What is the meaning of having an agent 'with general intelligence at human level'?
Who was more intelligent Albert Einstein or Charles Chaplin? If we want to give an objective and specific answer to this question, we should say immediately: ‘it depends on the kind of intelligence we are talking about’. Let us consider ‘intelligence’ at least in the very broad sense of H. Gartner, i.e., an initial taxonomy of this concept should contain at least the following kinds of intelligences: logical-mathematical, linguistic, spatial, musical, bodily-kinesthetic, in- terpersonal, intrapersonal, naturalistic and existential. So, the answer to the former question would require a deeper qualitative description. For example, we could require measur- ing how good was Einstein in acting and how good was Chaplin in understanding physics and mathematics. In general, a plausible answer would be that if Einstein and Chaplin were alive, we should perform a long test with both of them where we design particular tasks regarding at least the former seven types of intelligence. Moreover, in each of this topics, let us say, bodily- kinesthetic intelligence, we choose, for instance, a representative collection of sports and we create quantitative-qualitative tasks in any of the sports that allow us to compare the perfor- mances of both of them in a very well defined way. For example, the number of basketball free-shots scored, when one has 70 tries. This kind of general test would give us a global measure of the ‘intelligence’ of Chaplin and Einstein in a very comprehensive sense. In fact, due to the quantitative aspect of each of the tasks, we could obtain a specific numerical value of the (general) intelligence of each one in order to be able to compare them globally. So, we could finally get a concrete answer to our original question. In fact, let us assume that, hypothetically the performance of Chaplin on the linguistic, spatial, bodily-kinesthetic, interpersonal and music intelligences was notably better than Einstein’s, and that Einstein’s performance was better in the logic-mathematical and intrapersonal parts. So, we would conclude that, according to the performances of these two personalities during, let us say, more than 200 testing hours, Charles Chaplin is more intelligent than Albert Einstein. In fact, the broader we understand intelligence, the better we understand that Chaplin could ever be more intelligent than Einstein under those circumstances.
This has been criticized ever since. Especially if one traces the ontogenesis of an individual, that is, takes a developmental perspective on how adults acquire many capabilities through their life, it is evident that intellectual development is deeply influenced on bodily exploration
of the world; moreover, the kinesthetic capabilites change in close interaction to intellectual changes. The abstraction made by AI’s forefathers therefore seems much too rigid to us, and has to be complemented by an embodied perspective. Another fundamental point for judging intelligence is the one of specialization. A very common argument for (non-) comparing intelligence can be described as ‘each person is good in something very unique, which is quite specific and differs between people, therefore one cannot make comparisons concerning (global) intelligence’. To this argument there is a quite general remark: It used to happen that people that are very good at very specific fields, simultaneously start to develop an ability to be able to understand faster other fields, related (and not some much) with his (her) specialized field. For example, Bertrand Riemann in order to get a deeper understanding of the distribution of the prime numbers had to develop a new way of seeing functions with complex variables.
Grothendieck and Chevalley in order to obtain a better formalization of the classical notion of affine variety had to come to a more categorical understanding of the core properties of commutative rings and their prime spectra in terms of affine schemes. The work of Descartes on obtaining better models of geometrical objects led him to use methods of classical algebra in order to create his ‘analytic geometry’. Andrew Wiles’ effort on proving the insolvability of Fermat’s Last Problem (a pure arithmetical problem at the first sight) led him to find more precise connections between the areas of elliptic curves and modular forms.
The creative enterprise of Carl Levi-Strauss of making anthropology into a more formal science guide him to use Saussure’s structural linguistics and some aspects of combinatorics for developing his structural anthropology. Now, it is important to clarify that these phenomena occur, only if a person is flexible enough, intellectually speaking, to grasp deeply into other fields. Thus, a practical and vivid ‘intellectual openness’ would be necessary for getting broader scopes in our comprehension of very specific intellectual fields. Here, the expression ‘intellectual field’ encompasses all the possible areas related to at least one of the former types of intelligence, from classical sciences to any kind of sport and artistic activity. Finally, we want to draw attention to the fact that some people are able to learn regularly within some months (in the university) several kinds of scientific and artistic disciplines, to play different kinds of sports and music (instruments), to interact with many people at a wide range of levels in order to obtain specific and differentiated goals with any of them, to move to several places in minutes just by following quite abstract symbolic information expressed in oral and written language, to select at a quite unique and creative way styles of clothes, food and places to live in; to integrate him/herself relatively fast into a very sophisticated system of services, products, artificial laws and cultural schemes; to obtain a world wide perspective of the current political, social and economic state of the world just by processing light and sound configurations of an electronic device (e.g. smart phone, computer); to create quite innovative solutions in order to improve the life’s quality of other people and to modify whole pieces of natural environments in prefixed extraordinary ways; among many, many others. So, any agent having general intelligence at human level should be able to perform at least the sort of activities described before, since these tasks are typically made by humans.
Now, what is, objectively speaking, the state of the art of the current AI existing agents
regarding general intelligence?
Here, it is worth to mention that in the particular field of AGI there are at least two external sources which strongly influence the objective estimations of people concerning this issue, namely, the cinematographic and the machine industries. In particular, it is quite plausible to think that films like 2001: A Space Odyssey; Artificial Intelligence; I, robot; Ex- machina and The Star Wars Saga, among others; have a strong influence, at least cognitively speaking, on people’s own perception of the current state of the art of AI devices. Effectively, the goal of some of these films is precisely to produce the temporal illusion that a hypothetical researcher was (almost) able to produce AGI. Of course, the directors wish to present their stories in a very contextualized and realistic way and we could say that as pieces of art these movies are very creative contributions. However, the objective contribution of these films to the development of AGI is a quite different story. Now, how could we distinguish between scientific, fantastic and marketing-driven results on AI research?