OpenAI’s CEO, Sam Altman, recently admitted in a Reddit AMA that his company is facing significant challenges. One of these challenges is the company’s policy of keeping its AI technology locked away, a decision that is starting to haunt them as China’s DeepSeek moves ahead with open-source AI models. In his own words, Altman stated, “I personally think we have been on the wrong side of history”. This is a significant admission from a man who seldom acknowledges mistakes.
DeepSeek recently released an open-source model that is not only affordable but also powerful, and alarmingly close to what OpenAI has to offer. The model has been well-received by developers due to its accessibility and affordability, in stark contrast to OpenAI, where users are required to pay up to $200 a month for ChatGPT Pro.
Altman also revealed that his team is discussing how they can adapt to these changes before they lose control of the narrative. He stated, “We’re figuring out a different open-source strategy”. However, he also mentioned that this is not their top priority at the moment, as there seems to be a lack of consensus within the company.
Kevin Weil, OpenAI’s chief product officer, provided more insight into the situation. He mentioned that OpenAI is considering open-sourcing its older models that are no longer at the cutting edge. However, he did not specify which models or when this might happen, which has led to some skepticism.
Altman acknowledged another problem plaguing OpenAI – the lack of transparency in their models. Currently, if a user asks ChatGPT a question, it provides an answer without disclosing how it arrived at it. The company has previously stated that this is to prevent competitors from reverse-engineering its systems.
Contrarily, DeepSeek’s models are transparent. Their latest reasoning model, R1, shows the user the entire thought chain, step-by-step. This feature has been well-received by developers and researchers as they can understand exactly how the AI works and make necessary adjustments.
Altman conceded that this difference is to OpenAI’s disadvantage and said, “We’re working on showing a bunch more than we do today”. He added that the company is trying to find a balance between transparency and not providing an easy route for competitors.
Rumors of a possible price increase for ChatGPT were also addressed during the AMA. Altman dismissed these rumors, stating his intention to make ChatGPT less expensive. However, he also admitted that the company is losing money, as the $200-per-month Pro plan is not profitable, indicating the need for a viable solution.
During the AMA, Altman also admitted OpenAI’s insatiable need for more computing power. He explained that as models become larger and more sophisticated, large data centers become crucial.
When asked about AI’s role in weapon development, the discussion took a more serious turn. Weil responded that he had faith in the scientists working on the project and asserted that OpenAI is aware of the potential risks.
Finally, Altman touched on the concept of recursive self-improvement, where AI improves itself without human intervention. He confessed that his stance on the matter has changed, and he now believes a “fast takeoff” scenario is more likely.
On a more positive note, Altman also provided updates on future technology, including the upcoming reasoning model, o3, and the next iteration of their image-generation model, DALL-E 3. He also mentioned GPT-5 but did not provide a release date.
