What does AI mean by ‘intelligent’?
Up till now this question hasn’t been hugely important. But now that we’re deluged by optimistic hype about self-driving cars being controlled by intelligent AI systems, it really matters.
What if AI’s meaning of ‘intelligent’ has little to do with human intelligence? What if AI, since its inception in the 1950s, has purloined words used to talk about human intelligence (‘reasoning’, ‘knowledge’, ‘semantics’, ‘deducing’, ‘inferring’, ‘deciding’, ‘knowing’, …) and surreptitiously redefined them to refer to electronic things inside computers? And without a sound justification that these electronic referents are relevantly similar to the processes and structures inside human minds?
What, in fact, if AI terminology is a massive academic con job to get funding and attract students?
People believe that self-driving cars are controlled by systems that have human-like intelligence. Why? Because they’ve been told so, over and over. But on close examination it’s found that these vehicles are controlled by systems that are not even moronic imbeciles. The problem isn’t that more tweaks are needed, or that computers aren’t fast or parallel enough, or that the sensors aren’t sophisticated enough. The problem is really fundamental: AI doesn’t know how the mind works. It doesn’t know the basic principles.
A very few reports have raised this critical issue, for instance here. And I’ve discussed it before, for example here and here. But well-supported and technically detailed arguments would put the case more fully.
Wikipedia provides a great starting point, since it embraces the fake AI lingo. I’d like to start an analysis of key AI terms with the opening paragraph of the Wikipedia article titled Reasoning System:
“In information technology a reasoning system is a software system that generates conclusions from available knowledge using logical techniques such as deduction and induction. Reasoning systems play an important role in the implementation of artificial intelligence and knowledge-based systems” [emphasis added].
In the following few posts I’ll analyze in technical detail each of these claimed mental properties of AI systems, starting with the property, or characteristic, of deductive reasoning.
associative-ai.com