The truth is that "Artificial Intelligence" has been a misnomer since the term was coined in the 1950s. "simulated intelligence" would be a better term, and in fact in the paper that coined the term "AI," the authors themselves say their "study is to proceed on the basis of the conjecture that every aspect of intelligence can be so precisely described that a machine can be made to *simulate it*." Well, then call it a simulation, not an artificial version of intelligence. This is not a small distinction. A flight simulator can't get you across the American continent in 5 hours, but artificial flight can. And this is the proper way to understand what an LLM or some other AI technology is. Its job is to seem intelligent, not be intelligent. Look at the Turing test as well. It doesn't test what the computer is doing, rather it tests how the computer's output seems to a human observer. All that matters in a Turing test is appearance, in other words simulation. Secondly, intelligence has never been "precisely described" by anybody working on AI. 10 researchers will define intelligence 10 different ways. The same researcher might define intelligence differently on different days. This goes back to Timnit's critiques about the scope of AI. The whole "field" is fundamentally undefined, and continually falls back on woefully vague ideas about what intelligence is and how it works.
The truth is that "Artificial Intelligence" has been a misnomer since the term was coined in the 1950s. "simulated intelligence" would be a better term, and in fact in the paper that coined the term "AI," the authors themselves say their "study is to proceed on the basis of the conjecture that every aspect of intelligence can be so precisely described that a machine can be made to *simulate it*." Well, then call it a simulation, not an artificial version of intelligence. This is not a small distinction. A flight simulator can't get you across the American continent in 5 hours, but artificial flight can. And this is the proper way to understand what an LLM or some other AI technology is. Its job is to seem intelligent, not be intelligent. Look at the Turing test as well. It doesn't test what the computer is doing, rather it tests how the computer's output seems to a human observer. All that matters in a Turing test is appearance, in other words simulation.
Secondly, intelligence has never been "precisely described" by anybody working on AI. 10 researchers will define intelligence 10 different ways. The same researcher might define intelligence differently on different days. This goes back to Timnit's critiques about the scope of AI. The whole "field" is fundamentally undefined, and continually falls back on woefully vague ideas about what intelligence is and how it works.
Timnit real scholar