💰 AI pioneer Fei-Fei Li's World Labs raises $1 billion in funding ↗
World Labs - founded by Fei-Fei Li - just landed a huge funding round to push “spatial intelligence”: AI that can interpret and generate the 3D world, not only language and flat images.
The investor roster is loud: AMD, Nvidia, Autodesk, and others, with Autodesk putting real heft behind the effort (and advising). The argument is simple enough: 3D-native models could unlock sharper AR/VR and more capable robotics - which feels inevitable, and still devilishly difficult in practice.
🧠 British researcher raising $1bn to build superhuman intelligence ↗
David Silver (ex-DeepMind, central to AlphaGo) is reportedly launching a new London-based lab and trying to raise a gigantic round straight out of the gate. The rumored valuation is equally gigantic - the kind of number that makes you blink twice.
The core thesis is that AI is running into “data ceilings” and needs to learn more through interaction with the world (or high-fidelity simulations of it) to keep improving - not just keep vacuuming up the internet indefinitely. That line of reasoning tracks, at least on its face.
🧩 Nvidia Wins Meta Deal. Its Gain Is Broadcom's Loss. ↗
Meta is doubling down on Nvidia for a major chunk of its AI infrastructure - GPUs plus networking - tightening Nvidia’s grip on the “who gets to train the biggest models” supply chain. The scale is described in those slippery “tens of billions” terms, but the direction is unambiguous. (Barron's)
The sharper edge is the competitive subtext: Meta has been weighing alternatives, and this swing suggests Nvidia’s full-stack pitch (hardware + software + ecosystem glue) remains profoundly sticky. Broadcom and others have no reason to enjoy that narrative.
⚖️ UK SUPREME COURT REWRITES RULES ON AI AND COMPUTER PROGRAM PATENTS ↗
The UK Supreme Court handed down a decision in Emotional Perception AI Ltd v Comptroller General that’s being framed as a major shift in how UK law treats patent exclusions for computer-implemented inventions - especially the “computer program… as such” bit.
Translation: the boundary lines around what you can patent, when software and AI are involved, just got re-argued at the highest level. If you’re building AI tech in the UK, your lawyers are either celebrating or stress-eating, possibly both.
🌍 Tech billionaires fly in for Delhi AI expo as Modi jostles to lead in south ↗
A major Delhi AI expo is drawing heavyweight tech leaders alongside “global south” political figures, with plenty of subtext about who sets the rules - and who ends up dependent on whose models, chips, clouds, and standards.
It’s part summit, part power theatre: countries want the upside of AI without getting locked into lopsided relationships. The tone is familiar: everyone smiles for the cameras while quietly bargaining for leverage. Classic.
FAQ
What is World Labs and what does it mean by “spatial intelligence”?
World Labs, founded by Fei-Fei Li, is positioning “spatial intelligence” as AI that understands and can generate the 3D world, not just text or flat images. In practice, that implies models that can reason about objects, geometry, and physical space. The pitch is that 3D-native systems could be a step-change for applications where depth, motion, and physical constraints matter.
Why are companies like AMD, Nvidia, and Autodesk investing in World Labs?
The investor list suggests “spatial intelligence” sits at the intersection of compute, tooling, and deployment. Chip companies benefit if 3D-native training and simulation drive demand for higher-end hardware and networking. Autodesk’s involvement signals interest from the design and engineering software world, where 3D data is foundational. Strategically, backing a new model paradigm can also help investors shape - and gain leverage within - future ecosystems.
How could 3D-native AI improve AR/VR and robotics?
The argument is that AR/VR and robotics are bottlenecked by robust understanding of space, not just recognizing pixels. If models can reliably interpret scenes in 3D, AR overlays can anchor more precisely and feel less “floaty,” while robots can plan and manipulate more effectively. It’s described as inevitable but difficult, because physics and perception in physical environments are noisy and packed with edge cases.
Why are some AI labs saying we’re hitting “data ceilings”?
The idea is that simply training on more internet text and images delivers diminishing returns over time. If the most valuable public data is already consumed, progress may slow unless systems learn through interaction, experimentation, or simulation. In many pipelines, that means reinforcement learning, embodied tasks, or high-fidelity virtual environments. It’s a shift from “read everything” to “learn by doing.”
Who is David Silver and what is the new London lab reportedly trying to do?
David Silver, associated with DeepMind and AlphaGo, is reportedly launching a London-based lab and attempting to raise a very large funding round early. The reported thesis is about pushing past data limits by learning through interaction with the world or via simulations. “Superhuman intelligence” language tends to be aspirational, but the operational focus sounds like more agentic, environment-driven training approaches.
What is the Nvidia–Meta deal described as signaling about AI infrastructure?
The story frames Meta as doubling down on Nvidia for a major slice of AI infrastructure, including GPUs and networking. The implication is that Nvidia’s “full-stack” approach - hardware plus software and ecosystem - remains sticky at large scale. It also carries a competitive subtext: if big buyers keep choosing Nvidia, alternatives face a tougher path to win the most demanding training workloads.
How does the UK Supreme Court decision affect AI and software patentability?
The decision in Emotional Perception AI Ltd v Comptroller General is being described as materially changing how the UK interprets patent exclusions for computer-implemented inventions, especially the “computer program… as such” boundary. For AI startups, that can alter how confidently they file, draft, and defend patents in the UK. The practical takeaway is that the legal line shifted, so counsel will revisit strategies.
If I’m building AI in the UK, what should I do differently after this ruling?
A common approach is to review existing filings and future claims with patent counsel to ensure they align with the updated interpretation. Teams may also adjust how they describe the technical contribution - what the invention does beyond being “just software.” It’s also a good moment to audit IP plans across jurisdictions, since the UK and other regions may treat similar concepts differently.
Why is the Delhi AI expo being framed as geopolitics, not just tech?
The expo is portrayed as a gathering of tech leaders and “global south” political figures where influence and dependency are central themes. Beyond demos, the subtext is who controls models, chips, clouds, standards, and the rules that govern them. Countries want AI’s benefits without becoming locked into lopsided relationships. So the event reads like diplomacy with GPUs in the background.