Computational Intelligence and Consciousness

Eduardo César Garrido Merchán

In recent years there have been many advances in artificial intelligence, especially in the field of automatic generation of texts and images that sometimes compete successfully with human productions. In light of this, the media, and even some scientists, have rung the bells announcing that we are on the verge of creating conscious artificial intelligence, which would compete with human beings as our equal. But others believe that this goal, if it were possible (which is not clear), is much further away than some think.

In an article signed by Eduardo César Garrido Merchán and Sara Lumbreras and published in the journal philosophies with the title Can Computational Intelligence Model Phenomenal Consciousness, the authors review Bertrand Russell's analogy, which asserts that consciousness and intelligence are closely correlated. In other words, any entity that possesses consciousness will also possess a high level of intelligence, and vice versa. In a way, this analogy is similar to the Turing Test, which is much better known.

The authors of the article use Bayesian calculus to conclude that Russell's analogy is false. They also give some counterexamples to this analogy, such as the fact that people with low intelligence (subnormal people) are obviously conscious, a situation that also occurs in some cases of people in a coma.

To analyze the situation, the existence of two different concepts is proposed, alluded to by the title of the article:

1. Computational Intelligence: The ability to solve problems that belong to the field of computational problems. It is evident that human beings have a certain capacity of this type. It is evident that, in many of these problems, computer programs easily beat us. This concept is quantifiable, although difficult to measure, because criteria such as William Stern's IQ tests only provide a lower bound of the subject's real intelligence, as the test may be negatively affected by fatigue, mood, or even the desire to obtain a lower IQ.
2. Phenomenal Consciousness: Being aware of certain experiences, such as movements, colors, sounds, sensations, emotions, feelings, and (in the limit) one's own identity. These experiences are grouped under the name of qualia and are not quantifiable, since as their name indicates they are qualitative. For this reason, the authors of the article consider this variable as binary; that is, for a certain being, it can only take the values ​​zero (it does not have it) and one (it does have it).

Bertrand Russell

Russell's analogy relates these two concepts and considers that one arises from the other. Phenomenal consciousness would be something that emerges automatically when computational intelligence reaches a certain value, and vice versa. The authors of the article deny this relationship, and justify it with reasoning based on Bayesian calculation, whose conclusion is that the correlation between the two variables is zero (that is, one does not depend on the other).

Animals provide another counterexample, as their computational intelligence is very low (with a few exceptions, such as bees), but they obviously have a phenomenal consciousness that, although it may be less rich than that of humans, is by no means non-existent.

One of the most compelling arguments is the fact that every intelligent computing system built by man so far can be modeled as a Turing machine. This applies even to the most complex artificial neural networks used for deep learning, which, although they adjust hundreds of billions of parameters, are basically reduced to the execution of an algorithm (see this post in this blog).

Remember that the Turing machine has an infinite memory, so it is actually more general than any computer program, whose memory, although enormous, is always finite. However, given the simplicity of the design of the Turing machine, no one had thought of claiming that such machines could one day become conscious. It is therefore concluded that intelligent computing systems based on the von Neumann architecture, although much more intelligent (computationally) than man, will never be able to acquire phenomenal consciousness. To achieve this result, very different entities would have to be designed, more similar to the human brain. And it is not clear that this is possible.

The same post in Spanish

Thematic Thread about Natural and Artificial Intelligence: Previous Next

Manuel Alfonseca