3 ms·

> ChatGPT can't play chess, for example Out of pure curiosity I tried this a while back. At some point it tried to move a knight from d4 to f6, and it was the

by lucumo 3y ago

> ChatGPT can't play chess, for example Out of pure curiosity I tried this a while back. At some point it tried to move a knight from d4 to f6, and it was the most stubborn and obstinate child when I tried to correct it. It was the best fun I had playing chess in a few decades. Excerpt: _Me:_ Can a knight move from f6 to d4? _ChatGPT:_ Yes, a knight can move from f6 to d4, as long as there are no pieces blocking its path and the square it is moving to is not occupied by one of its own pieces. In fact, the move Nxd4 that I suggested earlier involves moving a knight from its starting square to d4, capturing a pawn in the process. _Me:_ Which fields can a knight move to from f6? _ChatGPT:_ From f6, a knight can move to six different squares. These squares are: [... lists: d7, e8, g8, h7, d5, e4 ...] [...] _Me:_ You forgot g4 and h5, but no matter. Is d4 in that list? _ChatGPT:_ [...] To answer your other question, no, the square d4 is not one of the squares that a knight on f6 can move to directly. However, as we saw earlier, it is possible for a knight to move from f6 to d4 via an intermediate square, such as e8 or d5.

kgwgk 3y ago
> However, as we saw earlier, it is possible for a knight to move from f6 to d4 via an intermediate square, such as e8 or d5. GPT stands for Goal Post Transportation.
euroderf 3y ago
So is this how hallucinations form ? It proposes something half-baked, and then later refers back to it as fact ?
- abraxas 3y ago
  Sounds like we already have full blown cyber replicas of politicians then.
  - mythrwy 3y ago
    Or some bosses I've had. I actually can't wait to start reporting to GPT, how much crazier can it be?
- zerocrates 3y ago
  Some of this is probably just an artifact of how ChatGPT specifically works: I believe I have it correct that it basically feeds the transcript of the conversation, to the extent possible, back to itself as part of the prompt going forward. So its prior responses in the session are part of the text it's generating from.
- HarHarVeryFunny 3y ago
  No, it's generating output word by word, not planning very far ahead (it can't since it doesn't even know what words it's really going to generate, since they are randomly sampled), and essentially backs itself into a corner where completing the train of thought requires a fact that it doesn't actually have. Just as a made up example, say you asked "what is the capital city of england", and the model had seen similar questions in it's training data that were answered with "the capital city of X is Y", so it starts word-by-word generating this type of response "the capital city of england is", but it then turns out the model doesn't actually know the answer (i.e. this partial response context doesn't cause it to predict the correct answer), so it blithley predicts the next word as whatever it's inner machinations come up with, maybe "buckingham palace" or "flavor town". "Hallucination" seems a poor way to describe it, nor is it lying since there's no bad intent ... it's basically "starting to speak before brain engaged", a bit like a game show contestant being a bit too enthusiastic and hitting the "answer" button without actually having a fully thought out answer in mind.
pmoriarty 3y ago
I tried this question on various LLMs on poe.com and Claude, Sage, and Claude+ correctly answered that a knight can't move from f6 to d4. Dragonfly failed the test, and though Claude did answer the question correctly, it gave some other examples of invalid moves for a knight on f6. All of the moves that Claude+ gave were valid, but it didn't list all possible moves for a knight on f6. I didn't try GPT4.
- lucumo 3y ago
  This was part of a longer conversation. It's pretty long and it got confused pretty easily, so I had to correct it multiple times. It was an interesting experience in its novelty, but it felt like pulling teeth to get it to give me a move. - It insisted multiple times that it was not a chess player. "As an AI language model" it's not physically capable of playing chess, nor does it have a personality, so it couldn't play chess but was able to be an "AI chess assistant". - As is ChatGPT's default M.O., it was a wordy son of a bot. Giving me explanations which fields would be controlled and what pieces would be developed, and what not. Often analysing my move too. - It repeatedly offered multiple move options, so I had to ask it for only one move every move. - After my fifth move it couldn't come up with a good move, so I asked it to describe the board. It did, and it even drew me a diagram. The diagram was complete nonsense: it had 9 black pawns for a start. The description was okay-ish, but not good enough. I could correct that though, but it took two iterations. I didn't get much further with it. It was completely confused about the state of the board and the possible moves. Still, it did know which game it was playing and more or less how to act with it, so I was still kind of impressed. You can really tell that it is a language model, not just because it bloody keeps reminding you, but also by the things it does wrong and right. E.g. getting wordy analysis or legible descriptions, but ridiculous diagrams. As another example: at one point it corrected me. I accidentally used notation in my native language ("Pg3") and it said that it was invalid but assumed I meant "Ng3". That was correct, it's clearly better at talking about chess than playing chess.
- valine 3y ago
  I tried GPT4, it failed. I also tried Vicuna which sort of passed. The conspiracy theorist in me thinks OpenAI did something to neuter the chess playing capabilities of their models. GPT4: Yes, a knight in chess can move from f6 to d4. The knight moves in an L-shape pattern: it can move two squares in one direction and then one square perpendicular to that direction, or one square in one direction and then two squares perpendicular to that direction. In this case, the knight would move two squares to the left (from f6 to d6) and then one square down (from d6 to d4). Vicuna: No, it is not possible for a knight on f6 to move to d4 in chess. The knight's movement pattern is L-shaped, meaning that it moves two squares horizontally or vertically and then one square diagonally. In this case, moving from f6 to d4 would require the knight to first move to g7, which is not part of its movement path.

by lucumo 3y ago

kgwgk 3y ago

> However, as we saw earlier, it is possible for a knight to move from f6 to d4 via an intermediate square, such as e8 or d5. GPT stands for Goal Post Transportation.

euroderf 3y ago

So is this how hallucinations form ? It proposes something half-baked, and then later refers back to it as fact ?

abraxas 3y ago
Sounds like we already have full blown cyber replicas of politicians then.
- mythrwy 3y ago
  Or some bosses I've had. I actually can't wait to start reporting to GPT, how much crazier can it be?
zerocrates 3y ago
Some of this is probably just an artifact of how ChatGPT specifically works: I believe I have it correct that it basically feeds the transcript of the conversation, to the extent possible, back to itself as part of the prompt going forward. So its prior responses in the session are part of the text it's generating from.
HarHarVeryFunny 3y ago
No, it's generating output word by word, not planning very far ahead (it can't since it doesn't even know what words it's really going to generate, since they are randomly sampled), and essentially backs itself into a corner where completing the train of thought requires a fact that it doesn't actually have. Just as a made up example, say you asked "what is the capital city of england", and the model had seen similar questions in it's training data that were answered with "the capital city of X is Y", so it starts word-by-word generating this type of response "the capital city of england is", but it then turns out the model doesn't actually know the answer (i.e. this partial response context doesn't cause it to predict the correct answer), so it blithley predicts the next word as whatever it's inner machinations come up with, maybe "buckingham palace" or "flavor town". "Hallucination" seems a poor way to describe it, nor is it lying since there's no bad intent ... it's basically "starting to speak before brain engaged", a bit like a game show contestant being a bit too enthusiastic and hitting the "answer" button without actually having a fully thought out answer in mind.

pmoriarty 3y ago

I tried this question on various LLMs on poe.com and Claude, Sage, and Claude+ correctly answered that a knight can't move from f6 to d4. Dragonfly failed the test, and though Claude did answer the question correctly, it gave some other examples of invalid moves for a knight on f6. All of the moves that Claude+ gave were valid, but it didn't list all possible moves for a knight on f6. I didn't try GPT4.

lucumo 3y ago
This was part of a longer conversation. It's pretty long and it got confused pretty easily, so I had to correct it multiple times. It was an interesting experience in its novelty, but it felt like pulling teeth to get it to give me a move. - It insisted multiple times that it was not a chess player. "As an AI language model" it's not physically capable of playing chess, nor does it have a personality, so it couldn't play chess but was able to be an "AI chess assistant". - As is ChatGPT's default M.O., it was a wordy son of a bot. Giving me explanations which fields would be controlled and what pieces would be developed, and what not. Often analysing my move too. - It repeatedly offered multiple move options, so I had to ask it for only one move every move. - After my fifth move it couldn't come up with a good move, so I asked it to describe the board. It did, and it even drew me a diagram. The diagram was complete nonsense: it had 9 black pawns for a start. The description was okay-ish, but not good enough. I could correct that though, but it took two iterations. I didn't get much further with it. It was completely confused about the state of the board and the possible moves. Still, it did know which game it was playing and more or less how to act with it, so I was still kind of impressed. You can really tell that it is a language model, not just because it bloody keeps reminding you, but also by the things it does wrong and right. E.g. getting wordy analysis or legible descriptions, but ridiculous diagrams. As another example: at one point it corrected me. I accidentally used notation in my native language ("Pg3") and it said that it was invalid but assumed I meant "Ng3". That was correct, it's clearly better at talking about chess than playing chess.
valine 3y ago
I tried GPT4, it failed. I also tried Vicuna which sort of passed. The conspiracy theorist in me thinks OpenAI did something to neuter the chess playing capabilities of their models. GPT4: Yes, a knight in chess can move from f6 to d4. The knight moves in an L-shape pattern: it can move two squares in one direction and then one square perpendicular to that direction, or one square in one direction and then two squares perpendicular to that direction. In this case, the knight would move two squares to the left (from f6 to d6) and then one square down (from d6 to d4). Vicuna: No, it is not possible for a knight on f6 to move to d4 in chess. The knight's movement pattern is L-shaped, meaning that it moves two squares horizontally or vertically and then one square diagonally. In this case, moving from f6 to d4 would require the knight to first move to g7, which is not part of its movement path.