SAN FRANCISCO — A new startup is called Recursive, with an “e.” Another is called Ricursive, with “i”. Both try to do the same thing: build artificial intelligence capable of improving itself without the help of humans, an obsession of Silicon Valley technologists for decades.
Ricursive Intelligence, based in Palo Alto, California, works on the specialized computer chips that power today’s chatbots. Founded by two former Google researchers, Anna Goldie and Azalia Mirhoseini, Ricursive aims to develop AI systems capable of improving the design of these extremely complex chips.
Also read:
Continues after advertising
If AI systems can produce better chips, they argue, those chips will produce better AI systems. And then the process will repeat itself over and over again, as technology evolves further and further.
“The idea of a self-improvement cycle that feeds back is what inspires us,” said Goldie, who did similar work with Mirhoseini at Google.
Ricursive has raised $335 million from venture capital firms including Sequoia, Radical Ventures, Lightspeed and DST Global. Although it is less than a year old and has fewer than 10 employees, the company is valued at US$4 billion.
The company is among several new AI startups that have raised huge amounts of money in recent months. Just last week, Humans&, founded in San Francisco by former researchers from laboratories such as Anthropic and Elon Musk’s xAI, raised US$480 million. The company is only three months old and is valued at US$4.48 billion.
Even as many financial analysts and industry insiders warn of an AI bubble, gigantic volumes of money continue to flow into the field. This is in part because the raw computing power required to develop AI technologies is extremely expensive.
If investors want to bet on a new idea, hundreds of millions of dollars have increasingly become the minimum entry fee to play the AI game.
Continues after advertising
“Recursion” is a term commonly used by mathematicians and computer programmers. It refers to a mathematical function or procedure that feeds on itself. After a procedure generates some information, it uses that information to generate something else. This process can continue indefinitely.
This mathematical idea has inspired AI researchers for decades. Instead of just building a mathematical function that feeds on itself, they seek to create an AI system that feeds on itself.
In 2017, as the latest wave of development in AI began to gain traction, Google created a technology called AutoML. ML was the acronym for “machine learning”, which refers to computer algorithms that learn skills by analyzing data.
Continues after advertising
With AutoML, Google took this idea a step further: it built a machine learning algorithm that learned to create other machine learning algorithms.
At OpenAI, creator of ChatGPT, researchers are developing what they call an “automated AI researcher.” Until autumn (in the northern hemisphere, between September and December) they hope to have a system capable of doing the work of a less experienced researcher, before gradually improving the technology, said company CEO Sam Altman.
This is similar to the goal of another new startup, Recursive AI, founded by Richard Socher, who oversaw AI research at cloud computing giant Salesforce.
Continues after advertising
The startup has not yet announced itself publicly, but its mission is already a topic of discussion in Silicon Valley’s highly connected community of AI researchers.
Recursive AI is also valued at $4 billion, according to a person familiar with its most recent fundraising round who spoke on condition of anonymity. The news was initially released by Bloomberg.
While technologies as old as Google’s AutoML have demonstrated that AI can help improve AI itself, these efforts are still a long way from a future in which humans can be taken out of the process, said Div Garg, CEO of AGI, a San Francisco startup working on developing increasingly intelligent computing technologies.
Continues after advertising
“They work well for very specific tasks,” he said.
At Google, Goldie and Mirhoseini developed AI technology capable of improving the design of the company’s internal computer chip. Called a tensor processing unit, or TPU, the chip was designed to develop and run AI technologies.
Now Ricursive plans to help other companies improve their chips in a similar way. And, over the years, his ultimate goal is to create a virtuous circle in which chips and AI evolve side by side.
“The first phase of the company is just accelerating chip design,” Goldie said. “But if we have the ability to design chips very quickly, why not use it ourselves? Why not build our own chips? Why not train our own models? Why not have them co-evolve?”
c.2026 The New York Times Company
