news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

New Intelligence Report

Editor: Ear Aeneas

【New Wisdom Introduction】The "AI Godmother" started a business and created an AI unicorn with a valuation of 1 billion in just 3 months.

The "spatial intelligence" startup company founded by AI godmother Fei-Fei Li was valued at $1 billion in just over three months.

Just now, people familiar with the matter revealed that World Labs has raised two rounds of funds from top technology investors such as a16z and artificial intelligence fund Radical Ventures.

They value World Labs at more than $1 billion.

In the latest round of financing, World Labs has raised approximately $100 million.

World Labs will probably soon enter the fray and join the fierce competition for commercialization in the entire technology industry.

Silicon Valley investors are optimistic about Fei-Fei Li

World Labs is the latest AI startup to receive huge investment following OpenAI’s release of ChatGPT in November 2022, showing that investor interest in GenAI is still surging.

In the past three months alone, investors have poured more than $27 billion into U.S. AI startups, according to PitchBook. That’s half of all startup funding in that period.

Just three months ago, the news that Fei-Fei Li, a famous Chinese-American artificial intelligence scientist and professor at Stanford University, founded a new company caused quite a stir in the industry.

This is also the first entrepreneurial project she has directly participated in since leaving Google and returning to Stanford in 2018.

Take a vacation and create a $1 billion unicorn

The "AI Godmother"'s emergence and founding of a company were foreshadowed at the beginning of this year.

According to her Stanford University profile, Fei-Fei Li will be on leave from early 2024 to late 2025.

Academic leave is an important institutional form of faculty development in American universities. Scholars have a leave of absence lasting six months to one year in a cycle of about four to five years, during which they can rest, serve as visiting scholars, engage in new research, etc.

As a result, it took only one academic leave for Fei-Fei Li to create an AI unicorn.

Now on LinkedIn, her work status has been changed to "something new".

Investors in the new company include a16z and Radical Ventures.

Fei-Fei Li is also one of the partners of Radical Ventures, an independent fund focusing on artificial intelligence and related innovations. It was registered in Canada and received investment from top scientists in the field of AI, such as Geoffrey Hinton and Fei-Fei Li, at the beginning of its establishment.

This entrepreneurial venture can also be said to be the implementation and practice of Fei-Fei Li's academic philosophy.

World Labs: Targeting the concept of spatial intelligence

In describing the startup, one person familiar with the matter cited a TED talk given by Fei-Fei Li in Vancouver in May.

In her speech, she said that her current cutting-edge research involves an algorithm that can reasonably infer what images and text look like in a three-dimensional environment and take actions based on these predictions.

The algorithmic concept of allowing computers to learn how to act in the real three-dimensional world is called "spatial intelligence."

In the field of AI, spatial intelligence refers specifically to the machine's ability to understand and process the spatial environment, enabling the machine to simulate these spatial cognitive functions of humans.

To illustrate this idea more intuitively, Fei-Fei Li showed a vivid picture in her speech: a cat stretched out its claws and tried to push a glass to the edge of the table.

“The human brain is able to assess the geometry of the glass, its position in three-dimensional space and its relationship to the table, the cat and other objects in a split second, then predict what will happen next and take steps to prevent it,” she explains.

Fei-Fei Li summarized the ultimate vision of spatial intelligence in her TED Evolution, "bringing us closer to a world where AI can not only see and create, but also interact with the physical world around it."

Allowing computers to "see" like humans is also a footnote to Fei-Fei Li's many years of research in artificial intelligence. Her autobiography is also titled "The World I See."

Last November, the laboratory led by Fei-Fei Li released an intelligent robot called "VoxPoser".

In the study, the general language big model was combined with computer vision, robotics and other technologies, allowing robots to directly understand human natural language instructions by accessing the big model and convert these complex instructions into specific action plans.

The "AI Godmother" will naturally not be satisfied with just creating a robot, but will set her sights on establishing a company that implements her own ideas.

Through her startup, Li aims to teach algorithms common sense to overcome the limitations of current technology, such as the hallucinations and false information that LLMs can generate.

It is crucial to build this “reasoning” capability before AI models can achieve AGI.

Researchers have different views on how to improve reasoning capabilities. Some believe that this can be achieved by building larger and more complex language models;

Others believe that future breakthroughs in artificial intelligence will lie in the creation of "world models" that can absorb visual information from the surrounding physical environment and then develop logic, imitating the way humans learn.

"World Labs" may be the answer given by Fei-Fei Li to this question.

Let robots take out the trash for humans

This area is indeed attracting great interest from investors, and some companies are developing intelligent robots that can understand and manipulate the physical environment around them.

For example, Skild, which is developing a "universal brain for all kinds of robots," last week raised $300 million from SoftBank, Amazon founders' investment fund, Lightspeed Venture Partners and others, valuing the company at $1.5 billion.

Despite these impressive achievements, Fei-Fei Li has a bigger vision for spatial intelligence: to train an AI system that can understand the complex physical world and the relationships between objects in it.

A venture capitalist familiar with Fei-Fei Li’s work said that the model developed by “World Labs” can understand the three-dimensional physical world and the location and function of things in space.

Therefore, this robot can successfully complete daily tasks such as "sweeping garbage into a dustpan" and "opening drawers and paying attention to avoid nearby vases" that are directly assigned by humans verbally.

The beginning of academic career: ImageNet

As a professor at Stanford University, she has achieved remarkable research results in the field of computer vision, won many international awards, and successfully transformed academic research into practical applications.

This leap from theory to practice shows that Fei-Fei Li not only possesses outstanding scientific research capabilities, but also has the foresight to commercialize technology.

As early as 2005, Fei-Fei Li, who was less than 30 years old and had just received a Ph.D. in electrical engineering from the California Institute of Technology, launched the ImageNet project with one of her graduate students.

ImageNet made her mark in the field of artificial intelligence. This project marked the beginning of her academic career and her most important contribution to the artificial intelligence industry to date.

Since then, she and her students launched the ImageNet Challenge, which has a history of 14 years since 2010 and is held every year. It has become the most watched top event in the field of computer vision.

Making AI people-centric

Fei-Fei Li wants the victory of AI to be not only a victory for science, but also a victory for humanity.

Scientists, including herself, have a responsibility to train AI to be "machines that follow good academic traditions, are willing to collaborate, and respect the opinions and expertise of others."

Therefore, "people-oriented" is also an important part of Fei-Fei Li's scientific research philosophy.

In 2019, Fei-Fei Li established the Human-Centered AI Institute at Stanford University, an interdisciplinary research institution with distinctive AI values ​​at its core.

In March 2024, in a conversation with Nvidia Chief Scientist Bill Dally, Fei-Fei Li once again clearly explained her views on AI, saying that AI should not be a privilege for a few people, and everyone must be involved.

Fei-Fei Li believes that she is not so much studying machines as she is studying the origin of the human mind.

She also firmly believes that humans are the creators of artificial intelligence and the objects of machine learning and imitation, and have abilities and values ​​that can never be replaced by machines.

The "AI Godmother" has not only achieved numerous scientific research results, but also intersected the two parallel lines of scientific rationality and humanistic care. How will World Labs continue to interpret her philosophy?

References:

https://www.ft.com/content/0b210299-4659-4055-8d81-5a493e85432f

https://www.xm.com/fr/research/markets/allNews/reuters/stanford-ai-leader-feifei-li-building-spatial-intelligence-startup-53830004