It’s easy to say generative artificial intelligence. Or general. Sometimes, however, we confuse terms that are not synonymous. General and narrow artificial intelligence they mean two different things. AND generative intelligence yet another. These are differences that also matter from a regulatory point of view. It is worth, then, to clarify things a little.
In the meantime, we can talk about “artificial intelligences“, in the plural, because there are different types and ways to make a computer capable of making calculations that in some way simulate human intelligence. Again, we must never forget that artificial intelligence is still a branch of computer scienceand not a science in itself.
In any case, it is scientifically proven that intelligent machines can exist. Alan Turingin a work published in 1950, Computing Machinery and Intelligence, he wondered if a machine can think. And, since “think” is a rather ambiguous and difficult term to define, he rephrased the question: can a machine do what we, as thinking subjects, do? Turing’s mathematical proof says that yes, it is possible. But at that time (and partly still today) it was not possible to achieve that result in practice.
Artificial intelligences
For this reason Turing formulated an experiment to recognize an artificial intelligencecapable of thinking like a person. Is called The imitation gamethe imitation game, but is generally called Turing test. The idea is that a machine capable of passing that test is capable of deceiving a human being, and thus is thus demonstrated to be “as if he were” capable of thinking.
The problem is that many artificial intelligences (or “expert systems”, as they have also been called) were capable of doing one thing surprisingly well, but not all others. They could act within a “restricted” scopetherefore, with respect to the objective of a “general” artificial intelligence, i.e. capable of doing everything a human being does and, potentially, better.
Agi, General artificial intelligence
General artificial intelligence (or Agi, from the English acronym), is therefore an AI capable of doing many different things. It is not limited to the “generative” aspects of some of the narrow artificial intelligences (those capable of doing only one thing: creating a drawing but not making a coffee or answering the question: What time is it?), but can do everything, including generalizing problems, deriving principles from concrete problems.
