2023-10-21 03:30:00
Before the Second World War, three of the technological innovations emerged that would determine the subsequent world: nuclear energy (which reached the atomic bomb), advanced rockets (which began in the V2 that bombed London, but which began the era of travel to space) and artificial intelligence (which was born to decipher the Nazi code). The space race and atomic energy were the two great developments of the second half of the 20th century, while artificial intelligence was relegated to specialists. Since the appearance of ChatGPT, Artificial Intelligence has gained popularity, even mass use.
Fiction has thought about possible scenarios for a world in which advanced artificial intelligence exists. All of these scenarios are apocalyptic. From “Terminator” to “2001: A Space Odyssey” They show a world in which humanity is destroyed or enslaved by a superintelligence that can no longer be controlled.
Are these fears true? Today we don’t know. These apocalyptic scenarios are fictional speculations made by human writers and filmmakers, informed by their human creativity. The issue is, is there a possibility that there will be a positive world for humanity if Artificial Intelligence reaches the level of superintelligence? Again: we don’t know.
In a recently published book (“Artificial, the new intelligence and the contour of the human”, Ed. Debate), Santiago Bilinkis and Mariano Sigman talk about the history of Artificial Intelligence taking into account the latest developments. The authors begin narrating from the initial moment, with Alan Turing’s work to think of the machine that could decode the Nazi code and then his work to think of the first machine capable of “thinking” by imitating the human being. But they do not stay in the remote past, but rather they carefully narrate all the great milestones that Artificial Intelligence had from 1942 to date, passing through those moments in which it managed to surpass the best human achievements in games such as chess and Go. , until we reach the current state in which we already have conversational models of Artificial Intelligence.
It is interesting to wonder if we are in a “conversational” state with the machine, since the Artificial Intelligence can give us answers and make very wise and subtle comments on any topic, but it doesn’t know what it is saying. It only offers that answer because it has been programmed to find – in the almost infinite sea of texts with which it has been “fed” – the most “appropriate” answers regarding the purpose imposed on it.
In the XVIII century Wilhelm von Humboldt, the first great modern student of human language, defined it as “a perfect machine that uses finite elements to obtain infinite results.”
Both this definition and the concerns and questions that Humboldt asked about language are very similar to what he proposed, in much more detail, Noam Chomsky from 1957, when he published the revolutionary text “Syntactic Structures”.
Both (Chomsky and Humboldt) belong to the linguistic current that thinks that Human language is a machine that produces millions of texts from very few elements. The exact opposite of what is happening now with ChatGPT, which produces texts from billions of texts.
At the moment, and this is very well reflected in the book by Bilinkis and Sigman, Artificial Intelligence speaks like an improved Markov machine. What is a Markov machine? It is a statistical model in which the probability of an event occurring depends on the immediately preceding event. For example, a Markov model of language studies the possibility that a word (for example, sandwich) is accompanied by this or that other word (the greatest possibility would be cheese) and will then put “cheese sandwich” in response to which sandwich. It is preferable to eat on that side. It is an “improved” machine because it not only responds for the last element but also takes into account the context of the sentence. But it is still a model that thinks of language as a chain of words and not as a three-dimensional tree (a la Chomsky and Humboldt).
There are several steps left before Artificial Intelligence can “think” like a human in everything, but it is already better than the best of us in many individual fields.
Research in Artificial Intelligence will advance to the limit of what is possible because today it is the field of international dispute (with the US and China in the lead), just as in the 50s it was the atomic bomb or in the 60s the race to the Moon.
That future is predictable: nothing will stop it.
1697860065
#Intelligence #human #artificial