Sam Altman admits this: the peak of Chetgpt is over

General Director of Openai, Sam Altman He admitted that the company has lost its advantage in the space of artificial intelligence in the middle of the introduction of a Chinese company Deepseek And his model of reasoning R1. Nevertheless, he says that the brand will continue to develop in the industry.

The head of the company admitted that Openai was “not on the wrong side of history” in terms of open source development for its AI models. Altman and several other Openai leaders discussed the state of the company and its plans for the future while “asking me” in Reddit on Friday, where the team was honest with curious enthusiasts on various issues.

“Personally, I think that we were on the wrong side of the story here, and we need to open another open source strategy. Not everyone in Openai shares this point of view, and this is not our main priority, ”said Altman.

TechCrunch noted that Openai previously followed the open source model in the past, but quickly changed the march when he gained popularity and saw the potential of profit in generative AI.

In response to the Reddit user, who asked how Depseek will affect Openai plans for future models, Altman said: “This is a very good model. We will produce the best models, but we will maintain less advantages than in previous years. ”The Openai product director Kevin Vale added that the company has the potential to make its oldest and less avant -garde -open source models. Nevertheless, I had no details about what models, no terms about when it could happen. “We will definitely think about doing more this,” he said.

The arrival of Deepseek put Openai into a state of contemplation on its next movements. After the recent launch of the O3-Mini model, the company is currently considering the possibility of discovering transparency in the reasoning model so that users can observe their “mental process”. This is a function that is already available in the DeepSeek R1 reasoning model, which is one of the things that makes it an extremely attractive offer.

Altman and Weil also examined rumors about increasing prices for ChatGPTThe Chatbot application uses most brand models. The general director spoke about the hope of making the price of ChatGPT “cheaper” over time. Currently, his most expensive level of subscription to Pro costs $ 200 per month, while its plus level is $ 20 a month.

Thanks to Amma, the Openai team put forward several nearby products, including the next model of O3 reasoning, which may have an attempt from several weeks to several months. The next schedule of the main model does not yet have a release date, but most likely it is called GPT-5. There may also be an overview of the Dall-E 3 image generator, which has not had an important update since it became known two years ago.

The leaders also considered the company’s cooperation with the Government of the United States, which included the creation of a data center project called Stargate. When he was asked whether the Openai models would be used to develop devastating weapons, in particular, nuclear weapons, Vail said that he was sure that the researchers participating in the project would not turn the results of the model into nuclear calculations. ”

AMA has been taking place after two weeks of dizziness since Depseek announced his reasoning R1, which, as they say, justifies Openai and completes the models in terms of productivity with significantly lower operating costs. The service is also free for users and open source code for developers, which makes it one of the main competitors.

So far there are assumptions that Depseek may have used an illegal method called distillation to extract Openai data To train their own models, experts indicated that the damage has already been caused. Even President Trump called the turn of events as a “call for attention” for the US AI industry. One of the largest brand investors, Microsoft, also began to support Deepseek in the Azure and Gitub cloud to democratize its proposals for AI for its Popilot+PC.