2025: AIs can reliably do 2-to-4-hour ML engineering tasks and sometimes medium-quality incremental research (e.g. conference workshop paper) with strong elicitation.
2026: AIs can reliably do 8-hour ML-engineering tasks and sometimes do high-quality novel research (e.g. autonomous research that would get accepted at a top-tier ML conference) with strong elicitation.
I don’t believe any of these summaries has been, is, or will be correct. AIs still can’t reliably do ML engineering tasks that take me ~30 minutes, at least not when I actually try it in practice. I will be shocked if an AI produces a medium-quality conference paper this year. General-purpose AIs will not do high quality novel research in 2026 (narrow AI along the lines of AlphaFold probably will to some extent).
I don’t believe any of these summaries has been, is, or will be correct. AIs still can’t reliably do ML engineering tasks that take me ~30 minutes, at least not when I actually try it in practice. I will be shocked if an AI produces a medium-quality conference paper this year. General-purpose AIs will not do high quality novel research in 2026 (narrow AI along the lines of AlphaFold probably will to some extent).