The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

In this episode, Justin Johnson, co-founder of World Labs, joins us to discuss world models and the emerging field of spatial AI. We explore why many researchers see capabilities beyond language as an important frontier for AI, and what it means to build models that can understand, generate, and simulate the environments around them.

Justin explains the different approaches to world modeling, including explicit 3D representations and generative models, and why there is still no established recipe for building these systems. We also discuss World Labs’ Marble system, which can generate navigable 3D worlds from images and other inputs, the challenges of evaluating world models, and the role of simulation, planning, and action. Finally, Justin shares his vision for models that bring these capabilities together, supporting everything from interactive virtual environments to agents and robots that can operate in the physical world.

🗒️ Full show notes: ⁠https://twimlai.com/go/775.

The conventional wisdom in AI is that the next breakthrough will come from more compute, more data, and larger models. But what if the next leap comes from somewhere else?

In this episode, Max Welling—co-founder and CTO of CuspAI and professor at the University of Amsterdam—argues that physics may provide some of the ideas behind the next generation of AI systems.

We begin with CuspAI’s work using generative AI to design entirely new materials for semiconductors, batteries, carbon capture, and clean energy. Max explains how foundation models for chemistry, agentic workflows, simulation, and automated experimentation are dramatically accelerating the search for new materials and reshaping scientific discovery.

The conversation then broadens into a deeper question. Beyond giving AI new scientific problems to solve, can physics also teach us how to build better AI? Max explores surprising connections between machine learning and thermodynamics, why waves may become a new computational primitive for neural networks, and how concepts like symmetry breaking and statistical physics could inspire AI architectures beyond today’s scaling paradigm.

🗒️ Full show notes: https://twimlai.com/go/774.

Text-to-image models have become remarkably good at producing realistic images. But realism isn’t the same as correctness. Ask for several distinct people, a specific composition, or a high-resolution image generated locally, and today’s models still struggle in surprising ways.

In this episode, Fatih Porikli, Vice President of Technology at Qualcomm, joins me to discuss what remains unsolved in image generation and several approaches his team presented at CVPR to address those challenges. We explore why better training objectives can improve controllability, how separating scene planning from rendering may lead to more reliable image generation, techniques for generating 16-megapixel images efficiently on edge devices, and new methods for eliminating the visible artifacts that often appear in AI-powered image editing.

Along the way, we discuss reinforcement learning for image generation, agentic image generation pipelines, on-device AI, and what the next phase of progress in generative vision systems is likely to look like.

🗒️ Full show notes: https://twimlai.com/go/773

For more than a decade, AI has advanced by training ever-larger models on ever-larger datasets. But as high-quality training data becomes harder to find and pretraining grows increasingly expensive, researchers are looking for new ways to keep foundation models improving.

In this episode, Damian Borth, professor of AI and machine learning at the University of St. Gallen, argues we’ve been overlooking an important source of knowledge: the models we’ve already trained. His group’s work on weight space learning treats trained neural networks themselves as data, learning from the distilled results of millions of GPU hours of optimization rather than starting from raw data each time.

We explore what it means to build foundation models of neural networks, how knowledge can be transferred across architectures and domains, why this approach could dramatically reduce the cost of developing specialized models, and whether future AI systems may be trained on collections of existing models instead of ever-growing datasets.

🗒️ Full show notes: https://twimlai.com/go/772.

In this episode, Alex Wiltschko, founder and CEO of Osmo, joins the show to discuss his goal of giving computers a sense of smell and what it takes to build olfactory intelligence.

We explore the science behind smell, from the hundreds of olfactory receptors in the human nose to the challenge of mapping the relationship between molecular structure and odor, ensuring safety regulations are met, and building foundation models for smell. Alex explains how graph neural networks and advanced embedding spaces allow AI to capture the multi-dimensional structure of scents, grouping them into perceptual neighborhoods, and creating a machine learning representation that predicts how molecules smell.

We also cover how Osmo built the largest proprietary olfactory dataset from scratch to train a fleet of predictive models, and how olfactory intelligence could eventually power applications far beyond fragrance, including disease detection, emotion sensing, and consumer devices.

🗒️ Full show notes: https://twimlai.com/go/771.