OpenAI makes GPT-5 ‘friendlier’ after widespread consumer backlash

About two weeks in the past, OpenAI launched GPT-5. The most recent AI mannequin within the GPT line, GPT-5 was put forth as the corporate’s “smartest, quickest, most helpful mannequin but” with “built-in considering” and “expert-level intelligence.” However the launch backfired for one vital motive.

A part of the modifications in GPT-5 concerned addressing the sycophantic positivity present in earlier fashions, the place the AI chatbot would incessantly reward the consumer to an undo diploma and emphatically comply with make the consumer really feel higher. Numerous customers disliked this, so GPT-5 was made to be “much less effusively agreeable” and “use fewer pointless emojis.”

However GPT-5 confronted its personal backlash—for being too chilly, too distant, too company. Now, the newest information is that OpenAI is rolling out an replace that may make the GPT-5 “hotter and friendlier,” reviews TechCrunch.

The modifications will reportedly embrace encouraging phrases equivalent to “Good query” or “Good begin,” however no outright flattery and fawning. “Inside exams present no rise in sycophancy in comparison with the earlier GPT-5 character,” OpenAI stated in a social media put up.