According to Microsoft's head of AI, Microsoft is working to achieve "true self-sufficiency" in the field of artificial intelligence by building its own powerful models and reducing its dependence on OpenAI. Mustafa Suleyman said in an interview that this strategic shift was made after the reorganization of relations between the two parties in October last year. The move prompts the $3 trillion company to build its most advanced technology independently rather than relying on outside partners.

The co-founder of Google DeepMind, who will join Microsoft in 2024, said: "We must develop our own basic model that is absolutely cutting-edge, equipped with gigawatts of computing power, and some of the world's top AI training teams."

The software giant is investing heavily in building and organizing the massive data sets needed to train advanced systems. “That’s our true self-sufficiency mission,” he added.

Microsoft is one of the earliest and largest investors in OpenAI and has always relied on OpenAI's models to drive its own AI tools, such as the Copilot software assistant.

Last year, Microsoft agreed to allow the startup to complete a corporate restructuring while retaining a $135 billion stake in the ChatGPT developer and ensuring continued access to its most advanced models until 2032.

But the agreement also gives OpenAI greater freedom to find new investors and infrastructure partners, which could transform it into a more direct competitor.

Microsoft is also diversifying its bets, investing in other model developers like Anthropic and Mistral. At the same time, the company is also accelerating the development of self-developed models. Suleyman said these in-house models will be released "sometime this year."

Suleyman specifically pointed out that Microsoft's goal is to gain more enterprise market share by promoting "professional-grade AGI" (general artificial intelligence), that is, developing powerful AI tools that can complete daily tasks for knowledge workers.

“White-collar jobs—that is, sitting in front of a computer, whether you’re a lawyer, an accountant, a project manager or a marketer—most of these tasks will be fully automated by AI within the next 12 to 18 months,” he said.

He also added that within the next two to three years, these AI agents will be able to achieve better coordination in the workflow of large institutions. These AI tools will also have the ability to continuously learn and improve and take more autonomous actions.

He said: "Creating a new model will be as easy as making a podcast or writing a blog. In the future, it will be possible to design AI that meets the needs of every institution and individual on the planet."

However, Microsoft faces fierce competition in the enterprise market. Anthropic has taken the lead in AI programming tools, while OpenAI and Google are also investing heavily in competing for lucrative enterprise-level AI contracts.

Microsoft expects capital expenditures to reach US$140 billion in the fiscal year ending in June this year to increase investment in AI infrastructure.

Investors worry that such spending is pushing up the AI ​​"bubble", thus suppressing the performance of large technology stocks. Microsoft's stock price has fallen more than 13% in the past month.

"There's no question these are unprecedented times, and I think the market is trying to understand what the path is going to be over the next five years," Suleyman said. But he added: "We all have no doubt that those returns will eventually compound and show up in top and bottom lines."

Suleyman also said that another key direction of Microsoft is to apply AI to medical care, and is committed to building "medical super intelligence" and using AI programs to help solve the problem of shortages of medical staff and long waiting times in the medical system. Last year, the company released an AI diagnostic tool it said could outperform doctors on some tasks.

He added that Microsoft's goal is to create "humanistic superintelligence," that is, AI technology that is always under human control, in response to concerns that some AI labs are advancing powerful technologies too quickly and may escape the supervision of their creators.

“We have to reset that premise and only bring into the world those systems that we know we can control and serve us in a subordinate way,” he said.

“These tools, like any technology before them, are designed to enhance human well-being and serve humanity, not transcend humanity.”