistically (instead of, you know, trying to get them to behave like real human beings) we have gotten to the point where computers can do two things: help people solve puzzles by showing them answers and letting them figure out the rest themselves and control repetitive tasks in manufacturing processes.

Now let's ask ourselves tow questions. First, can we program computers to think? Or, to put it another way, can we program computers to come up with useful solutions to new situations? Answer = of course not. Computers can do things human beings have done that can be translated into probabilistic processes or that involve specific and repetitive calculations over a limited field of possibilities. Hence Jill Watson and helping new computer science students see the error of their ways in simple engineering calculations. Could Jill teach a second level class in computational theory? No.

Second, can we use robots to replace human beings even in the repetitive situations mentioned (NB: this excludes the entire service industry and most of agriculture)? Answer = no.

Take the Airbus series of jet liners. They can take themselves off, navigate themselves to a specific destination, and land themselves. So why have a crew? Because you might fly your jetliner through a flight of Canadian geese and have both engine destroyed right after takeoff. So you would be well advised to have a Sully on board to take over. The flight computer would have frozen up. Same reasoning applies for most production processes and for all service industry.

Here endeth the lesson.

Comments