Artificial Intelligence
A podcast that I listen to, the Coder Radio, recently discussed how overhyped the field of Artificial Intelligence was and how the marketing departments of companies use the term when they shouldn’t. Which got me to thinking about when the term artificial intelligence should & shouldn’t apply. Intelligence is defined by Google Dictionary as “the ability to acquire and apply knowledge”. Pretty straightforward. IQ tests after all are claimed to be biased against the uneducated because they test the application of knowledge that some individuals haven’t had a chance to acquire. Artificial is a little more complicated. Again, the Google definition is “made or produced by human beings rather than occurring naturally.” But I would argue that to be more fitting definition of the word ‘synthetic’ than ‘artificial’. Instead I’d look at the root of the word, which is artifice, or “clever or cunning devices or expedients, especially as used to trick or deceive others.” By that definition, it would seem the marketing department would be well within their rights to call anything they wanted “Artificial Intelligence” as long as it posed at least a decent imitation of skill to the layperson.
The field of machine learning seems to be the most honest form of artificial intelligence out there. Genetic algorithms, neural networks, reinforcement learning, bayesian algorithms are all approaches for computers to acquire knowledge. Once trained, these models are then used in broader computer programs to apply the skills they’ve built in a computationally fast and efficient manner. But at the end of the day every single one of those trained models are by necessity of an exceptionally narrow scope. Worse, the optimizations that they arrive at are the result of meticulous human pruning, re-balancing, and tweaking. This in many ways makes them only one step removed from the human powered mechanical turk of the 18th century. Fascinating to see, but laborious to execute and ultimately dependent on human operation.
Which brings me to my final point, why does any of this matter in the slightest? Why the paragraph of pedantic word pondering above? My closing arguement is that all artificial intelligence up to now has just been clever tricks intended to mask the fact that we have yet to create a facsimile of intelligence. Siri, Google Assistant, and Alexa are no closer to the dream of true intelligence than the mechanical turk three centuries ago. Such an intelligence, if we ever develop it, will be as far separated from artificial intelligence is as calculus is to colors.
comments powered by Disqus