Several users of OpenAI's GPT-4 have been complaining for weeks that the AI model is performing worse, with some calling the system "lazier" and "dumber."
Apparently, this time, the customer is wrong.
"No, we haven't made GPT-4 dumber. Quite the opposite: we make each new version smarter than the previous one," Peter Welinder, VP Product at OpenAI, tweeted following an Insider report on the issues. "Current hypothesis: When you use it more heavily, you start noticing issues you didn't see before."
Transform talent with learning that worksCapability development is critical for businesses who want to push the envelope of innovation.Discover how business leaders are strategizing around building talent capabilities and empowering employee transformation.Know More "If you have examples where you believe it's regressed, please reply to this thread and we'll investigate," he added.
Several people responded, saying they had DM-ed Welinder with examples.
Ethan Harris, who lists himself on Twitter as a retired Army medic, replied to Welinder with a specific example.
Another user pointed to a recent post from Nihit Desai, a startup co-founder, who reported that performance degraded for data labeling tasks.
Yet another user on Twitter pointed to OpenAI's own developer forums, where many GPT-4 users have vented frustrations and posted examples of worse performance. Insider's story earlier this week also reported some of these examples. Many hours before publication, Insider asked OpenAI for comment on all this, and the company didn't respond.
You can check out the developer forums here and here.