Mike Carrell wrote:
I spent an appreciable fraction of my career in the early days of robotics with mechanization and robotics applications at RCA. Naïve chatter about human-less culture and artificial intelligence overlooks fundamental things. 1] brains are fundamentally different from computers. Brains learn, can be taught, but dont run on algorithms. Progress in ,say. household robots, is made with electronically simulated neural networks. These can learn tasks like a child or pet, but even the designers do not know exactly **how** the brain does its job. Same with synthetic, computer-simulated evolution by natural selection. It is incredibly efficient, but not something that can be programmed. A computer algorithm can create the illusion of intelligence, but it is a fake. Real electronic brains [where are you, Isaac?] may be built, but they be no more manageable than children or pets.
I do not think that is a forgone conclusion. There are life forms with advanced brain functions that are completely pre-programmed, manageable, and never unruly, such as a hive of bees. They gather materials from the environment, construct honey combs and other complex structures, and perform complex care of eggs and so on. Those are real brains but they do not learn. They exhibit no curiosity. They never rebel against authority or waste time, any more than your respiratory system does, which is also controlled by brain tissue. Bees have many remarkable mental capabilities, such as the ability to distinguish people and other animals from inanimate objects. They can do this better than the best artificial intelligence robots today. A similar level of artificial intelligence could perform many useful tasks, such as clearing the dishes from table and driving automobiles.
The 2005 DARPA Grand Challenge automobile managed to drive 300 miles autonomously in 7 hours. I do not know if it employs any artificial intelligence, or only ad hoc solutions to the problem. But it is capable of doing useful labor that only people could do a few years ago. The addition of artificial intelligence to the level of a bee-hive would enhance this performance, and reduce the likelihood of accident. I think we can accomplish this much in the next 50 to 100 years, even if we cannot make machines as smart as, say, Labrador retrievers. (When I say "a bee-hive" I mean the total brain tissue and aggregate capabilities of a hive of bees. They are capable of actions as a group far beyond the abilities of individual members.)
Whether these machines are actually intelligent or only act that way strikes me as irrelevant -- even meaningless. Google's translation software and the DragonSpeak voice input software do not understand language in any deep sense. I doubt they understand as much as a hive of bees would. But they are capable of useful language-related work so what difference does it make? Simulated intelligence is as good as the real thing, for practical purposes. I do not think we will need to worry about simulated intelligent machines disobeying orders any more than a queen bee has to worry about rebellious worker bees, pace "Maya the Bee." Perhaps if simulated or artificial intelligence reaches the level of a Labrador retriever we will need to worry about this. Dogs are in many ways as smart as we are, and a lot smarter when it comes to specific tasks and knowledge such as tracking and hunting down small animals or herding sheep. This is akin to saying that a peregrine falcon or a bat can outfly the best human airplane pilot. I can't imagine any pilot would dispute that! A bat does not have much brain tissue but gram for gram that particular tissue happens to be better at flying than human brain tissue can be, no matter how much training we subject it to.
This is getting off topic but . . . A recent issue of Time magazine has several articles about animal intelligence:
<http://www.time.com/time/health/article/0,8599,2008759,00.html>http://www.time.com/time/health/article/0,8599,2008759,00.html
It addresses the Big Question: Do Animals Think? That is a really dumb question, in my opinion. Of course they think! Do you suppose bees gather honey and build combs by magic? It is a function of their brain, and brains think. I guess what the Time authors mean is, are animals sentient (self-aware), or creative, and can they form thoughts, or talk. I studied these topics at Okayama U. Dept. of Biology 35 years ago, and nobody there doubted that animals think. Not only do they think, they sometimes outthink and outwit a professor. As I said here before, try keeping a Japanese badger (Nyctereutes procyonoides) from eating your watermelon and you will soon learn who is smarter when it comes to garden fences. I doubt they can disguise themselves as beautiful women or teapots as they do in Japanese folktales . . .
This Time article discusses a male bonobo named Kanzi with an English vocabulary of 384 words, including some abstract words and concepts such as "late" and "want." Years ago when people first taught primates sign language and other types of human language, a long, drawn-out argument ensued as to whether these animals really understood or whether they were making random movements while "aping" the people. I had no doubt the animals really understood because there were many instances in which the animals communicated information they had and the humans did not, such as the location of a toy, or what the primate wanted to eat, or some recent event the person did not witness. Opponents made irrational arguments similar to those against cold fusion. I recall one saying that the female chimpanzee Washoe might be imitating people because the person doing the ASL signing continued to sign subconsciously while speaking English to another person. That is, the person made slight motions of the hand while speaking, being fully bilingual. I am not surprised by this, but it has no relevance to whether the chimpanzee in question was actually communicating or merely imitating. Washoe might have been only imitating even if the person did not subconsciously use ASL; and she might have been willfully signing and expressing her thoughts even if the person did not always repeat everything in ASL. As I said, there is no way the person could have accidentally telegraphed the answers to some of the questions and topics of discussion because the person did not know the answers -- only Washoe did.
- Jed
