Progress of AI: 50 Years and Counting

[edpell]

How come AI never makes an progress? OK this may be slightly over stated but it has been 50 years since we thought we could make an AI soon and we are no where near an AI.

 

[Hurkyl]

It's very hard to convince people determined to dismiss every advance as being "that's not intelligence". I've seen some critics effectively admit that "a computer can't do it" is part of their definition of "intelligent".

 

[edpell]

The Turing test would prove intelligence to me. Even a typed conversation at the level of a six year old would work for me.

 

[Hurkyl]

I've seen some humans I think would fail a Turing test.

Part of the issue is that there isn't much use for that sort of intelligence. An AI that can converse at a kindergarten level isn't very useful. An AI that can accurately guess which of the trillions of web-pages on the internet you want to see is very useful.

 

[edpell]

The kindergarten level is a stepping stone to the post-doc level. How much do we pay doctors per hour? That is a big market with high profit margins.

That there are stupid humans I already know.

 

[Hurkyl]

The kindergarten level is a stepping stone to the post-doc level. How much do we pay doctors per hour? That is a big market with high profit margins.

A medical AI would be much more effective than a human AI that was trained in the human way to be a doctor.

 

[phyzguy]

How can you watch IBM's "Watson" software beat the best humans in the world at Jeopardy and say there has been no progress?

 

[adaptation]

How can you watch IBM's "Watson" software beat the best humans in the world at Jeopardy and say there has been no progress?

Seconded.

 

[Chronos]

AI success to date is largely due to massive computing power, not efficient decision making. Put a little natural selection into the mix and AI would be compelled to get real smart in a big hurry of go extinct.

 

[Adyssa]

I just recently finished this book, and it was an amazing read, I thoroughly recommend it if you are interested in AI, and it may change your perspective some.

Gödel, Escher, Bach: An Eternal Golden Braid - Douglas R. Hofstadter

I'm currently reading The Emporer's New Mind - Roger Penrose, which has a similar theme.

 

[chiro]

How come AI never makes an progress? OK this may be slightly over stated but it has been 50 years since we thought we could make an AI soon and we are no where near an AI.

One thing you should be aware of is that in the field of AI, (and in science in general), people usually work on very specific problems.

Sometimes people bring a variety of results together and create some great ideas for tackling general problems, but the majority of problems that people tackle are very specific.

When you have a problem and you want to solve it, you need to get a concise definition of the problem before you can have any chance of solving it.

Now think about a large problem such as general intelligence: in cases like this it's really hard to precisely describe the problem in great detail. It is easier however to do it for smaller problems because the scope is reduced and it is a lot clearer to actually state the problem, and hence make efforts to solving it.

There have been many great achievements and applications of math, engineering, and science to intelligent applications, and this in my view, is not something that should be trivialized: science, and in particular innovation is hard and should be appreciated for what it is.