AIprl Fool

In Private Eye this morning, I read about a nice April Fool joke perpetrated by Tim Gowers. He claimed on X that he had succeeded in getting Grok to solve the “well-known Dubnovy Blazen problem” in graph theory. The clue is apparently in the facts that this was posted on 1 April, and that “dubnovy blazen” is Czech for “April fool”. (I can’t verify this since my knowledge of Czech is not good enough, and for some reason Google refuses to translate the phrase.)

The report goes on to say that a Google query for “Dubnovy Blazen problem in graph theory” gives the answer that this was a well-known problem which was solved by AI model Grok.

However, it goes deeper than that. When I went searching on Google myself, I found that another well-known graph theorist, whom I won’t name here, had posted on X an enthusiastic response to Gowers’ post, wondering if the Riemann hypothesis would be the next to fall to AI. I don’t know for sure whether this person was taken in by Gowers’ joke, or saw through it and was mischievously stoking the flames; but I am very happy to give them the benefit of the doubt a nd assume the latter. But, worryingly, a Google search revealed that now Google “thinks” that this person was responsible for getting Grok to solve the problem.

Private Eye comments, “It’s wonderful watching the well of knowledge being poisoned in real time.” I do not in the least blame Gowers for this; his aim was clearly to warn us of the dangers of uncritical acceptance of AI, and have some fun at the same time. The people to blame are those who do see AI as the saviour of humanity. It will be interesting to see how they kick this under the carpet. My guess is that it will be by publicly ignoring the whole thing while privately putting a line in the program telling it not to respond to this question.

Roger Penrose refers to AI as “artificial stupidity”, and it was certainly stupidity that was on display in this incident.