Executive Code Podcast

AI that understands and generates speech directly. In this episode of Executive Code, Julian Salazar (Google DeepMind) joins Kirim to discuss WaveNet, Speech SSM, and what native speech means for the future of AI. youtu.be/h5fN55abB_c

Can light bouncing off a wall show what’s hidden from view? 🎧 Full episode on YouTube, Spotify, and Apple Podcasts. #Photonics #LiDAR #AI #ComputerVision #Imaging

Think of a laser hitting an object and bouncing back. That’s how we measure distance with light. 🎧 Full episode on YouTube, Spotify, and Apple Podcasts. #LiDAR #Photonics #AI #ComputerVision #Research

How can a camera measure single particles of light? 🎧 Full episode on YouTube, Spotify, and Apple Podcasts. #Photonics #LiDAR #AI #Optics #Research

Some AI models can predict efficiently but resist simple explanation. In this episode of Executive Code, Sanyam Agarwal and Oliver Broadrick discuss what happens when inference works—but representation fails. youtube.com/watch?v=qM9O6V…

Can AI learn to play it safe? In the new Executive Code episode, @Arnobg32 (NJIT) explains how risk-aware reinforcement learning adds risk budgets to make AI safer for robots, drones, cars, and beyond. 🎧 youtube.com/watch?v=UTiI-E…

What if you could rebuild the world just by tracking photons? @anagh_malik 🎧 Full episode on YouTube, Spotify, and Apple Podcasts. #Photonics #LiDAR #AI #ComputerVision #Research

That’s exactly what we dive into on the new Executive Code episode with @anagh_malik (PhD @ UofT, intern @ Apple). From capturing photons at the speed of light to seeing around corners — it’s sci-fi made real. 🔗youtu.be/YS4S_h_vIk4?si…

@BillyLincoln10 @emollick @billylincoln10 we’ve been diving into that on Executive Code too ! Bringing researchers on to break down their papers and opinions so people can actually get what’s behind all the AI. Check it out and tell us what you think!

What would be surprising is if LLMs *didn't* present the illusion of thought, since language is just a (lossy) wrapper of reality, and if it didn't have coherence, it wouldn't be useful. Our thoughts generate words. LLMs just invert the direction: they take the relationships between words to imply thought.

We really have not made a lot of progress on explaining the deep mystery of LLMs: How does a model using matrix multiplication to predict the next word manage to simulate human thought well enough to do all the very human-like things it does? And what does that mean about us?

What if your phone could run an LLM? Not in the cloud. Not on clusters. Right there—locally. That’s the promise of MODEGPT, the compression method we break down with Abhishek Patel in the new Executive Code. youtu.be/UwQQr4qqEJ4?si…

We’re constantly told that large language models need massive infrastructure. But what if they don’t? On this episode, we sat down with @mohammad_mmzf, author of SLiM—a new method for compressing LLMs without retraining and without major accuracy loss. youtu.be/HFS7Yc_TRWs?fe…

What if your chatbot didn’t start from zero every time? In this episode, we explore how the Memory Mosaics approach could power smarter customer service. 🎧 Full episode on YouTube, Spotify, and Apple Podcasts.

Why do language models still sound confident—and get basic facts wrong? In this episode we talk to @ChenShani2 , whose work explores how to move LLMs from surface-level pattern recognition to true concept-level reasoning. youtu.be/joH7JUec79w?si…

What if self-supervised learning could forecast the future—frame by frame? In this episode, we talk to @ziv_ravid about VJ-VCR, a new method that extends the JEPA framework to video. bit.ly/45JYusW Available on Youtube, Apple Podcast and Spotify