While people ridicule GPT-5, Sam Altman says that Openai will need “billion” infrastructure

People have ridiculed GPT-5, the new version of the largest language model in Openai, since the second when it was launched, with many users complaining that it is stupid, boring and not as good as the last LLM as the company has published. Sam Altman, the CEO of the company, has comforting words for those who can be concerned about the management of his business: AI is a bubble, and OH, also BTW, my company is about to spend France’s GDP to build our IA infrastructure.

This seems to be most of what Altman said at a dinner held in San Francisco on Thursday with a group of journalists and other OpenAi executives, according to The Verge. During this occasional conversation, Altman admitted that the latest version of software in his company had been struck off, but promised that the future was promising for his business, as well as his industry.

“I think we totally screwed up some things about deployment,” he said. “On the other hand, our API traffic has doubled in 48 hours and develops. We are out of the GPUs. Chatgpt has struck a new top of users every day. Many users really like the model switch. I think we learned a lesson on what it means upgrading a product for hundreds of millions of people in one day. ”

At the same time, Altman seems to agree with the criticisms of his industry who described him as a “bubble” similar at the start of the Internet. “Are we in a phase where investors as a whole are excited about AI?” Said Altman. “My opinion is yes.”

For a certain time, criticisms and commentators wondered if the excitement around the AI industry is due to a precipitated collapse. Certain recent industry movements – as a very unhappy day of stock for the data startup and IA infrastructure, the Coreweave Startup – added to these concerns. Many spectators have noted that, so far, AI is a silver pit, companies launching large sums of money in industry while dreaming of a day, in one way or another, to make a profit.

During his conversation with journalists, Altman added: “When bubbles occur, intelligent people are overexcited about a core of truth,” said Altman. “If you look at most of the bubbles in history, such as the technological bubble, there was a real thing. Technology was really important. Internet was really a big problem. People were excited.”

Whether AI is a bubble or not, Altman always wants to spend a certified mad money that builds the AI infrastructure of its business. “You should expect OpenAi to spend billions of dollars for the construction of the data center in the not very distant future,” Altman told journalists.

It is this truly absurd investment scale in AI that stimulates the occasional viewer to wonder what really use. Indeed, the only question that never seems to rise during conversations with Altman is whether a cost-to-service cost analysis has ever been managed on its industry. In other words, AI is really worth it?

This only question could easily be repressed in a variety of more specific questions. For example, could a good question be: is it really worth spending billions of dollars just to create a range of slightly fun chatbots that only give you specific information a certain percentage of time? Or: wouldn’t billions of dollars be better spent, as if to help the poor or improve our education system? Also: are chatbots a societal necessity, or they simply seem pleasant to have? What is the most useful of the AI that, say, a search engine? Can’t we just stick to search engines? The negative externalities associated with the use of AI (a massive energy imprint, alleged reduced mental capacities in users and a cheating plague in higher education) prevail over positive (access to a slightly more practical means of finding online information)?

I understand that these questions may seem somewhat obvious in a certain way, but it is also difficult to know if someone has asked them again. Hopefully, at the next relaxed SF dinner with Altman, someone will take them away. Gizmodo contacted Openai to comment, and we will update this message when we receive an answer.