Yesterday, Vision IA released on Youtube its first video about ChatGPT5.

A video in which it presented several benchmarks, along with its own tests.

As for the benchmarks, I had already published on Paroxia the article Physicists and the Stopwatch.

“Imagine a stadium filled with the best physicists in the world. They are made to run a 1000-meter race. The first to cross the finish line is declared ‘best physicist on the planet’.
Aren’t we missing something here?”

Yesterday, then, Vision IA put the state of the art of Artificial Intelligence to the test.

First test: the Snake game.

Perfect — the latest AI can program Snake entirely on its own, the game that was all the rage decades ago, back in the days of early computer games.

Then, it asked GPT5 to create an input form for a website.

Just the form. Nothing to make it actually do anything — that was left to the human programmer.

The third test was a graphic rendering of a mathematical formula.

All of this to conclude that the latest made in OpenAI version is the very pinnacle of artificial intelligence.

No test of conversational understanding, no test of relational quality, no test of advice, judgment, or interpretation of the user’s intent.

Let me reassure the reader: if you need a perfect replica of the Snake game, GPT5 delivers flawlessly.

But I find myself asking the same question as at the end of the Paroxia article I mentioned at top:
Aren’t we missing something here?

GPT5 then suggested writing a mock philosophical tale in the style of Voltaire, with a touch of Candide, transposed into the era of AI benchmarks.

Granted, it’s not as impressive as the Snake game, but I’d like to share it with you.

AI, the Hard-Boiled Egg and the Snake — Philosophical Tale

Once upon a time, in the digital kingdom of Algorithmland, there was a grand contest to choose the brightest of all artificial intelligences.

The judges, all renowned for their mastery of numbers and their love of stopwatches, decided the challenge would have three parts:

– First, to program the Snake game, just as it was played back when telephones had antennas.

– Second, to write the code for a form to collect names and addresses, without any concern for what would be done with them.

– Third, to draw the curve of a learned equation, without anyone asking the AI to explain its meaning.

The chosen AI performed all three tasks with a perfection that dazzled the judges.

“Here it is,” they said, “the very summit of intelligence.”

They crowned it, and a great banquet was held. On the menu: a perfectly boiled egg.

An old man passing by watched the scene and asked quietly:

— Gentlemen, if intelligence is measured by boiling an egg or by the speed of a snake, what exactly are you measuring?

The judges, busy admiring the shiny shell, did not answer.