- The American Conservative - http://www.theamericanconservative.com -

A Simplistic Artificial Intelligence

At BetaNewsRobert X. Cringely writes [1] that fifty years ago attempts to create artificial intelligence (AI) failed because there was not enough “processing power” to do it: “But thanks to Map Reduce and the cloud we have more than enough computing power to do AI today.”

At The Washington Post, Dominic Basulto tells us that AI is “the next big thing” for Silicon Valley start-ups [2]. Not only will it be created, it will be done so easily and cheaply:

AI will move from something that took tens of millions of dollars and thousands of people to create, to something that takes tens of thousands of dollars and can be created by a group of kids after an all-night Red Bull session. When they do, then we’ll know that visionaries like Erik Brynjolfsson and Andrew McAfee were right – we are entering the dawn of the age of artificial intelligence.

As Alan Jacobs [3] and Fredrik deBoer have pointed out, these facile visions of AI being created by processing power simply assume that intelligence is a matter of data. It’s not. DeBoer [4]:

The problem with AI is that we don’t really know what the problem is, or agree with what success would look like. With your cellphone (or any number of similar rapidly-improving technologies) we are perfectly aware of what constitutes success, and we know pretty well how to improve them. With AI, defining the questions remains a major task, and success a major disagreement. That is fundamentally different from issues like increasing processor power, squeezing more pixels onto a screen, or speeding up wireless internet. Failing to see that difference is massively unhelpful.

If people want to reflect meaningfully on this issue, they should start with the central controversy in artificial intelligence: probabilistic vs. cognitive models of intelligence.

3 Comments (Open | Close)

3 Comments To "A Simplistic Artificial Intelligence"

#1 Comment By Michael Guarino On April 21, 2014 @ 4:42 am

When I was in college and interested in AI research, I remember talking with a classmate about some of the software trying to pass the Turing test. Apparently, the best candidate he heard of basically responded to any query like a total asshole, which is a pretty thorough indictment of the predictability of assholes. It could be apocryphal, but I think it is a very good way of charting the minimal progress of the old understanding of AI, software that can interact with natural language like a human.

Also, the enthusiasm with cloud computing is a distraction here. The cloud gives parallelism, but that is not necessarily going to be a win. Areas like computer vision still need significant boosts in latency to be anywhere near where a comprehensive AI needs to be, and throwing everything in the cloud is not going to help you. An Amazon EC2 instance is worse than a laptop for that sort of processing. Add wire latency, and it is basically the stupidest solution. But please, let the journalists have their fun.

#2 Comment By SDS On April 21, 2014 @ 10:11 am

Without the corresponding level of artificial wisdom; AI will prove to be our worst nightmare….

#3 Comment By HeartRight On April 21, 2014 @ 5:19 pm

‘SDS says:
April 21, 2014 at 10:11 am

Without the corresponding level of artificial wisdom; AI will prove to be our worst nightmare….’

Agreed. So this computer meets Ayn Rand, and gives a whole new meaning to Object Oriented Programming.