Hot topic: Fei-Fei Li officially announces new AI company, Jen-Hsun Huang appears in new financing

Li Feifei's first entrepreneurial project, World Labs, announced that it had successfully raised US$230 million (approximately RMB 1.6 billion) in funds.

The financing was led by a16z, NEA and Radical Ventures, with participation from AMD, Adobe, Databricks’ venture capital arm and Shinrai Investments LLC. In addition, Nvidia’s founder Jensen Huang also participated in the investment.

The project has also received support from many outstanding individual investors, including LinkedIn co-founder Reid Hoffman, former Google CEO Eric Schmidt, Salesforce CEO Marc Benioff, Google AI Chief Scientist Jeff Dean, Turing Award winner Geoffrey Hinton, OpenAI founding member Andrej Karpathy, famous venture capitalist Jim Breyer, Silicon Valley angel investor Ron Conway, billionaire Ram Shriram, etc.

This is Fei-Fei Li's first entrepreneurial venture, and she chose the most difficult problem to solve in the field of artificial intelligence - spatial intelligence.

The official website shows that the company is committed to building a large world model (LWM) to perceive, generate and interact with the 3D world.

At the beginning of 2024, Fei-Fei Li applied for a two-year vacation to start her own business. In April this year, she founded World Labs and started her AI entrepreneurial journey.

Currently, the World Labs team has grown to 20 people, with Fei-Fei Li as co-founder and CEO;

The other three co-founders are Justin Johnson, Christoph Lassner and Ben Mildenhall, all of whom are well-known experts in the fields of computer vision and computer graphics.

World Labs has big ambitions. Its goal is to elevate AI models from the 2D pixel plane to a complete 3D world, including both virtual and real worlds, and to endow them with the same rich spatial intelligence as the human world.

World Labs revealed that the first phase will build a spatial intelligence model LWM with a deep understanding of three-dimensional, physical, and spatial and time concepts;

Next, the big model will support AR; after that, the big model will be used in robotics technology to improve self-driving cars, automated factories, humanoid robots and other fields.

Huang Renxun and Li Feifei have already established a deep friendship

Huang received his master's degree in electrical engineering from Stanford University in 1990;

More than ten years later, Fei-Fei Li also joined Stanford University and was promoted to tenured professor at the university in just three years.

In 2012, in the ImageNet competition led by Fei-Fei Li, the [AlexNet] project of the University of Toronto won the championship.

In order to cope with the huge computing needs, Alex, the project leader, adopted NVIDIA's GPU and used the CUDA programming interface to successfully achieve fast training and computing of convolutional neural networks.

Before 2012, Huang Renxun was worried about the marketing of GPU.

However, with the success of the competition, Huang Renxun's products became famous, and technology giants such as Microsoft, Google, and Amazon began to purchase Nvidia's GPUs.

With the further development of artificial intelligence technology, especially the launch of ChatGPT, NVIDIA, as a technology supplier, has gained huge economic benefits, and its market value once approached 3 trillion US dollars.

Therefore, some people praised Fei-Fei Li as the person Huang Renxun was most grateful to. Huang Renxun also provided indispensable support on Fei-Fei Li's AI research path.

In 2015, Fei-Fei Li founded the non-profit organization AI4ALL, which is dedicated to promoting AI education with a humanitarian spirit to various student groups, educational institutions and companies. Huang Renxun provided initial financial support for this.

Fei-Fei Li recently recalled Huang Renxun's generous support on social media: [I sincerely thank Jensen for his contribution to the field of artificial intelligence. In 2016, he donated the world's first NVIDIA DGX-1 to our lab. ]

Spatial intelligence will become the next frontier of computer vision

In the past two years, the rise of text-prompted image and video models, as well as LLMs, has heralded the potential of AI in the field of vision.

But they are just the tip of the iceberg.

In Fei-Fei Li's view, we need spatially intelligent AI that can model the world and reason about objects/places/interactions in 3D space and time.

In addition, spatial intelligence has strong creativity and can present the images in our minds in reality. It is with spatial intelligence that humans can reason, act and invent.

From simple sand castles to towering city visualizations, it's essential.

In a recent interview with Bloomberg, Fei-Fei Li said that human spatial intelligence has actually evolved over millions of years. It is the ability to understand, reason, generate, and even interact in a 3D world.

While current generative AI is language-based, Li sees a frontier in which systems build entire worlds using the rich details of physics, logic, and physical reality.

Spatial intelligence enables people to think in three dimensions, allowing them to perceive external and internal images, and to reproduce, transform or modify images, so that they can move freely in space and manipulate the position of objects at will to generate or interpret graphic information.

In a broad sense, spatial intelligence includes not only the ability to perceive spatial orientation, but also visual discrimination ability and image thinking ability.

For machines, spatial intelligence refers to their ability to process visual data in three-dimensional space, make accurate predictions, and take actions based on those predictions.

This capability enables machines to navigate, operate, and make decisions in a complex three-dimensional world just like humans, transcending the limitations of traditional two-dimensional vision.

[Nature has created a virtuous cycle of observation and action, which is enabled by spatial intelligence.] In her view, by enhancing AI's spatial reasoning capabilities, more complex interactions and navigation in the physical world can be achieved.

The development of spatial intelligence needs to be based on advanced technologies, including three-dimensional visual perception technology, spatiotemporal data processing and analysis, and deep learning.

These technologies will enable machines to perceive and understand three-dimensional space like humans, allowing them to make more intelligent decisions and actions.

Ending:

The era of spatial intelligence has arrived. The improvement of computing power, the deepening of data understanding and the evolution of algorithms have laid a solid foundation for the development of spatial intelligence.

Spatial intelligence goes hand in hand with linguistic intelligence, and is even more fundamental to some extent. It has become the inevitable direction of the development of artificial intelligence.

Some references: Founder Park: "AI Godmother" Fei-Fei Li's startup company officially announced" Investment Circle: "Huang Renxun invested in Fei-Fei Li", Titanium Media AGI: "Valuation exceeded 7 billion in 3 months, why did Fei-Fei Li's AI company become the world's fastest unicorn?", Chief Intelligence Officer: "AI Godmother Fei-Fei Li joins hands with World Labs to launch a 3D space intelligence revolution"

Recommended reading:

For business cooperation, please add WeChat:

18948782064

Please be sure to specify:

「Name + Company + Cooperation Requirements」