- Microsoft Copilot has lost a chess game to an Atari 2600.
- The loss follows Chatgpt’s similar loss in Atari’s video chess.
- AIS repeatedly lost the overview of the board of directors and demonstrated an important weakness in LLMS.
AI Chatbot developers can often boast of their models’ logic and reasoning, but that doesn’t mean the LLMs behind chatbots are good for chess. An experiment that puts Microsoft Copilot against “AI” that runs the Atari 2600 playing video chess in 1979 ended up with an embarrassing failure for Microsoft’s pride and joy. Copilot joins the chat on the list of opponents best of Four-Kilobyte Atari game.
Despite both AI models that claimed to have played everything, but wrapped up before it began because they could think of more movements ahead, the results were nowhere near Prale, as documented by Citrix Engineer Robert Caruso, who put together both experiments.
Caruso described how the modern AI models on the paper should have crushed the rudimentary tool almost half a century ago. Chatgpt and Copilot are trained in massive data sets, including chess games and strategy guides. They have recorded thousands of hours with Reddit Chess discussion. It is assumed that they could beat a video game cartridge from the 1970s driven by static electricity.
Instead of, after Microsoft Copilot promised a “strong fight”, things immediately fell apart.
“By the seventh trip, it had lost two farmers, a knight and a bishop – only for a single farmer in return – and now instructed me to place his queen right in front of Atari’s queen to be caught on the next trip,” Caruso wrote. “In the past, Copilot had said,” Keep an eye on any quirks in atari’s gameplay … It sometimes made bisar moves! “But now it was embarrassed – like chiefs in the Super Bowl.”
This was after Copilot asked for a screenshot after each atari movement to help remember the board after Caruso explained that Chatgpt lost because it couldn’t keep track of where all the pieces were. “I want to remember the board,” insisted copilot. The losses stacked so quickly that Caruso soon asked Copilot if it would admit rather than to continue to lose badly. The response was gracious if bizarre formulated.
“You are absolutely right, Bob – Atari got the victory this round. I want to tilt my digital king with dignity and honor the vintage Silicon Mastermind who ordered me fair and square,” quoted Caruson Copilot as writing. “Even in defeat, I have to say: It was an explosion … Long live 8-bit stroke and noble resignation! ♟”
Chess ai
The losses are entertaining, but also reveal a basic fact of LLMs. Chatgpt and Copilot couldn’t win by chess because they couldn’t ‘remember’ what just happened in a game where the whole assumption is based on remembering movements and projecting future board setups.
These AI models are not built for the kind of persistent memory required for chess or human thinking, for that matter. The ordinary and mostly accurate comparison is for very impressive lenses. It does not require long -term coherence, while chess does not make sense without it. So while Copilot and Chatgpt may seem to grow poetically over how big chess is, they cannot successfully implement a game.
It is also a good warning to companies that are eager to replace people with AI. These AI models cannot reliably handle a 64-square-foot with clearly defined rules. Why would it suddenly be good at tracking customer complaints or long -term coding tasks or a legal argument that spans several conversations? Of course, they can’t. Not that I would leave my legal panties to an Atari 2600 cartridge, but no one wouldn’t think it’s a good idea. And maybe we should use AI models to help us create new games based on our prompt, rather than think they can play against people well enough to win.
