On February 2, the Financial Association reported (Editor: Shi Zhengcheng) that as a summary of the recent shock to AI practitioners and Wall Street by the DeepSeek open-source large model, Sam Altman, the CEO of OpenAI, which previously adhered to a closed strategy, candidly admitted that this tech giant stands on the "wrong side of history" on the open-source issue.

During a Q&A session on the USA platform Reddit, a participant asked Altman if his company would consider releasing some of the technologies in its AI model to better demonstrate how its system works.

Altman replied, "In my view, we are on the wrong side of history on this issue. Now we need to come up with a different open-source strategy."

OpenAI's leader also admitted that not all colleagues Hold this view.

OpenAI, founded in 2015, initially stated that it would release its AI models and research as long as it served the public interest. However, with the explosive popularity of ChatGPT and the influx of billions of dollars in capital, the company no longer open-sourced its models, citing "concerns about competitive pressures and sharing too much information."

This practice also angered another co-founder of OpenAI, Elon Musk, who is suing OpenAI, accusing Altman of betraying the company's original intention. According to past reports, OpenAI has not only transformed into "Close AI" but has also been dismantling its initial non-profit structure to facilitate larger-scale financing.

According to this week's latest reports, OpenAI is seeking a $40 billion round of financing with the assistance of SoftBank, with a potential valuation of up to $340 billion. Meanwhile, the number of paying users for ChatGPT reached 15.5 million at the end of last year, tripling in nearly a year. Even at the minimum price of $20 per person, this equates to at least $0.3 billion in monthly revenue.

At the same time, the capital-led and closed development competitive logic spearheaded by OpenAI is facing upheaval due to the rapid improvement and cost reduction of open-source model capabilities. The definition of open-source models means that anyone can freely download and use them, and adjust the specific operational methodologies by changing the underlying code.

As the first step of the 'reopening', Altman stated that DeepSeek is pushing OpenAI to reveal more of its reasoning model's 'thought process', which was previously intentionally hidden by OpenAI due to concerns about competitors scraping data to train their own models. In contrast, DeepSeek's R1 model demonstrates a complete chain of thought.

OpenAI's Chief Product Officer Kevin Weil responded, stating: 'We are working to showcase more content than we currently do. The specifics are still to be determined, and showing the entire chain of thought could lead to competitive extraction, but we also understand that people want it, so we will find the right way to balance it.'

In other Q&A sessions, Altman mentioned that the company's next flagship reasoning model o3 will be released in 'a few weeks, but less than a few months', while there is no timeline for the release of 'GPT-5'. Regarding the proprietary model DALL-E 3, which has already fallen behind mainstream competitors in 'text-to-image' capabilities, he also indicated that efforts are being made to develop new products.

Editor/danial