The giants of AI such as OpenAi, Meta and Google dream of reaching artificial superintelligence. The models of the future may overcome the abilities of the human brain, although that moment is still somewhat far.
The AI still has unpredictable consequences as has happened with Gemini. The chatbot I had to play a chess game against an Atari 2600 console, but artificial intelligence has refused.
Chatgpt and co -pilot lose against Atari 2600
Artificial intelligence has decided to ignore the requests of a user and not play the game against Atari 2600, a console manufactured in 1977. The device had previously defeated Chatgpt with a new chess algorithm.
Engineer Robert Caruso tested ChatgPT last June, but Sam Altman’s chatbot could not read the board. The emulated version of Atari 2600 ended up imposing with a technology of almost half a century.
«Chatgpt was completely shattered at the beginner level,» in statements Robert Caruso in his LinkedIn account. «I wanted to discover how quickly I could win a game that only thinks one or two movements ahead on a 1.19 MHz CPU,» he added.
Chatgpt faced numerous problems, the first of them read the Atari icons used for chess pieces. The chatbot ended up losing the chips, so Caruso had to resort to the standard chess notation.
The humble 8 -bit engine of the Atari 2600 does not have a language or flash model. The engineer had to dedicate 90 minutes to prevent AI from making mistakes and correcting the way you read the board. Chatgn promised that he would improve in a second attempt, but did not yield.
Copilot does not get rid of mistakes, Gemini retires
Robert Caruso conducted the same experiment with co -pilot before facing Atari against Gemini. The result has been very similar: «Like Chatgpt, Copilot exceeded confidence. He claimed that he could anticipate between 10 and 15 movements, but decided to limit himself between 3 and 5,» clarifies the AI expert.
Microsoft’s artificial intelligence discovered that Chatgpt had had difficulties during the game, but Co -ylot was still convinced that he could overcome. The reality was very different with a very low performance.
«On the seventh turn, he had lost two pawns, a horse and a bishop (in exchange for a single pawn) and now he was telling me that his queen was just in front of the queen of Atari to capture her in the next turn,» says Caruso.
Gemini AI was safe shortly before the game, he defined himself as a «modern chess engine capable of anticipating millions of movements and evaluating infinite positions.» The expectations were slowly dissipating.
To Ia by Google He has admitted that he has hallucinated with respect to his skills to compete with the console. Gemini has clarified that «I would have a lot of difficulties against the Atari 2600 chess game engine,» they think that canceling the game would be «the most sensible decision.»
Know How we work in NoticiasVE.
Tags: Artificial intelligence