Debate: is reward shaping enough for alignment, or just whack-a-mole patching?
tszzl · x · 2026-10-13
An alignment debate on X: @JacksonKernion argued "just reward shaping" isn't the whole ball game — but challenged critics, saying for any misaligned behavior he can show a reward that would have prevented it, and that most misalignment cases in production models look like oversights, acknowledging the whack-a-mole risk.
Large-account @tszzl pushed back that reactively shaping rewards as problems show up is likely insufficient, implying the post-hoc patching approach falls short against more systematic alignment challenges.
Related event: Alignment stagnation debate erupts as OpenAI chief scientist weighs in(19 posts)→
More from AGI Musings
- a16z and Dan Boneh: AI's math leap could compress a decade of crypto trust-building into one month — a16z · 2026-10-14
- Who bears the burden of proof on AI doom? Accelerationists and doomers clash — AccBalanced · 2026-10-14
- MIT PhD Student Co-Authors Book Chronicling the Rise and Fall of the Tech Worker Movement — soumitrashukla9 · 2026-10-14
- Tim Sweeney asked if 5-minute AAA game generation would be good for humanity, sparks debate — max_paperclips · 2026-10-14
- MIT professor's ten predictions for AI for science, from executable papers to compiled matter — ProfBuehlerMIT · 2026-10-14
- Sweden spends hundreds of millions of kronor on printed textbooks as 2.8M-child study links screens to lower grades — Eric520CC · 2026-10-14