If continual learning is nearly solved, why do we still pick reasoning effort tiers?
akbirthko · x · 2026-09-20
The author argues: if continual learning is "solved" or labs are close to RSI (recursive self-improvement), why do models still expose low/medium/high/ultra reasoning effort settings? The goal should be getting the job done, not making users choose how many tokens to spend.
More from AGI Musings
- Charlie Warzel urges a middle path between AI doomers and skeptics — round · 2026-09-20
- Muse's Spotify connector seen as start of consumer actions moving into LLM chats — anikembhavi · 2026-09-20
- The Guardian Q&A: AI doom looks like stalled systems and misaligned goals, not sci-fi apocalypse — nordicinst · 2026-09-20
- Reviewer warns: 8 slop papers, all rejected — and it closes doors to PhD admissions — AnkaReuel · 2026-09-20
- Andrew Yang warns of AI bubble as organizations pull back on spending over low ROI — Polymarket · 2026-09-20
- Is Scale 'Solving' Computer Vision? Researchers Clash Over 3D Understanding, Robotics Timeline — Sadlylifes41u · 2026-09-20