Researcher says interpretability will be automated eventually, but the PhD was still fun
aryaman2020 · x · 2026-07-27
The author says interpretability research will eventually be automated, "like everything else," but still enjoyed doing a PhD in the area even while a trillion-dollar company was also working on it.
They add that the experience might have been different if they had been doing RL instead, framing the post as a candid reflection on research careers, competition, and why people still choose a field despite obvious industrial pressure.
More from AGI Musings
- LLMs still fail at temporal reasoning, and a hierarchical HMM is proposed for extreme long contexts — beffjezos · 2026-07-27
- AI agents may lose to UIs on repetitive work, but win on novel tasks — bendee983 · 2026-07-27
- The singularity is still 3–4 years away, says the poster — DionysianAgent · 2026-07-27
- A viral Anthropic tone-policing dispute gets framed as billionaire-driven language control — tszzl · 2026-07-27
- What should “human values” mean when we ask AI to align with them? — Salt_Progress8049 · 2026-07-27
- AGI may arrive long before people agree it has arrived — VraserX · 2026-07-27