Two distinctly different visions in the artificial intelligence community are at war, and the outcome could determine the future of humanity.
The Yin
Just over a decade ago, I attended a discussion dinner sponsored by the Churchill club — a group I sorely miss. Formed by the wives of Silicon Valley executives, it was a worthy venture that sought to shed light on various important topics via discussion panels. Hats off to you ladies. Well done.
This particular evening, the event was held at the Computer History Museum in Mountain View, CA. By the way… If you’re into computers, that site is an absolute must-visit.
The subject matter this particular night was AI and chess, the guest of honor being the current world champion and one of the all-time great chess players — Magnus Carlson. Look up the magnificently counter-intuitive, brilliant end-game in a match against Vishy Anand — a wonderful player in his own right. It was astounding.
Why a chess master in a discussion of artificial intelligence? Because nerds have been programming computers to play the game since they came into existence. In other words, computer chess (as well as music, and other games) might well be considered the genesis of the search for artificial intelligence. How far has it come? Carlson admitted that the best he could likely do against a top modern chess program is to draw.
Mr. Carlson, to my way of thinking — wise beyond his years — also stated that he often uses chess simulations to set up problems for him to solve, but he has no interest in competing against them. Put bluntly, he feels that the technology should be a tool in human hands and there’s no satisfaction to be gained or lost by competing with them.
The Yang
Alas, I attended another Churchill club discussion entitled Augmenting human intelligence that proved far more disturbing. At least to me. You can view the discussion on YouTube by clicking on the link above.
The two featured guest speakers were John Wilkens, the driving force behind IBM’s Watson, and Yoshua Benigo, a Montreal professor. The difference in their attitudes and beliefs was as stark as night and day. Mr. Wilkens sees AI as a tool for humans to exploit, and to use for their benefit as they see fit.
Yoshua on the other hand seemingly believed that there are responsibilities that should be handed over to AI. In other words, for AI to not only come up with answers, but to implement solutions with the wisdom and impartiality it’s creators might lack.
At the time, self-driving cars were in their infancy so the discussion took a turn towards morality, or more specifically morality as it applies AI. The classic conundrum for driver-less vehicles is that when hitting something is inevitable, what does it choose?
A worst case conundrum would be having to choose between a senior citizen with a walker and a mother with her child. You have to hit one? Which will it be?
I have a distant non-blood relative that actually worked on a similar problem for the U.S. Military’s armed drones. Yes folks, there is at least a consideration of giving machines the power of life or death.
What is AI?
As happens with just about everything that marketing types, or other poorly informed or incentivized people get involved with, the definition of AI has shifted.
Originally conceived as the complete replication of human intelligence sans the emotions and primal impulses, it’s now used to describe anything that involves heuristics and pattern recognition. Is that the end-all of intelligence? One might ponder if the emotions and primal urges aren’t actually part of the equation.
Most of what current AI does is simple aforementioned pattern recognition and replication. Siri is an example. Phoneme recognition and libraries reconstruct words, definitions and lingual patterns are searched to reconstruct context, then contexts are searched to forms a concept, etc.
This indeed mimics what we do as humans, but…
Why is AI be better at some things?
The strength of AI is the amount of data it can scan and process to achieve a result. Our brains can only hold so much — the reason we specialize in certain fields. An AI can cross field boundaries with impunity, perhaps arriving at solutions that might escape us mere mortals.
And AI also won’t suffer from memory issues. If the data is there, it can find it. Sometimes we humans forget.
Why might AI not be better in all ways?
The human mind is a very, very complex mechanism that functions using methods we still only vaguely understand. It can arrive at conclusions it seemingly has no right to. Call it what you will, a spark, intuition, a sixth sense. All that is very difficult to program.
And therein lies another issue with “AI”. It’s using data that we provide. There’s an old saying: Garbage in, garbage out. Artificial intelligence relies on what we humans provide it, and we are fallible. Not to mention biased, often misinformed, etc.
The Future
Alas, while I’m firmly in the “AI as tool” camp, there are certainly forces that are not. You can see that in the current gold rush of data centers, and a premature willingness to fire workers and replace them with AI.
Mostly because many merchants have no allegiance except to their money and they’ve been sold on the idea that AI will save them money. Currently, it won’t and some businesses are learning that the hard way. AI is not nearly as far along as some would have you believe.
All I can say is good luck to all of us. We’ve had atomic weapons for over eighty years and we haven’t had a nuclear war yet, so there’s hope.