Reacción a "An AI system could win a medal at the International Mathematical Olympiad, according to a study"

It's very rare for an article in Nature not to meet the highest quality standards. If it has passed the rigorous peer-review process (by people more specialized in the subject than I am), it's almost certain that the quality of the work is top-notch.

AI is increasingly occupying and surpassing areas of human knowledge. This is beautifully illustrated by researcher Hans Moravec, as Max Tegmark explains in his book Life 3.0 on AI: a panorama of human skills can be visualized as a landscape with its plains and mountains. The plains or valleys would represent the human skills that are easier for a computer, such as "arithmetic" and "memorization," the slopes like "theorem proving" and "chess," and the high mountain peaks with names like "locomotion," "eye-hand coordination," and "social interaction." The advance of computer performance is like the slow rise of floodwaters. Half a century ago, they began to inundate the lowlands, displacing human calculators and record-keepers, but allowing most of us to remain dry. Now the waters have reached the slopes, and the frontier posts we have there are considering withdrawal. We feel safe on our peaks, but at the current rate, these too will eventually disappear. "Submerged within another half-century," and at that point, AI will have surpassed all human cognitive abilities.

[Regarding possible limitations] Almost the opposite: in the case of the article mentioned, the Mathematical Olympiads focus on a very specific type of problem that requires relatively little mathematical background, but extensive use of that knowledge and a great deal of ingenuity. I suppose that acquiring a greater amount of mathematical knowledge must be relatively easy for a machine, and what has always been considered most human—ingenuity—is what this work has demonstrated: that an AI can achieve it.