Nicolas MØUART
An interesting #transparency problem with LLMs, which are anything but scientific in practice. It's not clear from where the data is collected for most, if not for any #LLMs. In a case like surface gravity, the results are not always matching data provided by official instances like NASA, but it depends on the sources used I guess (and measurements). International standards, that's how it was usually solved in the past. Sadly #GDPR is not one for now. #AIAct
NB: Neptune's is 11.27 m/s² @ #NASA
Mar 29 at 12:06pmWeb
Show threadNicolas MØUART1d ago
it's really fun, ask Qwen3.5 to calculate it, and it does find 11.27 m/s² eventually (and the calculus is correct), the explanation provided in regard to the discrepancy in results is debatable, sounds like hallucination, but Google returns 11.15 too upon request. #astrophysics