Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
A recent paper published in Nature highlights that larger and more instructable language models (LLMs) become less reliable as they are scaled up. Amid the chaos at OpenAI, the study finds that these models, which are trained on more data, tend to attempt more questions they cannot answer accurately, increasing the likelihood of errors. Unlike earlier models that avoided questions they couldn't answer, newer models are more prone to fabricating responses. This phenomenon underscores the need for a shift in AI design towards enhancing reliability. Experts, including Carissa Veliz, note that large language models do not know the limits of their own knowledge.
'Large language models (LLMs) seem to get less reliable at answering simple questions when they get bigger and learn from human feedback.' https://t.co/EgXLAV1asT
Scaling up and shaping up LLMs increased their tendency to provide sensible yet incorrect answers at difficulty levels humans cannot supervise, highlighting the need for a shift in AI design towards reliability, according to a @Nature paper. https://t.co/5gVG5yQvrK https://t.co/JbHJ7KB0HG
AIs get worse at answering simple questions as they get bigger https://t.co/EJodSXp74o ✍️@stokel via @newscientist #AI #LLMs #ScalingUp #Hallucination 💡“Large language models do not know the limits of their own knowledge” - @CarissaVeliz @PerBBerggreen @sonu_monika…
