A year ago, Igor Babuschkin left xAI, the artificial intelligence start-up he founded with the tech mogul Elon Musk.
Now, he is among a growing number of voices across Silicon Valley arguing that A.I. should not be dominated by any one company. He believes this transformational technology should be controlled by the people who use it, much like a personal computer or a smartphone.
Mr. Babushkin’s new start-up, River AI, wants to put a new kind of computer inside people’s homes, so they can own and use A.I. without oversight from anyone else, including River AI.
Backed by $1 billion in funding, the company plans to share its A.I. technologies as open-source software. This means that anyone — including businesses, independent software developers and consumers — would be free to use and modify these technologies.
“We hope that in the future, A.I. will be trained to benefit you as the individual, rather than being a centralized thing that is controlled by one big corporation,” Mr. Babuschkin said in an interview.
Mr. Babuschkin, 35, plans to build a new kind of computer server that would allow everyday businesses or individuals to easily run open-source technologies on their own hardware. That would let them control the private data they shared with A.I. systems, rather than sending it through machines operated by River AI or another company.
The personal data people share with A.I. will be enormous, Mr. Babuschkin said. He envisions his company’s technology operating as a digital assistant that people can use across multiple devices, talking to it much as they would with Apple’s Siri assistant.
He said he wanted to provide ways for people to “retrain,” or modify, this technology as they saw fit. In this way, they could define how A.I. behaves and how they can put it to use, rather than following the rules created by companies like Anthropic and OpenAI.
Based in Palo Alto, Calif., River AI has about 20 employees, and it is still months away from releasing its own open-source technologies. But Mr. Babuschkin and his fellow researchers, many of whom previously worked for xAI, OpenAI and Tesla, hope to build systems that can learn new behavior on the fly, as people use them.
He envisions a future where people can change the behavior of these systems with a few spoken commands. “We want a democratized world where anyone can control and shape A.I. for their own needs,” he said.
Open-source A.I. technologies have become a hot topic in recent weeks, after two Chinese start-ups released open-source systems that are nearly as powerful as the leading American systems.
U.S. companies have not been as aggressive in releasing open-source systems, which many executives, investors and other industry experts believe will slow the progress of American technologies.
Mr. Babuschkin and his investors, including several prominent venture capital firms and the chipmaking giant Nvidia, hope his new company can fill this void.
“Igor is one of the very few people who can do it — and can do it in way that is competitive with the ecosystem of Chinese open-source companies,” said Hemant Taneja, chief executive of General Catalyst, the venture capital firm that is leading the start-up’s funding round. “Hopefully, this is America’s answer.”
Over the past several years, companies like Anthropic and OpenAI have spent billions of dollars creating the world’s most powerful A.I. technologies. But they maintain strict control over how these technologies operate and how people use them.
Executives at these companies argue that A.I. is too dangerous to be handled in any other way. But others worry that this approach will give Anthropic and OpenAI enormous power over the wider tech industry and perhaps even the global economy.
Critics of those closed systems also argue that if companies like Anthropic and OpenAI limit access to the most powerful technologies, everyday businesses will not have the tools they need to fight cyberattacks.
They also contend that the philosophy favored by Anthropic and OpenAI will stifle innovation across the industry. Innovation happens far more quickly, they argue, when companies share their technologies as open-source software, allowing anyone to use and modify them for free.
Companies like Meta and xAI have offered open-sourced technologies in the past. But in recent months, they have not kept pace with the leading systems, including the technologies emerging from China.
“What will it take for the American ecosystem to have an open-source infrastructure?” Mr. Taneja asked. “We don’t think anyone is doing that well.”
Mr. Babuschkin, who has also been a key researcher at OpenAI and Google, said he wanted to build a company that was fundamentally geared toward the open-source philosophy.
“The way that the big A.I. companies like Anthropic and OpenAI are building A.I. may not be the way of the future,” he said. “Increasingly, this is just not something the public is behind. We don’t want these A.I. companies to rule the world and control this superpowerful technology.”
(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news content related to A.I. systems. The two companies have denied those claims.)
The post His Start-Up’s Goal: A.I. That Is Trainable and Not Controlled by a Big Company appeared first on New York Times.
