The new ChatGPT is stupider than expected

Altman presented this artificial intelligence (AI) model as the smartest, fastest and most useful to date, purporting to have the intelligence of a PhD student in any field”.

According to OpenAI, GPT-5 was to be more reliable: it hallucinates much less than previous models, and admits when it doesn’t know instead of inventing a seemingly convincing answer, OpenAI claims.

However, since its launch, GPT-5 has been criticized by users for its duller, drier personality, shorter answers and lack of engagement compared to GPT-4o.

But above all, users are complaining that the AI can’t answer simple questions. For example, one user asked ChatGPT to count the number of times the letter “b” appeared in the word “blueberry”. The chatbot answered three, instead of two.

The same is true when asked to draw the last American presidents: the portraits are badly done, the dates are wrong and the names are misspelled.

Sam Altman admitted that GPT-5 was experiencing difficulties and promised to remedy the situation. According to Altman, the problem stems from one of GPT-5’s innovations. The latter is supposed to combine different models and solicit the most relevant according to the complexity of the user’s query. Prior to GPT-5, users had to manually select the model best suited to their needs. Now, GPT-5 chooses itself, but the settings still need to be improved, suggests the OpenAI boss.

“We expected some bumpiness as we roll out so many things at once. But it was a little more bumpy than we hoped for!” reacted Sam Altman.

In rolling out the new model, OpenAI basically retired its older models, including the very popular GPT-4o, which had been on the market for over a year. As a result of this decision, those who loved the old version of the chatbot could no longer use it. Over 4,000 people signed a petition on Change.org to force OpenAI to reinstate it.

Sam Altman, CEO of OpenAI, heard them and announced the gradual return of the 4o model in a post on X on August 9.