Reflections on "API Science" and Its Negative Impact on AI Research
tessera_antra · x · 2026-09-01
Researcher tesseraantra reflected that linguists from the GPT-3 era were right in suggesting that "API science" is detrimental to the field. The argument is that this practice reifies meaningless API affordances as if they were real features of the language model itself, leading to a corruption of research focus and a misunderstanding of what is actually being studied.
More from Companies & People
- Call for OpenAI to release 70k+ message board logs — scaling01 · 2026-09-01
- “Programming is dead”: Blogger argues ‘dead’ means mature and stable — bibryam · 2026-09-01
- Li Haoyi Joins AMI to Work on World Models and Physical AI — sainingxie · 2026-09-01
- AI Makes the Org Chart Less Important as a New Routing Layer — bibryam · 2026-09-01
- Engineers must find unique value that models cannot replace — rseroter · 2026-09-01
- Apple accuses OpenAI employee of stealing trade secrets to train AI agent — Polymarket · 2026-09-01