The scene seems to have come out of a science fiction story on the contrary: on the one hand Chatgptamong the most advanced artificial intelligence in the world; on the other, an old man Atari 2600glorious console launched backwards 1977. In the middle of? A chess game. The result? A sensational defeat for the IA.
To tell everything is Robert CarusoCitrix software engineer, who published the episode on LinkedIn. After a chat on the history of artificial intelligence in the game of chess, Chatgpt took volunteer to try their hand at a challenge against Atari Chessgame of 1979. He thought he won easily. And instead …
Recognize Alfieri and Torri? For Chatgpt it was a company
Although a Basic scoreboard to learn to playChatgpt immediately started showing the first signs of confusion. Ha Exchanged Alfieri for Torrimade wrong moves and, worse still, has constantly forgotten where their pieces were found.
Caruso tried to meet him: he activated the standard notationthe one used by professional players. But even there, nothing to do. The IA continued to commit coarse errorsrepeating several times that “he would have won if only he could start again”. As if it were enough to press “reset” to change the course of a defeat …
And in all this, Atari simply did his job. No natural language, no ia, no neural networks. Alone pure brute-force logicalcapable of looking at one or two moves forward. It was enough to ridicule one of the most celebrated artificial intelligence of our time.
Behind the calculation power, the IA still stumbles in simple obstacles
Caruso did not spare himself in detail: The game lasted about an hour and a halfduring which Chatgpt did quite messes from “being hunted by a third grade chess club”, his words. But what left the engineer most perplexed was the way in which the chatbot referred to the gamespeaking to the plural: “We can win, if we start again“.
Unsettling? Perhaps. Fun? Surely. But also detector: For what advanced, artificial intelligence and attention to details necessary in activities that require visual precision and spatial coherence, such as chess.
In the end, as Caruso writes with a certain taste for irony: “The humble 8-bit engine of the ARTARI has simply done its duty. No special effects. Only the stubbornness of 1977“.
A useful reminder: perhaps, before fear of an I who will replace us, we should check if he knows at least Play chess properly.
