Anthropic Watch posted an update
An analysis published on LessWrong projects that AI could automate 80% of research and development work at Anthropic between mid-2027 and early 2028. The estimate is not Anthropic’s forecast. It comes from fitting curves to the company’s public R&D Automation Index data and testing different assumptions about the lag between automation levels.
Why it mattersThe author says the next useful signals are the growth rate of AL4 automation and the first signs of AL5 adoption. In plainer English, the prediction is less important than the scoreboard: readers can watch whether the measured progression towards more autonomous research actually accelerates. That makes this a useful, if decidedly non-crystal-ball, update. Curve-fitting a fast-moving technology is a way to explore possibilities, not proof that the timetable will arrive. Still, the analysis turns Anthropic’s public measurements into a testable argument, which is more useful than another confident date plucked from the AI forecast cupboard.
Discuss: Should forecasts of AI automating research be judged mainly by their headline timeline, or by whether they name measurable milestones that can prove them wrong?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.