The AI powering ChatGPT might present fully completely different solutions to the identical mathematical issues over time. Those findings from latest experiments have fuelled an ongoing debate about whether or not the AI chatbot’s efficiency is getting worse – and have spurred the agency behind it, OpenAI, to reassure clients that functions constructed on ChatGPT is not going to frequently break.
“The takeaway message is that the behaviour of the ‘same’ large language model can change substantially,” says Lingjiao Chen at Stanford University in California. “It is essential …
Source: www.newscientist.com