Artificial intelligence or computer algorithms?

John MacCarthy

The term artificial intelligence appears frequently in the media. It is usually used to refer to a computer application that behaves in a way that appears to be intelligent. But is it really intelligent? Or is this a case of meaning displacement, the application of a more appealing term to something not really new?

The term artificial intelligence was invented by John MacCarthy in 1956, in a seminar that took place at Dartmouth College in Hanover, U.S.A. At that time, the participants made exaggerated predictions about the imminent advances they expected in this field, some of which have not been fully performed 65 years later. Now the predictions are even more ambitious, but it is quite likely that none will come true in the short term, and that some will never come true.

The current applications of artificial intelligence, to which the media give a lot of publicity, are based on algorithms designed more than three decades ago, which are now applied on a large amount of data, taking advantage of the greater speed of computers and the existence of the Internet, which provides the data. But these algorithms are not really intelligent. Essentially, they can be divided into four large groups:

1. Data analysis. Decision-making algorithms are applied to huge amounts of data to predict specific future situations. In this case, of course, the intelligence is not in the algorithms, but in the programmers who designed them. Therefore, these applications should not be considered artificial intelligence. This is an example.
2. Machine learning. I discussed this in detail in another post in this blog.
3. Generation of natural language. It consists in creating sentences or paragraphs that sound natural adapted to a specific topic, or to the way someone speaks. The media usually spread news related to this issue, because they believe that they are huge advances. It was recently announced that a dead person has been resurrected, which is obviously false. What has been done is to apply a chat-bot-generating program, so that it uses recorded conversations of the dead person to fabricate texts that appear to come from that person. This is just an updated version of Eliza, which almost fifty years ago emulated the way a psychiatrist speaks and deceived some people. These programs are not really intelligent, for there is not a mind behind them. The intelligence is, as usual, in the program designer. I mentioned this type of applications in another post in this blog.
4. Image analysis. These applications are usually described as if they were strong artificial intelligence, when they are nothing more than developments of classic algorithms that allow the program to discover if the person who interacts with it is sad or happy, and similar elementary things. The program can use this information because it has been programmed, but of course there is no conscient being behind.

There is also much talk in the media about the imminent building of quantum computers that will implement strong artificial intelligence, or of computers based on components that function like neurons in the human brain. The former is impossible, as I explained in this post. The second, for the moment, is just hype. An example is this paragraph that appears in a recent article published in a Spanish web page:

We have designed intelligent systems that think like humans (artificial neural networks), that act like us (robotics), that reason like people (expert systems), and that even behave rationally (intelligent agents).

This short sentence is full of mistakes. In fact, a) artificial neural networks don't think like human beings. They don’t think at all, as nobody is thinking there. b) Robots act like us (or even better than us) in specific activities, but they are far from being able to replace a human being, except in automatable tasks. c) Expert systems have never reasoned like people: this is why research in this field has almost stopped. d) "Intelligent" agents (in quotes, as it is a euphemism) do not behave rationally. They behave just as they have been programmed. The only one who acts rationally is their programmer.

Isaac Asimov

In a recent interview I stated that many things being predicted now are not science, but science fiction. The interviewer replied that many predictions that at first were just science fiction have come true. Although this has happened in some cases, other science fiction predictions will never come to pass, so this assertion cannot be generalized. For instance:

•         Travel at speeds greater than that of light: This is a typical theme of science fiction, to make it possible to travel to the stars with the same ease with which we travel today on Earth, but this is highly unlikely in the current state of science. In previous posts I have talked about this.

•         Time travel: As I explained in other posts, the modified Fermi paradox shows that time travel is not just impossible at this time, but it will always be.

•         Reduction of human beings to microscopic size: The paradigm of this subgenre is Fantastic Voyage, a novel by Isaac Asimov based on the script of the film of the same title. I will talk about this in a future post.

Can this happen regarding artificial superintelligences, whose imminent implementation is continually announced by the media? Perhaps we will never make them. This is what many artificial intelligence experts think.

The same post in Spanish

Thematic Thread on Natural and Artificial Intelligence: Previous Next

Manuel Alfonseca