Sam Altman: Openai was ‘wrong side of history’ on the open source

One-day product releases answered questions from open researchers, engineers and managers, including Openai CEO Sam Altman, Altman Widely changing Reddit AMA Friday.

Openai finds a slightly dangerous situation. Fights the perception of the EI race Chinese companies such as DeepSeekThe IP claims that Openai claims are stolen. The Chatgpt manufacturer was working Lift their relationship with Washington to the shore and at the same time a momentary occasionally Ambitious data center projectreported to be groundwork For one of the biggest financing tours of history.

Altman, Deepseek has reduced Openai’s management in the EU and he said he believed that he was “wrong side of history” while opening Openai’s technologies. Although Openai has open source models in the past, the company generally liked the property, indoor source development.

“Understand a different open source strategy,” (I think it is necessary) “said Altman. “Everyone does not share this view in Openai, and this is not our current highest priority (…) better models (we move forward)

In response to a follow-up, Openai’s Chief Product Officer Kevin Weil said he thought the oldest models called the most modern in Openai. “We will definitely think about our work more,” he said without passing more detail.

Philosophy, Altman, Deepseek, who did not want to release Openai, said the “process of thoughts”, the company’s potential for the so-called substantial models said. Currently, Openai models hide their disputes, a strategy designed to prevent competitors for their models for their models. On the contrary, Deepseek’s justification model shows the R1, a full thinking chain.

“We are working on showing a gang more than we show today – (to show the model of model thinking) it will be much more soon,” he said. “We also know that they want to have a competitive distillation, which shows all the chains of everything in TBD, but they want people (at least power users), so we will find the right way to balance it.”

Altman and Weil, WHATGPT, Chatgpt, the chatbot application launched in many models of Openai, wanted to increase the price increase. Altman said they wanted to make Chatgpt “cheaper” over time.

Altman said before Openai, Pratiest Chatgpt Pro, which cost $ 200 per month, lost money in Chatgpt Pro.

Weil in a slightly related issue, “Better” and more “better” and more singers continue to see the evidence that more settlements are caused by more computing. Stargate, Openai’s projects such as Openai are a large part Recently announced The Mass Information Center Project, Weil said. The growing user serving the base of Openai, which has increased the required requirement and continued.

He asked to improve the recursive self-improvement by strong models, Altman said he thought he thought “fast flight”, once he believes. Recursive self-developing is a process that the AI ​​system can develop its intelligence and opportunities without human access.

Of course, it should be noted that it is notorious for the excess of the Altman. It did not pass long Reduced Openai’s Bar for AGI.

A Reddit user asked Openai models, self-evolving or not to develop destructive weapons – special nuclear weapons. This week, Openai announced a partnership to provide US national laboratories to US national laboratories to US national laboratories.

Weil said he trusted the US government.

“I recognized these scientists and they are AI specialists in addition to world-class researchers,” he said. “They understand the strength and boundaries of the models and do not think that there is no chance that a few model access to a nuclear calculation.

The Openai team was given a few questions of a more technical nature, O3’s next substantiation model, O3, Several weeks, “Altman” said the company’s next flagship can be “foolish” model, GPT-5, land (“still one There is no schedule, “he said. About two years ago, 3 years ago. Image generation Tech, Dall-E 3’s debut and model has improved the bounce and limits from 3 There is no longer a number of criteria tests.

“Yes! We are working on it,” said Weil, a Dall-E 3 follow up. “And I think it’s worth waiting.”