Subvolts take AI is optimizing its own hardware, unlocking efficiency beyond bigger models and data. Feb 27, 2026 - Developers
- Learn how AI models can optimize GPU kernels and improve infrastructure efficiency for better performance.
- Business
- Understand how AI hardware optimization impacts compute economics and could define future AI scaling.
This episode provides a detailed look at AI infrastructure and optimization, a crucial area for scaling AI progress beyond model size.
#324 Sharon Zhou: Inside AMD's Plan to Build Self-Improving AI
Audio streams from Craig S. Smith. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Eye on AI features Sharon Zhou, AMD's VP of AI, discussing self-improving AI. The episode explores how language models optimize GPU kernel code, the role of reinforcement learning in post-training, and kernel optimization as a scaling factor. Zhou explains GPU efficiency's impact on training and inference costs, continual learning challenges like catastrophic forgetting, and how hardware profiling aids self-improvement. The conversation also touches on compute economics, synthetic data, and infrastructure's role in AI's next phase.
Imagine AI learning to tune its own engine for better performance. This episode explains how AI models are starting to write code that makes computer chips run faster and more efficiently. This matters because it can make AI cheaper and more powerful for everyone.
This episode offers a deep dive into AI infrastructure, a critical but often overlooked area. Listeners learn how AI models are becoming more efficient by optimizing the hardware they run on, a key factor in scaling beyond larger models and more data. This is relevant for anyone tracking AI's practical advancements, especially those focused on compute efficiency and cost reduction. It connects to frontier labs' efforts in optimizing their own hardware and software stacks for massive AI deployments.
- HostCraig S. Smith
- GuestSharon ZhouVP of AI, AMD
- Guest: Sharon Zhou, VP of AI at AMD
- Show: Eye on AI
- Episode Title: Inside AMD's Plan to Build Self-Improving AI
- Length: 46 minutes
- Key Topic: Language models writing and optimizing their own GPU kernel code.
- Key Topic: Kernel optimization as an overlooked scaling lever in AI.
AI is not just getting smarter. It is getting faster by learning how to optimize the hardware it runs on. In this episode, Sharon Zhou, VP of AI at AMD and former Stanford AI researcher, explains how language models are beginning to write and optimize their own GPU kernel code. We explore what self improving AI actually means, how reinforcement learning is used in post training, and why kernel optimization could be one of the most overlooked scaling levers in modern AI. Sharon breaks down how GPU efficiency impacts the cost of training and inference, why catastrophic forgetting remains a challenge in continual learning, and how verifiable rewards from hardware profiling can help models improve themselves. The conversation also dives into compute economics, synthetic data, RLHF, and why infrastructure may define the next phase of AI progress. If you want to understand where AI scaling is really happening beyond bigger models and more data, this episode goes under the hood. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI (00:00) Preview and Intro
(00:25) Sharon Zhou's Background and Transition to AMD
(02:00) What Is Self-Improving AI?
(04:16) What Is a GPU Kernel and Why It Matters
(07:01) Using AI Agents and Evolutionary Strategies to Write Kernels
(11:31) Just-In-Time Optimization and Continual Learning
(13:59) Self-Improving AI at the Infrastructure Layer
(16:15) Synthetic Data and Models Generating Their Own Training Data
(20:48) AMD's AI Strategy: Research Meets Product
(23:22) Inside the NeurIPS Tutorial on AI-Generated Kernels
(30:59) Reinforcement Learning Beyond RLHF
(39:09) 10x Faster Kernels vs 10x More Compute
(41:50) Will Efficiency Reduce Chip Demand?
(42:18) Beyond Language Models: Diffusion, JEPA, and Robotics
(45:34) Educating the Next Generation of AI Builders
- GPU kernel code
- Sharon Zhou
- The specific instructions that tell a graphics processing unit (GPU) how to perform a task.
- Reinforcement learning
- Sharon Zhou
- A type of machine learning where an AI learns by trial and error, receiving rewards or penalties for its actions.
- Catastrophic forgetting
- Sharon Zhou
- A problem in continual learning where an AI forgets previously learned information when it learns new things.
- RLHF
- Sharon Zhou
- Reinforcement Learning from Human Feedback, a technique to align AI behavior with human preferences.
55 min42 min51 min34 minUkraine's Secret Weapon: The Points System Winning the Drone War | Andrii Hrytseniuk
WithAndrii Hrytseniuk
Questions people ask
- What is self-improving AI?
- Self-improving AI refers to models that can learn how to optimize the hardware they run on, such as writing and optimizing their own GPU kernel code.
- Why is GPU kernel optimization important for AI?
- Kernel optimization directly impacts the cost of training and inference for AI models, making it a significant scaling lever that is often overlooked.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Craig S. Smith.