Ahead of the Plasticity Curve: AI, Biology and Technology
summary
Argues that brain-inspired choices drive practical AI progress: Arm adds dot‑product and matrix‑multiply instructions, bfloat16 and wider vectors, and a SME architectural boost ~100x in ML throughput; GPUs target ~5x ML gains and software alone gives ~2x speedups. Shows product results: Ethos‑U55 delivers ~75x vs Cortex‑M55 (M55 itself ~10x older MCUs), on-device examples include an AI hearing aid and a fridge saving ~10% energy. Presents tooling: TOSA (67 tensor primitives) and Bumblebee NAS for platform‑aware model search.
AI-written summary · about
chapters (6)
related talks
26mKeynote: Artificial Innovation Accelerates Artificial Intelligence
35mAI innovation begins with trust
01h 14mTissue vs silicone musings on the future of deep learning hardware and software
34mThe Future of Computing: AI technologies in the exascale era
58mThe Next Decade of Software Development
18mDriving Innovation with AI
24mTools & Techniques for a cost optimized Analytics & AI pipeline
30mBeyond GPUs: Why AI Agents Need a Need New Hardware