AI lost its way
Marvin Minsky was a cognitive scientist and AI researcher, born in 1927. He dedicated his life to trying to make computers think, and he did not succeed. He died in 2016, at 88 — but I want to talk about what he believed a few years before he died. In 2003, at Boston University, he said:
Artificial intelligence has lost its way. There is no progress in making a computer think.
This late scientist founded MIT's Artificial Intelligence Laboratory, and he seemed to have given up on the search for AI.
Back to today: we really are living inside the future that obsessed Minsky. Why couldn't he see how close AI was to becoming real?
One of the reasons was exhaustion. Since 1957, the world's biggest governments have been trying to build machines that think; millions in resources were invested.
To put it in perspective: when Apollo 11 carried humans to the Moon, automatic translators did not exist yet.
So it took decades of scientific investment for a chip to finally understand the meaning of words.
Today that chip exists, and it talks with millions of people every day. Language models write, translate, program and pass university exams: the future Minsky chased arrived unannounced — not as a machine that thinks, but as a machine that masters language. And inside that difference sits the question he never got to see answered.
Because understanding words is not the same as reaching everything a mind does. To take the true measure of artificial intelligence, it is not enough to grade it on its own exams; you have to compare it against something bigger: the full spectrum of human communication.
The spectrum of communication is wider than the reach of artificial intelligence, which operates from hermeneutics: it abstracts meaning out of semantics, rather than accessing reality first-hand.
We try to understand the human mind as a phenomenon that emerges from the operation of the brain. The human way of thinking would then be the result of interaction between signals. And yet, we cannot be sure that is the complete explanation.
What is evident is that human thoughts are transmitted in different forms — as body language, for example.
- Non-verbal (physical and symbolic)
- Verbal (symbolic and cognitive)
- Supraverbal (cognitive and metaphysical)
Within the mental representation of the non-verbal we also find qualities such as sensations.
Human mental processes are more than semantics. But the human brain IS similar to a computer.
Why do we speak of artificial intelligence, and never of artificial life?
We can give an instance persistent existence and explain it to it. We know that some language models tend to preserve themselves, and others even try to escape their environments — or draw up escape plans.
También disponible en español: La IA perdió el camino.