OpenAI CEO Sam Altman on Saturday promised that the corporate will work on making GPT-5 “hotter” in conversations after customers complained in regards to the lack of wit, creativity, and relatability with the brand new AI mannequin. Altman additionally famous that his group had underestimated how a lot a number of the issues that customers appreciated about GPT-4o mattered, even when the brand new mannequin does higher in some ways.
Quickly after the GPT-5 launch, OpenAI introduced that it was pulling the plug on all its older fashions, together with GPT-4o, GPT-4.1, GPT-4.5, GPT-4.1-mini, o4-mini, o4-mini-high, o3, and o3-pro. As GPT-5 began rolling out to customers, they had been stripped of the choice to speak with the older variations of the chatbot, resulting in a widespread uproar.
OpenAI in the end relented and allowed customers to work together with GPT-4o, however provided that they’ve a ChatGPT Plus subscription, whereas free customers proceed to be restricted to GPT-5.
Throughout its launch weblog, OpenAI claimed that GPT-5 is quicker and extra correct than all its earlier fashions whereas having decrease hallucination probabilities (making stuff up) and fewer instances of sycophancy (being too agreeable). The AI startup additionally claimed important advances throughout quite a lot of duties, together with coding, reasoning, writing, health-related questions, and multi-modal skills.
Nonetheless, social media reactions counsel that GPT-5 may typically be slower than the earlier fashions whereas additionally offering shorter replies that present shallow emotional depth.
Sam Altman on modifications coming to GPT-5:
After the backlash on social media, Altman took to X to share the modifications that can be coming to GPT-5 within the coming days and weeks.
“We for positive underestimated how a lot a number of the issues that folks like in GPT-4o matter to them, even when GPT-5 performs higher in most methods.” Altman wrote.
“We’re going to concentrate on ending the GPT-5 rollout and getting issues steady after which we’re going to concentrate on some modifications to GPT-5 to make it hotter. Actually good per-users customization will take longer.” he added
Altman famous that customers have very totally different opinions on the strengths of GPT-4o in comparison with GPT-5, and OpenAI will work on offering customers with methods to customise various things in the long run. As an illustration, Altman mentioned some customers like emojis within the responses they get from ChatGPT, whereas others need ‘chilly logic,’ and a few need heat and a special form of emotional intelligence.
