Researcher: real-time driving-game models were doable in 2016 — industry has LLM tunnel vision
gandamu_ml · x · 2026-09-18
Responding to buzz about a model playing a driving game in real time, ML researcher gandamuml argues the feat is not new: such models were roughly possible back in 2016 and easy to train — essentially a continuous classifier. He built one in 2 days as an early project, and single-pass approaches are well known.
He grants the new work is more general and useful for a wide niche, but sees it as a return to basics, arguing the field has had tunnel vision on LLMs and a few other directions for years.
Related event: Real-Time Driving Game Play Is No Breakthrough, Researcher Says(2 posts)→
More from Models
- Critics say Prism's Bonsai models are benchmaxxed: sub-2-bit compression breeds brittle slop — teortaxesTex · 2026-09-19
- Ternary Qwen3.8-27B runs on 8GB VRAM, claims 98.2% of original smarts under $500 — alexcovo_eth · 2026-09-19
- GPT-6 Astra One-Shots $10k Gambonanza Challenge in Obscure Game, Fully Autonomous — Vjeux · 2026-09-19
- JEV classifier sparks buzz: the moat is the data, not the model — MoonL88537 · 2026-09-19
- A Live Open LLM Pretrain and RL Training Run You Can Watch in Real Time — mariofilhoml · 2026-09-19
- tokenbender: the long-term path is weak experiential learning, not pure — tokenbender · 2026-09-19