Successful Artificial Intelligence
by JS
I have been thinking a bit about what constitutes successful artificial intelligence research. I have heard the usual saying that once a problem in artificial intelligence is solved it is no longer considered AI. I think we can unpack this dictum a bit more. First, I do not think that successful research in artificial intelligence necessarily requires that some “real” world problem be solved. The goal of AI should be to better understand the phenomenon of intelligence. This is a goal shared by other disciplines as well (and as a consequence AI is becoming very interdisciplinary), but the AI approach to this question seeks to model intelligence using computation.
Now getting back to the fallicy that AI needs to solve problems in the real world, consider that the real world does not care whether the solution to a problem is “intelligent.” What makes a line of research successful by the standards of the real world are
- that there is a real need for a solution to a particular problem
- that research uncovers a sufficiently good solution.
You’ll notice that neither of these criteria actually requires that we understand intelligence. To put this in more concrete terms, consider Google. The success of search on the web is a product both of the need for automated search and the ability of page rank to generate good search results, but should we expect that a clear and complete understanding of how Google search works also conferes a deep understanding of intelligence?
One particularly important critique of this view is that we should expect research in AI to yield systems of increasing intelligence, and since many currently open problems can be solved by more intelligent systems, we should expect that AI results in “real” world solutions in addition to a deeper understanding of intelligence. The former is an epiphenomena of the latter. Without observing the growth of practical solutions, we should rightly doubt progress in the field. The problem with this view is that practical solutions to problems seem to be punctuated (requiring a precise alignment of criteria 1 and 2 above), whereas our understanding of intelligence can increase gradually over time.
Comments
Better understanding of intelligence is a worthy goal for AI and entirely achievable. And as you note it will spin off applications that become increasing shopisticated in their dealings with us normal humans. (But this will be because the designers have thought long and hard about what’s required and how the system should function and react.)
You are absolutely right that as soon as intelligence is built into something, it becomes simply information processing and no longer intelligence. At least that’s the perception.
I like these common sense AI fundamentals. Where AI goes off the deep end is the holy grail search for a system that independently evolves to something more than what its designers intended and actually does so in a useful manner. We get deep into philosophy and even religion here when we start the argument about the ability of natural systems to develop complexity and intelligence without a designer. I’m still waiting for someone to prove this can happen. (nobody has artificial life crawling out of a test tube yet!)
I remember playing with something Chris Adami at Caltech wrote back in the 90′s (a kind of artificial life using processing code) that claimed to evolve and by so doing achieve more things than were ever anticipated by the designer. But imo it never really proved anything about the ability to create systems that independently evolve to become something more than their creator intended and develop intelligence on their own. So I’m still waiting for proof on this issue, although I haven’t followed the debate recently on this topic. (Unfortunate, the scientific debate usually gets polluted by fanatics on both sides, who stop dealing with the issue in a scientific / rational manner and start name calling.)
Some more thoughts: I think the real problem is “A” in AI. It should be N for natural. That would get people off the holy grail , tilting at windmill approach that sometimes characterizes the A part of AI. NI would be really smart people seeking to understand “I” better and become more “I”. They would be designing really smart systems that have a lot of “I” built into them. Entirely natural and very successful.
Eventually people who didn’t see the N might think it was A, but that’s simply the wizard in oz effect.
The question of how intelligence came into the universe is an interesting one, but is sort of besides the point of my post. It is a phenomenon that we can study, either through psychology or using computation.
As for where life comes from (and by extension intelligence), we have a convincing theory, evolution, which describes a mechanism and a system of deducing hypotheses that have all turned out to be correct (so far).
Beyond the normal vigilant skepticism that is the responsibility of all scientists, I don’t see a need to posit some deeper philosophical or theological theory for the origin of life. In particular, I don’t think we can conclude from lack of proof that something mysterious is missing. The theory (and the evidence) might merely (and somewhat prosaically) be incomplete.
Bookmarked