Unhappy with the humiliating ChatGpt in the hands of the Atari 2600 emulator video chess, Robert Caruso attempted again this time with Microsoft's co-pilot.
In theory, the results are the same, and the co-pilot would take similar driving. Still… what happens if Copilot wins where ChatGpt is not possible? “There's no reason to think about that,” Caruso writes, but “Imagine if Microsoft products surpass ChatGPT, everyone will explode their heads.”
So Caruso launched Stella Emulator and had a pre-match chat with Copilot to explain what stumbled over ChatGpt. He told the chatbot that one of the main reasons ChatGpt was lost was that it was unable to track the board. If the co-pilot suffers from the same challenges, it is rarely plagued to play.
With confidence that only AI chatbots could gather, Copilot claimed that it was hilariously good, not only could it play chess. Caruso said, “he argued that he could think of 10-15 moves, but “I thought he would stick to a 3-5 move against 2600 because it would be a “suboptimal movement” that could make “suboptimal movements” rather than sticking to deep calculations.”
Humans return to the Go-Playing AI system
read more
And do you want to track the board? Copilot said, “We're reminiscing about previous moves and making strong efforts to maintain gameplay continuity, so the match should be smoother.”
Copilot admitted that it has the same spatial memory gap as ChatGpt, but stated that it can analyze the current board and choose good movements. Caruso needs to give the chatbot a screenshot of the board after the bite movement and feed capillot moves manually to video chess.
The game was in progress!
Now anyone with experience with today's generative AI systems will know what happened. Copilot's Hubris has been left behind. The move was…interesting, and although we managed to lose two pawns, knights and bishops, the powerful Atari 2600 video chess was just one pawn. Eventually, Caruso asked Copilot to compare what the board looked like in the last screenshot the board had pasted, and the chatbot admitted they were different.
“Chatgptdéjàvu.”
There was no way Microsoft's chatbot could win with this handicap. Still, it was elegant in defeat. [to the] The mastermind of vintage silicon made me fair and square. ”
Caruso's experiments are interesting, but it also emphasizes the absolute confidence that AI can vent nonsense. Copilot (like ChatGpt) could have been trained in the basics of chess, but could not create a strategy. The problem was exacerbated by the fact that the position of the chessboard was understood to appear to be significantly different from reality.
The morality of the story is to be aware of the confidence of the chatbot. LLM is clearly good at a few things. The 45-year-old chess game is clearly not one of them. ®
