摘要：SoftBank founder Masayoshi Son and OpenAI CEO Sam Altman unveiled the 50-50 collaboration at an event in Tokyo, the capital of Jap

TMTPOST – SoftBank Group and OpenAI have announced a joint venture aimed at developing and selling artificial intelligence (AI) services to businesses across Japan. The partnership marks one of the most significant efforts to expand OpenAI’s enterprise customer base beyond the United States.

SoftBank founder Masayoshi Son and OpenAI CEO Sam Altman unveiled the 50-50 collaboration at an event in Tokyo, the capital of Japan. The venture, which will be housed under SoftBank’s telecom arm, plans to recruit 1,000 employees from SoftBank to market OpenAI’s products to industries ranging from automotive to retail. SoftBank Group companies, including key investees such as PayPay and LY, are expected to collectively spend $3 billion annually on OpenAI’s technology.

This partnership shows SoftBank’s growing role in global AI development. The company has been at the forefront of major AI initiatives, from leading the $100 billion "Stargate" project in the U.S. to investing in data centers and AI research in Japan.

With this deal, SoftBank joins major tech players like Meta and Microsoft in making billion-dollar bets on AI infrastructure, further solidifying its commitment to shaping the future of AI.

“If more is better, we should do a lot,” said Son, addressing concerns about the need for large-scale AI investments. His remarks appeared to counter the emergence of DeepSeek, a Chinese startup that has challenged the assumption that massive spending is necessary for cutting-edge AI development.

DeepSeek has been making some promotional moves recently. They might try to mimic or catch up, and their costs could be much lower, Son remarked.

Altman reiterated the fundamental advantage of scale in AI development. “Some people suggest using small models with compress, I think people still don't understand how much how exponential the return is. The cost is exponential too but I think the return is even more exponential. To the smartest model we can make and that will require the biggest compute.”

Altman’s response underscored the deep divide between OpenAI and DeepSeek in terms of technological approach. While DeepSeek may focus on cost-efficient models, OpenAI remains committed to maximizing intelligence through sheer computational scale.

When discussing the speed of AI development, Son presented a startling calculation, saying that right now, the models are improving tenfold each year. Meanwhile, chip performance is also advancing at a tenfold rate. On top of that, the 'Stargate' project is massively increasing the number of chips—scaling by another factor of ten annually. So, 10×10×10... In one or two years, that's a 1,000-fold improvement. Then, another year of 10×10×10 means another 1,000-fold increase. A thousand times a thousand is a million-fold improvement, Son explained further.

Altman reiterated his belief in the exponential nature of AI progress: "I don’t think people truly understand how exponential the return is."

OpenAI has unveiled a groundbreaking AI research agent, "Deep Research," marking a major step forward in artificial intelligence capabilities. The AI system is designed to conduct complex research tasks, analyze vast amounts of data, and generate insights with minimal human input.

Altman pointed out the critical role of computational power in advancing AI. "To get small models to do incredible things and to really push the frontier of intelligence, that's going to require a huge amount of compute," he said. The most value will be created at that frontier, and finally being able to do this at scale is fantastic, he added.

OpenAI's focus on large-scale computation mirrors the early evolution of the internet. When broadband first arrived, people questioned its necessity. But as high-resolution images and video content emerged, the demand for bandwidth skyrocketed, Altman explained. AI is following a similar trajectory—more computational power leads to exponentially greater returns, he added.

The newly launched Deep Research tool is designed to handle tasks that might take humans hours, days, or even weeks. It can scan web pages, analyze images and PDFs, synthesize information, and generate detailed reports. Unlike ChatGPT, which provides instant responses, Deep Research processes information systematically, mimicking the work of a sophisticated analyst.

"This is just the beginning," Altman noted. “There are more to come about synthesizing knowledge, eventually AI will be inventing new knowledge.”

OpenAI also introduced its first AI agent Operator, which can navigate web pages, interact with user interfaces, and complete tasks on behalf of users. The company envisions AI agents evolving into fully autonomous digital assistants capable of making decisions and taking actions without human supervision.

"With ChatGPT, you could talk to it about anything, with agents, you'll be able to do anything," Altman explained.

The launch of Deep Research represents a significant step toward this vision, demonstrating AI’s ability to execute complex tasks with minimal oversight. The technology is expected to be particularly valuable for professionals in finance, law, and scientific research.

AI agents like Deep Research could revolutionize productivity, serving as virtual research assistants capable of handling economically valuable tasks at scale, Altman noted, adding that OpenAI believes that these tools will eventually transform industries by automating high-level decision-making and analysis.

"I think this is one of the best things OpenAI has ever launched and it really points at what's going to be possible with AI agents," Altman said.

It's the start of a new era where AI doesn't just assist—it actively works alongside us to generate real economic value, he added.

来源：钛媒体