YC Paper Club Explores AI Compute Beyond GPUs: Optical, Neuromorphic and Biological Computing
ycombinator · x · 2026-10-02
Y Combinator's latest Paper Club session tackles "alternative compute": modern AI has been shaped by the tight coupling of transformers, backpropagation and GPUs, but as compute efficiency plateaus, the next generation of machines may be built very differently.
Three main directions covered:
- Optical computing: performing computation with light, including building a diffusion model with light and why optical computing isn't everywhere yet.
- Neuromorphic computing: brain-inspired chip architectures, what the brain can teach chip design, and where the field stands today (talk by Alok Vasudev).
- Biological computing: training living brain cells to play Doom, using reinforcement learning on biological neurons, and how brain cells learn from feedback (talk by vytalow).
The session also covers optimization beyond backpropagation, including zero-order optimization and SOMA. Total runtime is about 80 minutes, with a full chapter timeline provided.
Related event: YC Paper Club Explores Alternative Compute Beyond GPUs(2 posts)→
More from Infra
- Microsoft open-sources NVX, an ultra-light OpenVMM-based micro-VM sandbox for agentic workloads — unixterminal · 2026-10-02
- Amazon reportedly to offload $8B worth of Nvidia AI chips to outside investors — Polymarket · 2026-10-02
- Decision models now run on-device in llama.cpp, says Hugging Face CEO — ivan_bezdomny · 2026-10-02
- Firmus eyes $7bn IPO, set to be Australia's second-biggest ever, amid datacentre backlash — nordicinst · 2026-10-02
- Micron bets on custom HBM: NVHBM with Nvidia uses TSMC-manufactured base die — pstAsiatech · 2026-10-02
- Google's 81-satellite TPU datacenter plan called "launch economics cosplay" — 15-min bursts admit power is the problem — TansuYegen · 2026-10-02