One of the known issues in the field of Artificial Intelligence is that there is no generally accepted formal or informal definition of intelligence.
I have learned about this problem from Legg & Hutter and Chollet.
Each of these two papers proposes a definition for machine intelligence:
Def. 1 [Legg & Hutter]: The ability to achieve goals in a wide range of environments.
Def. 2 [Chollet]: The intelligence of a system is a measure of its skill-acquisition efficiency over a scope of tasks, with respect to priors, experience, and generalization difficulty.
Both papers start with a review of previous attempts at defining intelligence, and then proceed to present a proposal for a formal definition.
One point of disagreement is whether the definition for machine intelligence should capture only human-like intelligence [Chollet] or more general, non-human types of intelligence, Legg & Hutter call this Universal Intelligence [Legg & Hutter].
Differences between definitions
The agent under Def 1.
- it does not need know the initial prior
- it aims to be maximally intelligent for any given environment, in other words, Universal
The agents under Def 2.
- it aims at emulating human-like intelligence
- relies on its builder to provide and control priors
- the provided priors are human priors
References
- arXiv:0712.3329v1 - Shane Legg, Marcus Hutter, Universal Intelligence: A Definition of Machine Intelligence
- arXiv:1911.01547v2 - François Chollet, On the Measure of Intelligence
