LLMs as World Models Research

On Friday’s show, the DAS crew discussed how AI is shifting from text and images into the physical world, and why trust and provenance will matter more as synthetic media gets indistinguishable from reality. They covered NVIDIA’s CES focus on “world models” and physical AI, new research arguing LLMs can function as world models, real-time autonomy and vehicle safety examples, Instagram’s stance that the “visual contract” is broken, and why identity systems, signatures, and social graphs may become the new anchor. The episode also highlighted an AI communication system for people with severe speech disabilities, a health example on earlier cancer detection, practical Suno tips for consistent vocal personas, and VentureBeat’s four themes to watch in 2026.

Key Points Discussed

CES is increasingly a robotics and AI show, Jensen Huang headlines January 5

NVIDIA’s Cosmos world foundation model platform points toward physical AI and robots

Researchers from Microsoft, Princeton, Edinburgh, and others argue LLMs can function as world models

“World models” matter for predicting state changes, physics, and cause and effect in the real world

Physical AI example, real-time detection of traction loss and motion states for vehicle stability

Discussion of advanced suspension and “each wheel as a robot” style control, tied to autonomy and safety

Instagram’s Adam Mosseri said the “visual contract” is broken, convincing fakes make “real” hard to assume

The takeaway, aesthetics stop differentiating, provenance and identity become the real battlefield

Concern shifts from obvious deepfakes to subtle, cumulative “micro” manipulations over time

Scott Morgan Foundation’s Vox AI aims to restore expressive communication for people with severe speech disabilities, built with lived experience of ALS

Additional health example, AI-assisted earlier detection of pancreatic cancer from scans

Suno persona updates and remix workflow tips for maintaining a consistent voice

VentureBeat’s 2026 themes, continuous learning, world models, orchestration, refinement

Timestamps and Topics

00:04:01 📺 CES preview, robotics and AI take center stage

00:04:26 🟩 Jensen Huang CES keynote, what to watch for

00:04:48 🤖 NVIDIA Cosmos, world foundation models, physical AI direction

00:07:44 🧠 New research, LLMs as world models

00:11:21 🚗 Physical AI for EVs, real-time traction loss and motion state estimation

00:13:55 🛞 Vehicle control example, advanced suspension, stability under rough conditions

00:18:45 📡 Real-world infrastructure chat, ultra high frequency “pucks” and responsiveness

00:24:00 📸 “Visual contract is broken”, Instagram and AI fakes

00:24:51 🔐 Provenance and identity, why labels fail, trust moves upstream

00:28:22 🧩 The “micro” problem, subtle tweaks, portfolio drift over years

00:30:28 🗣️ Vox AI, expressive communication for severe speech disabilities

00:32:12 👁️ ALS, eye tracking coding, multi-agent communication system details

00:34:03 🧬 Health example, earlier pancreatic cancer detection from scans

00:35:11 🎵 Suno persona updates, keeping a consistent voice

00:37:44 🔁 Remix workflow, preserving voice across iterations

00:42:43 📈 VentureBeat, four 2026 themes

00:43:02 ♻️ Trend 1, continuous learning

00:43:36 🌍 Trend 2, world models

00:44:22 🧠 Trend 3, orchestration for multi-step agentic workflows

00:44:58 🛠️ Trend 4, refinement and recursive self-critique

00:46:57 🗓️ Housekeeping, newsletter and conundrum updates, closing