Openai investigators, engineers and executives, including executive director Sam Altman, took Company and in general at the time the current star technology, artificial intelligence. As expected, the Entry of the Deepseek Chinese model sneaked into almost all the issues.
Openai became the Fashion company After the launch of the Chatbot Chatgpt and the generative AI model, Dall E. the American firm specializing in AI research, initially founded as a non -profit company, had achieved investments of large technological ones such as Microsoft and prepared to lead the largest project World Infrastructure World, the The Stargate Project valued at 500,000 million dollars and where the company has committed at least 18,000 million dollars.
But Deepseek has arrived to put the technological sector up and place Openai in a somewhat precarious position. The company has to fight now against the perception that He is giving ground in the race for the leadership of AI To Chinese companies such as Deepseek (and other Asians). At the same time, you have to underpin your relationship with the new Trump administration and convince investors of the need to spend half a billion dollars on the aforementioned Stargate.
There are few who think Deepseek has punctured the AI bubble, A clearly inflated sector both in the amount of investment and in the cases “benefits” for companies and consumers. In addition to the doubts arising if you have to continue feeding an insatiable resources, both of processing and energy, the emergence of this model also has geostrategic consequences of consideration by confirming that China is in a position to compete and win in the states Joined.
OpenAI, less leadership in AI
With all the above on the table, Reddit’s interest was clear and did not disappoint. Executives commented on the main issues that affect the company and that, due to their dominant position in the IA segment, ending affecting the rest of the industry. In addition, CEO himself, Sam Altman, admitted what anyone can see, that Deepseek’s emergence has reduced OpenAi leadership.
Open source
An important section having that Deepseek is open source under MIT license and that this development model can be imposed in the future. Altman said Openai had been “On the wrong side of the story” With regard to the code opening of its technologies. While Openai had opened the code of its models in the past, the company has generally favored a closed -owned and code development approach.
«I personally believe that We need to devise a different open source strategy«Said Altman. “Not everyone in Openai shares this opinion, and it is not our top current priority … we will produce better models (in the future), but we will maintain a lower advantage than in previous years”.
In a subsequent response, Kevin Weil, director of OpenAi products, said OpenAi is considering the possibility of publishing older models that are no longer generating. “We will definitely think about doing more things like that”he said, without going into more details.
‘Thought process’
In addition to reconsidering the opening of release of the code, which many consider such a philosophical and strategic issue, Altman acknowledged that Deepseek had pressed them to reveal more than the functioning of the so-called reasoning models, such as the O3-mini launched over the weekend and showing its «Thought process».
Currently, Openai models hide their reasoning, a strategy aimed at prevent competitors from extracting training data For your own models. In contrast, Deepseek’s reasoning model, R1, shows its complete chain of thought.
«We are working to teach much more than we show today; (Show the model thinking process) will be very soon »Weil added. «It is not yet known with certainty what will happen; Showing the entire chain of thought leads to competitive distillation, but we also know that people (at least advanced users) want it, so we will find the correct way to balance it.
Precio
Altman and Weil tried to dissipate rumors that Chatgpt, the Chatbot platform through which OpenAi launches many of their models, would increase in the future. Altman said he would like to make Chatgpt “cheaper” over time, if possible.
Openai executives said in their day that Openai was losing money in his most expensive chatpt planthe ‘pro’ model that costs 200 dollars per month. In a related thread, Weil said Openai continues to see evidence that a greater computing power leads to “better” and higher models.
It is the argument used so far to sustain the need for investments such as Stargate, the massive data center project recently announced by the company, with a billionaire Price. “Serving a growing user base is also promoting the demand for computing within OpenAi,” they described, although they did not enter the discussion of use of much more efficient models in costs and energy consumption, such as Deepseek.
Weaponry
A Reddit user asked if Openai models, with or without automatic improvement capacity, would be used to Develop destructive weapons, specifically nuclear weapons. The question, which has not been answered since the incidence of AI in the ‘Lethal Autonomous Weapons Systems’, arrived in reference to the announcement of an association with the United States Government to deliver its models to national laboratories, of the United States, partly for nuclear defense investigation.
Weil’s executives simply said they trusted the government … «We know these scientists and are experts in AI, in addition to world -class researchers. They understand the power and limits of the models, and I do not believe that there is the possibility that they simply convert the result of a model into a nuclear calculation. They are intelligent, they are based on evidence and perform a lot of experimentation and work with data to validate all their work ”.
New versions
The OpenAi team was asked several more technical questions, such as when the next reasoning model will be launched; When the next model “without reasoning” company, the GPT-5 could arrive, or when they could reveal a successor of Dall-E 3, the company generation model of the company that has remained quite behind behind the competence.
“We are working on all of them”Altman said, but without specifying a throwing schedule. “I think it will be worth the wait”said the CEO of a company that still has to digest the cataclysm caused by the launch of its rival, Deepseek, and other different models that will arrive from Asia.
