Thứ Hai, 28 tháng 9, 2026

To Seek a Newer World - Fei-Fei Li | 2026.09.09

 Source: https://x.com/drfeifei/status/2104664403681632453

To Seek a Newer World by Dr. Fei-Fei Li

Copy:

To Seek a Newer World

World Labs is joining . This is a huge moment for , our team, and for me, and I wanted to take a moment to share what this means and why I’m so excited for this next chapter.
“Come, my friends, ‘Tis not too late to seek a newer world.” – Lord Tennyson, Ulysses

A coming of age story

AI to me is both personal and civilizational.
I entered the field as a fresh graduate student in the year 2000, chasing my curiosity for solving the problem of visual intelligence for machines and the brain. I didn’t know it then, but these were the first steps of a career that’s been (to my amazement) at the center of AI’s coming of age. The early work from my lab called ImageNet helped usher in the age of modern AI, along with neural network algorithms and their implementation on GPUs. Since then, I’ve spent my entire career pushing AI from largely an academic curiosity to become the most innovative and transformative technology of our era.
Throughout this time, from my research in computer vision, deep learning, robotics and healthcare, to founding Google Cloud’s AI unit as a chief scientist, to becoming the founding director of Stanford’s Human-Centered AI institute, my work has always been guided by the desire to solve real world problems, hand in hand with humans for societal betterment. When generative AI began its transformation, and frontier AI research at scale moved from academia to industry, I knew the path with the deepest and broadest potential for positive impact would be creating a startup to tackle the most audacious technical problems.
Together with my incredible co-founders Ben Mildenhall and Justin Johnson, two world class AI researchers who have each moved the field in significant ways, I founded World Labs at the beginning of 2024. Our goal was to build a frontier lab focusing on some of the most important challenges in AI, which if solved, would dramatically improve AI’s ability to tackle real world problems.

The world is not just made up of words

We built the company guided by the foundational belief that language alone isn't sufficient for model development. The universe isn’t made up of words, it’s made of real things. So many of the most important problems for AI to solve, from science to entertainment to robotics, require models to reason over the structure and behavior of the physical world. Tackling this problem domain requires a unique team with wide ranging skills including data, computer vision, large model training, systems and even computer graphics. I feel very lucky to have built a founding team of researchers and engineers that contains world experts across these areas. Together, we set out to train new classes of foundation models to tackle the hardest problems in spatial intelligence, and we’ve made significant strides toward our goals.
Since our founding in 2024, World Labs has built leading AI spatial intelligence capabilities for everything from creative work to design. We built the world leading model training team for images, video and spatial reconstruction. And, with the acquisition of SceniX, we’re building towards an industry leading capability for robotics simulation.
Recently we released Atlas, a first of its kind omni model architecture that solves a key outstanding problem in spatial intelligence: new camera view prediction. Like LLMs can predict the next token from a line of text, Atlas, trained from scratch, can predict the next view from an input of 2D images, outperforming state of the art results even by specialized models. It has essentially solved a long standing problem in computer vision called sparse reconstruction, by combining generative models with multiview geometry. This has direct and far reaching consequences: from design and engineering to science and robotics.
We’ve seen incredible interest in Atlas across many domains: RL environments for robotics; scene generation for therapy and entertainment; and real world reconstruction for real estate, design and construction, and so much more to come.

Following our North Star to AMD

The research and technical breakthroughs we achieved at World Labs in such a short amount of time have only fueled our ambition and given us a clear vision for the potential for AI to solve spatial and physical world problems. Now that we have tangible proof of the possibilities, we want to do everything we can to accelerate the future.
To do this requires scaling our efforts, widening our reach, and getting closer to the hardware. Without having a focused hardware effort, AI is hobbled in efficiency. And scale. And for our purposes, remains trapped in the digital world.
Which brings me to AMD. We’ve partnered with AMD since our earliest days. AMD’s CEO, Dr. Lisa Su was an early investor and believer in our mission, and we shared a belief that AI could propel society toward a better future. She has also been a great friend to me as we each journey through the world of technology.
We began a deep technical partnership last year starting with model training and inference optimization on AMD GPUs. We’ve seen up close how Lisa is boldly leading AMD as a major player in the hardware ecosystem with incredible momentum. As our teams worked together, we realized what a natural fit it would be to bring together our AI ecosystem of software and hardware, foundation models and applications.
Just like I knew it was time to do a startup to jumpstart the work of creating foundation models beyond LLMs and achieve the most impact in AI, it’s again very clear that the right path for myself and World Labs is to continue our mission as a part of AMD. Together we’re committed to keep pushing the frontier of AI to solve real problems, and doing so openly with the community and the entire ecosystem. We’re unwavering in our mission, and even more excited to continue building a world leading frontier research organization that is durable for the decades ahead. And we are committed to providing the best open models and platforms for running AI end to end from hardware to software to models to the data platforms needed to fuel it all.

Solving real problems, for human betterment

We’re extremely thrilled to be part of a shared journey towards the future we want to build together; where powerful AI platforms and systems are developed and deployed to help solve the most fundamental problems that humanity faces: from drug discovery, to unleashing human creativity; from transforming learning to unlocking the next generation of robotics.
And I can’t be more excited to join as an Executive Vice President and Chief Scientist, working directly with Lisa and her team. My career has led up to this moment - to continue growing and running a frontier research team, within a leading AI company, as a scientist and a technology leader. With this expanded platform and opportunity, I feel a deep sense of humility, gratitude and hopefulness that we will build a future we believe in, guided by the same North Star that has always inspired me to build technology in service of humanity. I am beyond excited for the future and grateful for the opportunity for World Labs and AMD to build it together.
---

Fei Fei has a lovely reflection blogpost that hints at the main reasons:

Since our founding in 2024, World Labs has built leading AI spatial intelligence capabilities for everything from creative work to design. We built the world leading model training team for images, video and spatial reconstruction. And, with the acquisition of SceniX, we’re building towards an industry leading capability for robotics simulation.

Recently we released Atlas, a first of its kind omni model architecture that solves a key outstanding problem in spatial intelligence: new camera view prediction. Like LLMs can predict the next token from a line of text, Atlas, trained from scratch, can predict the next view from an input of 2D images, outperforming state of the art results even by specialized models. It has essentially solved a long standing problem in computer vision called sparse reconstruction, by combining generative models with multiview geometry. This has direct and far reaching consequences: from design and engineering to science and robotics.

We’ve seen incredible interest in Atlas across many domains: RL environments for robotics; scene generation for therapy and entertainment; and real world reconstruction for real estate, design and construction, and so much more to come.

---
Cecile G. Tamura' post:
> You've probably heard of these three people:
> Fei-Fei Li. Creator of ImageNet—14 million labeled images.
In 2012, AlexNet won on it, and deep learning truly turned the corner.
> Justin Johnson. Her Stanford PhD student. In 2016, they made style transfer real-time—nearly 1,000 times faster than prior methods. He later joined the faculty at the University of Michigan before returning to co-found a company with his advisor.
> Ben Mildenhall. First author of the landmark NeRF paper.
In 2020, he used a small neural network to represent complex 3D scenes, spawning an entirely new subfield of computer graphics.
> In 2024, they joined forces to build World Labs, setting their sights on spatial intelligence.
> By late 2025, they launched Marble—a model capable of taking a single sentence or image and generating a fully interactive 3D world you can walk into.
> AMD, an early backer through its venture arm, saw where this was heading. Shortly after, AMD bought the entire team for $8.2 billion in an all-stock deal.
> Fei-Fei Li steps in as EVP and Chief Scientist at AMD, reporting directly to Lisa Su.
> In her own words: *To advance the next generation of AI, we must tie together model research, systems, and compute.*
> AMD didn't buy an app. They bought the minds defining what the next generation of physical AI models will look like—and put them right next to the silicon.
---

Fei-Fei Li Finally Sold

What's happening: Fei-Fei Li, the Stanford professor whose ImageNet dataset taught machines to see, left Google Cloud in 2024 to found World Labs and build spatial intelligence. AMD said Monday it will buy the lab for $8.2 billion in stock. Li becomes executive vice president and chief scientist, reporting to CEO Lisa Su. Her title before the detour: chief scientist of AI at Google Cloud.

How this hits reality: The title got longer. The job did not change. She left a chief executive's org chart and she rejoined one, with "executive vice president" bolted onto the same two words she already had. In between: a $1 billion round AMD helped fund, then an $8.2 billion exit AMD paid for, in stock. Two years, and the delivered artifact is a bigger business card. No product. No revenue. One announcement that names a title. ImageNet was handed to the field for free and the field is still standing on it. This one cost $8.2 billion and the field gets nothing to hold.

Key takeaway: Spatial intelligence has not shipped. Fei-Fei Li has.


Không có nhận xét nào:

Đăng nhận xét

US 2026 Midterms

Credit : Giang Công Thế. Credit : Gemini