Source: Zinc Finance
Last week, in a program, OpenAI CEO Sam Altman said, "In order to prevent OpenAI from being surpassed, there is pressure every day. Plan to meet with the leadership of DeepSeek."
It has to be said that it is rare to see Sam Altman so humble. Even if he was driven out of the CEO position by the company's directors and senior executives, he calmly took out his mobile phone to take a selfie, turned around and went to Microsoft to regain power.
Now, with DeepSeek popular all over the world, Sam Altman and OpenAI have undoubtedly been hit unprecedentedly.
Since ChatGPT-4 opened the first year of AI, OpenAI has been regarded as the "absolute leader in the AI world" and is the target of countless Internet giants and AI startups.
Now, the pursued has become the pursuer in turn.
The Spring Festival has just passed, and the world has ushered in the "DeepSeek moment", which allowed DeepSeek to reach 100 million users in just 7 days.
In comparison, the ranking of the time it takes for Internet products around the world to reach 100 million users shows that ChatGPT took 2 months, second only to DeepSeek; TikTok took 9 months to rank third; Pinduoduo took 10 months and WeChat took 1 year and 2 months, ranking fourth and fifth.
The number of users is just the appetizer for DeepSeek to surpass ChatGPT. What scares Sam Altman the most is thatunder the impact of the free DeepSeek, no user is willing to choose the paid ChatGPT.
When facing the company's biggest challenge, OpenAI also had several strong struggles for survival.
The first card Sam Altman played was to cut prices.
Just when DeepSeek caused OpenAI's API subscriptions to plummet by 37% in two weeks, the pampered OpenAI panicked and urgently announced a 50% API price cut and opened up some source code.
But this did not alleviate OpenAI's ever-declining situation. Even the Wall Street Journal stood up and publicly mocked: "This is just a stress response before the collapse of the closed-source empire."
In order to ease the downward trend, Sam Altman made a series of stupid moves and played the second card: relaxing content review.
Last Tuesday, February 14, OpenAI threw a bombshell: officially lifting the ban on ChatGPT's "adult content" generation function.
Abroad, chatbots have been under much criticism. In particular, in November 2024, a sensational "AI-instigated suicide" incident occurred in the United States. Google's large model Gemini directly issued a message to a 20-year-old college student: "Human, please die, please."
The outside world has been calling on AI companies to strengthen content review, but OpenAI has gone the other way and further lifted the ban on ChatGPT's "adult content" generation function.
The implementation of this measure has caused the ChatGPT application to surge in daily downloads by 280%, but the user retention rate the next day has plummeted to 11%.
The data will not lie, which shows that "curiosity traffic" is just a mirage, and it is difficult to retain real users.
All the tricks have been declared invalid, and OpenAI has finally come to the crossroads of fate:
Is it to continue to close the source, maintain the status quo, and blindly believe that its product strength is enough to match the free DeepSeek?
Or to overturn all its business plans and turn to open source?
Sam Altman and OpenAI have always been firm "AI conservatives": insisting on closed source, delaying open source, and pursuing profit maximization.
The "API+subscription system" that ChatGPT has always implemented is also a powerful business moat for OpenAI. The so-called "API+subscription system" is very simple to understand, that is, paid membership.
At the end of January 2023, ChatGPT launched a Plus subscription option of $20 per month; at the end of 2024, ChatGPT added a Pro subscription option of $200 per month.
Paid users can enjoy various permissions, such as priority access, shorter queue time, higher concurrent use, etc.
After OpenAI established the "order of the AI world", almost all AI companies and large model companies are "feeling OpenAI across the river" and using the "API+subscription system" as the unified standard of the AI industry.
But the most classic business war model on the Internet is to defeat payment with free. Especially in the Internet world, it is undoubtedly a very difficult task to get users who are used to free to pay.
According to Mistral, a foreign open source AI developer, the vast majority of customers are using smaller free models, and less than 10% of customers use the largest paid models.
As DeepSeek presses forward, Sam Altman publicly admitted in a Reddit Q&A event: "OpenAI is on the wrong side of history on the issue of open source, and now needs to come up with a different open source strategy."
The day before yesterday, that is, February 18, Sam Altman acted as a "riddler" on a foreign social platform, hinting at two directions of OpenAI's open source plan:
The first is a fairly small o3-mini-level model that still needs to run on a GPU; the second is a side-end model suitable for smartphones.
Sam Altman undoubtedly turned on the switch for OpenAI's full transformation to open source. In the past six years, OpenAI has not open sourced any large models since GPT-2.
I didn't expect that the task that Musk couldn't complete was easily solved by DeepSeek.
But, fortunately, although the difficulties in front of them popped up one after another, Sam Altman and OpenAI still had one last trump card to turn things around.
That is: GPT-5.
Turn the time back to two years ago, on March 14, 2023, OpenAI released GPT-4, opening the "first year of AI". It was a "ChatGPT moment" that was no less than the current "DeepSeek moment".
Two years later, the call for GPT-5 is getting louder and louder, but Sam Altman only handed over a "semi-finished product" GPT-4o on May 14, 2024, and the new generation flagship GPT-5 has always been in a state of dystocia.
Sam Altman did not want to launch GPT-5, but it was too difficult to make a GPT-5 that met the psychological expectations of users.
In mid-2023, OpenAI launched the first training code-named "Arrakis", but the test results showed that the training of larger-scale AI models takes a very long time, which will cause the overall cost to soar.
From May to November 2024, OpenAI conducted several small-scale trainings and the second large-scale training, but the problems of too little data and insufficient data diversity still exist.
Sam Altman said frankly: "The cost of training GPT-4 is about $100 million, and the cost of training AI models in the future will reach $1 billion. Now, the training of GPT-5 for a month has cost $500 million, and it has not achieved the desired results."
After the rise of DeepSeek, OpenAI was forced to overturn the closed-source logic and move towards full open source, and it is bound to redesign the plan for GPT-5.
The latest news is a hint from Sam Altman on the social platform. He said that GPT-5 will be released "within a few months", perhaps in late 2025.
It's just that I don't know whether OpenAI can still withstand the bombardment of many open source models led by DeepSeek at that time.
ChatGPT said: Shinhan Bank has added cryptocurrency tracking and educational tools to its SOL app, becoming the first South Korean bank to offer such services. The move signals growing acceptance of crypto in traditional banking, with plans to expand into crypto purchases and wallet management soon.
WeatherlyAn Australian fintech company, Block Earner, has launched the country’s first mortgage that accepts Bitcoin as loan security. This became possible after a court ruled the product doesn't need a financial licence, letting people use crypto to buy homes without selling it.
AnaisScale AI has cut 700 jobs, including 200 full-time staff and 500 contractors, as part of a major restructure. The company is narrowing its focus to coding, language and audio data after rapid growth created overlaps and confusion.
WeatherlyTokenize Xchange will shut down its Singapore operations by 30 September 2025 after failing to get a licence from the Monetary Authority of Singapore. The company plans to move to Malaysia and the UAE, and all 15 local staff will lose their jobs.
WeatherlyStrategy now holds over 600,000 Bitcoin worth nearly $71 billion, making it one of the biggest corporate treasuries in the US. Its chairman, Michael Saylor, has hinted at another large Bitcoin purchase with a cryptic post online.
AnaisMicrosoft is likely to follow the EU’s voluntary AI rules, while Meta has refused to sign, saying the code is too vague and strict. The EU’s code asks AI developers to share data details, follow copyright rules, and manage risks before full enforcement starts in August 2025.
AnaisA Colorado dentist is on trial for killing his wife by poisoning her with cyanide and chemicals found in eye drops. Prosecutors say he later asked his daughter to fake a video making it look like the wife took her own life.
WeatherlyA Japanese crypto group has asked the government to introduce a flat 20% tax on crypto trading profits. They say the current tax system is too high and stops people from investing.
WeatherlyThe SEC Chairman said Ethereum is not considered a security, which has encouraged companies to increase their Ethereum holdings. This clear regulatory view has helped boost Ethereum’s price and sparked more interest from investors and firms.
AnaisFederal Reserve Chair Jerome Powell is facing a criminal referral for allegedly lying to the Senate about the true cost and scale of the Fed’s $2.5 billion headquarters renovation. Internal documents suggest the project includes luxury features Powell did not mention, raising concerns over misleading testimony and public spending.
Weatherly