Physicists Test Formula to Predict AI Chatbot Answer Breakdowns
George Washington University researchers say a formula could estimate when chatbot responses deteriorate, with early support from tests on small models.
Physicists at George Washington University have proposed a formula to estimate when an AI chatbot’s responses will shift from good to poor, according to Decrypt.
Initial experiments using small models supported the approach, offering preliminary evidence that such changes in response quality may be predictable.
The findings remain early-stage. The reported tests provide support at a small scale, but do not establish whether the formula works for larger models.
Source: Decrypt
