There’s an attention-grabbing new entrant in the competitive world of AI startups. Computer scientist and Stanford University professor Fei-Fei Li, known as the “godmother” of AI, is building a new company “that uses human-like processing of visual data to make AI capable of advanced reasoning,” according to a Reuters report.
The report says the startup’s investors include Andreessen Horowitz and Radical Ventures, which has also funded LLM model builder Cohere, self-driving car company Waabi, and AI search engine You.com, and which Li joined as a scientific partner last year. The new startup’s name, funding amount, and valuation have not yet been made public.
Li is co-director of the Stanford Institute for Human-Centered AI and was formerly chief scientist of AI and machine learning at Google. But arguably her biggest claim to fame was the 2006 debut of ImageNet, a huge collection of labeled images—like “dogs” or “cars”—that became essential to teach computers to recognize objects in photos. Six years later, neural network research based on the ImageNet dataset, led by AI “godfather” Geoffrey Hinton, helped jump-start the deep learning revolution and led directly to today’s generative AI boom.
At a TED talk in Vancouver last month, Li reportedly said advances in “spatial intelligence,” or the ability to predict and derive insights from objects’ relationships to each other—like AI that can recognize 3D structures, or robots that can navigate their environment—will enable robots and computers to learn from the real world. While a representative from Radical Ventures told Fortune “we are not commenting at this time,” it’s clear that Li’s reputation, as well as the intense industry interest in spatial computing (considered an essential stepping stone towards human-level AI) makes this startup investment enticing.
Of course, the new company will be up against many others in the race towards increasingly human-like intelligence, including the potential for artificial general intelligence, or AGI. Most experts say AGI remains a long-term goal that will still take several years—or even many decades—to achieve.
Still, it’s interesting that Li is jumping into the profit-seeking startup space, describing herself on her LinkedIn as a “newbie.” It’s in sharp contrast to her lobbying over the past couple of months of DC leaders to make sure that top AI academic researchers can keep up with industry giants.
Last month, the Washington Post reported that Li asked President Joe Biden to fund a “moonshot investment” in computing power and data sets to make sure the high cost of AI models does not compromise the independent study of the technology.
“The public sector is now significantly lagging in resources and talent compared to that of industry,” said Li in the Washington Post interview. “This will have profound consequences because industry is focused on developing technology that is profit-driven, whereas public-sector AI goals are focused on creating public goods.”
Considering her track record of giving birth to the deep learning foundations that underpin today’s generative AI models, it’s no surprise that Li’s new company is getting attention. It remains to be seen, however, whether she’ll be able to add to her already considerable success in the very difficult startup realm.
More news below.
Sharon Goldman
[email protected]
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
Today’s edition of Data Sheet was curated by Rachyl Jones.
NEWSWORTHY
Investing in twins. The Biden administration is allocating $285 million in Chips Act funding to establish research on digital twins, among other topics, TechCrunch reported. Digital twins, which are virtual models of physical objects and systems used for testing, could reduce the cost for developing and manufacturing chips, director of the National Institute of Standards and Technology Laurie Locascio said. Access to these semiconductor chips, which are necessary to develop artificial intelligence, is a top priority for government officials as it competes with China in tech innovation.
Robots for all. Hugging Face, known for its massive library of open-sourced AI models, has launched a collection of models, datasets, and other tools to help individuals build robots, VentureBeat reported. Former Tesla scientist Remi Cadene, who joined Hugging Face in March, created the toolkit, called LeRobot.
Momenta IPO. General Motors-backed Momenta, a Chinese autonomous driving company, has confidentially filed for an initial public offering in the U.S., Bloomberg reported, citing sources familiar with the matter. Momenta could reportedly raise $200 million to $300 million from selling shares. GM invested $300 million in the company in 2021.
Another OpenAI deal. OpenAI has partnered with Stack Overflow, the question and answer site for software developers, to improve how its large language model answers questions about programming, TechCrunch reported. Fifteen years of responses from developers on the Stack Overflow platform will train OpenAI’s ChatGPT to answer software and coding questions. After originally banning ChatGPT responses on its site, Stack Overflow will also begin integrating AI into its platform.
ON THIS DAY IN TECH HISTORY
May 6, 1998
At an event, Steve Jobs announced Apple’s first major consumer product, the iMac, since his return as the company’s CEO the year prior. Marketed as a way to reach the internet quickly, the iMac was an instant success, and more than 25 years later, it is still a core product for Apple.
IN CASE YOU MISSED IT
Bought a Model Y? Tesla may charge you to use the battery’s full range, by Chris Morris
Elon Musk has turned Tesla into a meme stock as he tells Wall Street to value the EV maker like an AI company, top economist says, by Jason Ma
An AI-controlled fighter jet with the Air Force secretary onboard flew against a human pilot in a secret dogfight: ‘It’s a security risk not to have it’, by the Associated Press
Democrats race to design ways to best use AI in 2024 campaign after Trump and allies outmaneuvered them on social media in 2016, by the Associated Press
Warren Buffett saw an AI deepfake video of him that was so good he says he could imagine it tricking him into sending money overseas, by the Associated Press
Apple’s China sales tell a different story than what analysts have heard for months. Where is the disconnect?, by Rachyl Jones
BEFORE YOU GO
Robinhood’s SEC troubles. The U.S. Securities and Exchange Commission issued a Wells Notice, or a warning that it plans to initiate enforcement action, to popular stock trading platform Robinhood over its crypto business, CoinDesk reported. The SEC is alleging violations of its rules around processing securities transfers and registering people and firms exchanging securities.
In a response, Robinhood executive Dan Gallagher said the assets available on Robinhood are not securities. The trading platform previously delisted tokens named in the SEC lawsuits against Binance and Coinbase.