news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Smart Things
Compiled by Cheng Qian
Edit Heart

In just 3 months, "AI Godmother" Fei-Fei Li created a unicorn.

Zhidongxi reported on July 17 that just now, according to the Financial Times, citing people familiar with the matter, the "spatial intelligence" startup founded by the famous Chinese computer scientist Fei-Fei LiWorld Labs' valuation has surpassed $1 billionThe startup mainly uses human-like visual data processing technology to enable AI to have advanced reasoning capabilities.

According to two people familiar with the matter, since its establishment in April this year, World Labs hasTwo rounds of financing, with investors including top tech investor Andreessen Horowitz and AI fund Radical Ventures. It is understood that the latest round of financing may reachApproximately US$100 million。

Li, Andreessen Horowitz and Radical Ventures did not respond to requests for comment.

According to data from PitchBook, a world-renowned data analysis agency, investors have invested more than $27 billion in U.S. AI startups in the past three months, accounting for about half of all startup financing in the same period. World Labs is the latest AI startup to receive a large investment.

Fei-Fei Li rose to fame in the field of AI and was called the "godmother of AI" for developing the large-scale image dataset ImageNet. This image dataset helped create the first generation of computer vision technology that could reliably identify objects.

Fei-Fei Li is the first Sequoia Professor of Computer Science at Stanford University and co-director of the Stanford Human-Centered AI Institute (HAI). She led Google Cloud's artificial intelligence business from 2017 to 2018, served as a board member of social platform X from 2020 to 2022, and currently serves as an advisor to the White House Artificial Intelligence Task Force. Among them, the Human-Centered AI Institute was founded by Fei-Fei Li during her partial sabbatical at Stanford University in 2019, with the aim of using emerging technologies to improve the human condition.

World Labs will attempt to create AI by developing human-like visual data processing technology.“Spatial Intelligence”。

In April this year, Fei-Fei Li gave a TED talk on "spatial intelligence" in Vancouver, describing the potential of machines to understand and navigate three-dimensional space. The related cutting-edge research involves an algorithm that can reasonably infer what images and text will look like in a 3D environment and take actions based on these predictions. (Top AI scientist Fei-Fei Li's speech: Robot evolution is inseparable from spatial intelligence)

To illustrate the idea, she showed a picture of a cat with its paw extended, pushing a glass against the edge of a table. Li said the human brain can assess “the geometry of the glass,” “its position in 3D space,” “its relationship to the table, the cat, and everything else” in a split second, then predict what will happen and take steps to prevent it.

“Nature has created this virtuous cycle of seeing and doing, fueled by spatial intelligence,” she explains.

She believes that large libraries of labeled images have been crucial to recent AI breakthroughs, such as training self-driving cars to navigate their environments and training AI models to correctly identify objects based on visual cues.

Fei-Fei Li's vision for space intelligence is ambitious: She wantsTraining a machine to understand the complex physical world and the relationships between objects in it。

“[WorldLab] is developing a model for understanding the three-dimensional physical world; essentially the dimensions of objects, where things are and what they do,” said a venture capitalist familiar with Li.

Conclusion: Simulating human visual processing helps AI understand the physical world

When it comes to breakthrough AI emergence, some researchers believe that reasoning capabilities can be improved by building larger and more complex existing models; others believe that the path forward involves using new "world models" that can obtain visual information from the physical environment around them to develop logical abilities and replicate the way infants learn.

Some of the AI ​​startups that are attracting investor interest are also developing intelligent robots that can understand and manipulate the physical environment. Skild AI, for example, is building a "universal brain for all kinds of robots" and was valued at $1.5 billion last week after raising $300 million from SoftBank, Amazon founder Jeff Bezos' investment fund, Lightspeed Venture Partners and others.

The "space intelligence" project that Fei-Fei Li focuses on is also a typical example.

Source: Financial Times