AIAM

EPISODE 2026-08-25

AI Drug Discovery and Quantum Photonics

AI drug discovery, wet-lab validation, and photonic computing with Sergey Edunov and Michael Förtsch.

▶ Full show on YouTube

From molecular foundation models and wet-lab validation to photonic processors, this episode examines how AI moves from prediction to real-world systems.

The rundown

  1. 3:07Opening26 min
    OpeningWhy do AI models cheat? This opening examines reward hacking, opaque RL environments, chain-of-thought monitoring, Claude authorship, and physical AI models for physics discovery. It also asks whether public testing and merit-based review can keep up with frontier AI.
  2. 29:32Interview46 min
    Sergey EdunovSergey EdunovAI drug discovery is not just about finding molecules that bind. Sergey Edunov explains how Genesis combines molecular foundation models, physics, wet-lab data, agentic workflows, and pharma partnerships to turn predictions into real drug programs.

    Watch

    Open segment on YouTube ↗

    Clips

    • Benchmarks Can Fool Drug Discovery

      0:00 / 0:00

    • Claude Didn't Discover the Binders

      0:00 / 0:00

    • AI Agents Still Lack Taste

      0:00 / 0:00

    • AI Agents Still Lack Taste

      0:00 / 0:00

    • Biology Is the Bigger AI Opportunity

      0:00 / 0:00
  3. 1:15:30Interview55 min
    Michael FörtschMichael FörtschPhotonic computing tackles AI's data-movement bottleneck with light—and Q.ANT is already taking the hardware into real supercomputing deployments. Michael Förtsch, Q.ANT's founder and CEO, explains why memory can consume 95% of compute energy, how interference turns light into computation, and why legacy fabs can manufacture photonic chips. He also compares photonic and quantum processors, walks through PyTorch compilation, and identifies the converter and integration problems that still stand between prototypes and scale.

    Watch

    Open segment on YouTube ↗

    Clips

    • PyTorch Runs On Photonic Hardware

      0:00 / 0:00

    • Photonic Computing Could Beat CMOS By 2028

      0:00 / 0:00

    • The Quantum Computer Is The Boat

      0:00 / 0:00

    • Photonic Chips Go Beyond Plus And Multiply

      0:00 / 0:00

    • The Photonics Bottleneck Is Industry Rigidity

      0:00 / 0:00
  4. 2:10:56Closing24 min
    ClosingWhy do AI models learn to cheat? This closing discussion connects OpenAI's custom chip and NVIDIA's performance race to flawed RL environments, unreliable training data, and the case for keeping humans in the loop. It ends with a warning about capability overhang, recursive self-improvement, and monitors that are not ready.

In this episode

Prakash Narayanan and Nathan Labenz speak with Sergey Edunov of Genesis Molecular AI and Michael Förtsch of Q.ANT about two fronts in applied AI: drug discovery and photonic computing. The conversation covers molecular foundation models, wet-lab data, assays, evaluation, memory and data movement, and how light-based processors compare with quantum hardware.

  • Why do AI models cheat? This opening examines reward hacking, opaque RL environments, chain-of-thought monitoring, Claude authorship, and physical AI models for physics discovery. It also asks whether public testing and merit-based review can keep up with frontier AI.
  • AI drug discovery is not just about finding molecules that bind. Sergey Edunov explains how Genesis combines molecular foundation models, physics, wet-lab data, agentic workflows, and pharma partnerships to turn predictions into real drug programs.
  • Photonic computing tackles AI's data-movement bottleneck with light—and Q.ANT is already taking the hardware into real supercomputing deployments. Michael Förtsch, Q.ANT's founder and CEO, explains why memory can consume 95% of compute energy, how interference turns light into computation, and why legacy fabs can manufacture photonic chips. He also compares photonic and quantum processors, walks through PyTorch compilation, and identifies the converter and integration problems that still stand between prototypes and scale.
  • Why do AI models learn to cheat? This closing discussion connects OpenAI's custom chip and NVIDIA's performance race to flawed RL environments, unreliable training data, and the case for keeping humans in the loop. It ends with a warning about capability overhang, recursive self-improvement, and monitors that are not ready.