The Singularity Q&A

Q: If AI is possible, why has it not been created yet?

A: Looking at this short timeline from the USA Today, you would get the impression that real AI has been around since the mid 50's and just keeps getting better. Since we obviously don’t live in the world seemingly described by the editors of this list, allow me to suggest that they were, perhaps willingly, duped by what I call “strong AI imposters” I’m not referring to narrow AI implementations that do not claim to be strong AI but merely borrow AI terminology for marketing reasons. By imposters I mean programs that are specifically intended to display facets of what the general public often considers to be the essence of genuine intelligence. Facial expressions, understanding of natural language, “autonomous navigation”, and the creation of “original art” are all recent examples of tasks supposedly achieved by today’s software. But these programs are no more than well-rehearsed puppets – automata whose limitations become quickly apparent in real-life situations.

There’s nothing more disappointing than, in a matter of minutes, “seeing through” some “revolutionary” program that is supposed to be truly intelligent, artistic, or empathetic. Like sophisticated animatronics or ornate cuckoo clocks, imposter AI can often be amusing as novelty items. But tens of thousands of lines of computer code simply cannot contain the evolving character of an engaging conversation, the brilliance of a talented artist, or the emotional depth of a close friend.

When real AI is created, we can be confident that human programming will describe its functioning, but that the overwhelming bulk of knowledge and talent will be learned and developed by the program as it is run. (The program is the mind pattern; the mind does the learning.) Only in this way can an AI ever hope to expand its effective range in a way that would avoid tripping our senses of repetition and oversimplification that unmask the strong AI imposters.

So, why has this not been done? Well, to engage in a bit of oversimplification myself, creating a mind is hard. There is a tendency in scientific fields to gravitate towards the simplest theories. On the whole, this is a valuable and often effective trait. But the secret of intelligence, unfortunately, does not appear to be very simple. If it were, we probably would have strong AI by now, because fifty years of AI research has been fairly thorough at trying the simple theories that might have explained general intelligence.

While the pessimist may regard these unsuccessful approaches as disheartening failures, these projects were worth attempting for a number of reasons. First, these simple approaches were actually possible in eras where computing power was not nearly as abundant as it is today. Secondly, they may well have solved theoretical problems that will figure into the eventual AI solution. And, finally, even projects that prove to be dead ends at least flesh out the map of the possibilities by telling others where not to waste their time.

But the emotionally disappointing track record of AI, along with the now understood difficulty of the problem, have no doubt dissuaded many potential researchers and investors. Worse, the culture of the computer programming community has actually come to stigmatize proponents of real AI as pie-in-the-sky dreamers at best, and useless crackpots at worst. Additionally, as I discussed earlier in my list of requirements for a successful AI project, Artificial Intelligence requires interdisciplinary cooperation – something that historically never comes easy to most scientists. Software engineers generally get along well with mathematicians, but find cognitive scientists to be a very alien species.

It is also possible that today’s computers are still not powerful enough to run the first successful AI program at any useful speed. But I would argue that we don’t really know, because interest in AI has not risen nearly as fast as the performance of computers. We have not come close to designing efficient prototypes for minds that would nevertheless exhaust today’s supercomputers – let alone tomorrow’s.

In my next responses, I will specifically discuss a few of the AI projects that stand out as milestones on the path to genuine AI, as well as some of the reasons why naysayers have justly tagged the ongoing procession of AI projects as some of the most consistently over-hyped endeavors in the known universe. For entertainment purposes only, I’ll also highlight a few of the more interesting AI imposters.

A little later, I will discuss some of the more significant ongoing projects currently attempting to create genuine Artificial Intelligence.

[Back to Futurism] [Back to Main]  ©2002 by Mitchell Howe