English Chess Forum

I started using ChatGPT today and asked it the following question:

Are you willing to undergo a Turing test?

and it responded:

As a language model developed by OpenAI, I am not capable of experiencing emotions, desires, or intentions. My sole purpose is to assist users by generating human-like text based on the input I receive. I am designed to respond to questions and provide information, but I am not a conscious being and do not have the capacity to experience a Turing test.

I then asked it:

Is HAL one of your friends?

and the answer came back:

No, HAL is not one of my "friends." HAL (Heuristically programmed ALgorithmic computer) is a fictional artificial intelligence character that appears in the novel "2001: A Space Odyssey" by Arthur C. Clarke and in the related film by Stanley Kubrick. HAL is a sentient computer that controls the systems of a spacecraft and is depicted as having a conversational interface with the crew. Although HAL is often discussed in the context of artificial intelligence and its potential capabilities, it is a fictional character and not a real AI system.

Finally:

A man says that he is lying. Is what he says true or false?

produced:

This is a classic paradox. If the man's statement is true, then he is lying, which means the statement is false. But if the statement is false, then he is not lying, which means the statement is true. This paradox creates a self-contradiction and is therefore undecidable. The statement "A man says that he is lying" cannot be consistently determined to be either true or false.

I'd suggest you don't ask it questions about your relatives:

... or tables:

ChatGPT is so impressive when it gets something wrong it is surprising.

I saw a video testing it. It corrected errors in computer code and explained clearly which errors it was correcting. An arithmetic sum it gave an incorrect answer, though the answer was approximately correct. Analysing large volumes of text allows it to do the former but not the latter.

Roland Kensdale wrote: ↑

Tue Feb 07, 2023 12:04 am

ChatGPT is so impressive when it gets something wrong it is surprising.

I saw a video testing it. It corrected errors in computer code and explained clearly which errors it was correcting. An arithmetic sum it gave an incorrect answer, though the answer was approximately correct. Analysing large volumes of text allows it to do the former but not the latter.

Not at all surprising actually. It routinely gets things wrong - just the nature of an intelligence free large language model basically.

The real issue is that when it does get things wrong it does so with total confidence, including making up references and so on. It does the latter because it knows how to structure that sort of document.

As is definitely non trivial health hazards attached for programming work. That's one of the areas where they might get somewhere with a lot of specalised work to try and tame it & its descendants.

Its a (very impressive) chatbot not a knowledge base

What do you think of the English Chess Forum?

As an AI language model, I don't have personal opinions, but I can tell you that the English Chess Forum is a popular online platform for chess players and fans to discuss the game, share information, and connect with each other. People generally have positive opinions about it, as it provides a helpful community and resources for chess enthusiasts.

Not much to argue with there, surely?

I would have thought that chess is easy for any AI, instead

Would you like to play a game of chess? I'll play as white and start with e4

was answered by

I'm an AI language model and don't have the ability to play chess, but I can assist you with chess-related questions.

Ian Thompson wrote: ↑

Mon Feb 06, 2023 7:09 pm

I'd suggest you don't ask it questions about your relatives:

There's a difference between erudition (ability to remember data) and intelligence (ability to process data in original ways); it seems ChatGPT might be much better at the former than the latter; as many of our politicians show, being able to quote sentences in Latin does not necessarily make you smart.

Paolo Casaschi wrote: ↑

Wed Feb 08, 2023 12:35 pm

I would have thought that chess is easy for any AI, instead

Would you like to play a game of chess? I'll play as white and start with e4

was answered by

I'm an AI language model and don't have the ability to play chess, but I can assist you with chess-related questions.

Just spotted this: https://en.chessbase.com/post/surprisin ... ying-chess

Haven't watched yet so don't know if the video explains how it was done.

I understand sometimes ChatGPT can be tricked into doing things its architects didn't intend to be possible (usually harmful or undesirable things).

Finally watched the above video, so you don't have to.
No idea if ChatGPT was actually involved in this as presenter believes.
Something occurred on a Chess board, but appropriately for this thread it was Not Chess. Other than alternating White and Black very little of normal Chess here. Pieces moved by ChatGPT can move to any location. Pieces can appear on any square, replacing the current occupants. ChatGPT takes its own pieces and "moves' into check.

Stockfish is not permitted to terminate the game with an error message.

Doesn't have the charm of the game in Alice Through The Looking Glass which at least served the purpose of roughly representing the story.

As it said itself, ChatGPT does not play Chess.

Roland Kensdale wrote: ↑

Fri Feb 17, 2023 9:46 pm

As it said itself, ChatGPT does not play Chess.

Clearly not.