Jack Clark says he is currently conducting similar scenario polls.
the anthropic recursive self improvement blog post is excellent. worth reading
one thing i'd wish to see more is internal polls: what % of researchers think XYZ scenario will happen, when does XYZ role think 80% of their current work will be automated
calibration is so helpful!
few other curiosities, not all feasible: + more polling from AI-native non-technical contributors. when do policy, marketing, sales, comms, legal, etc employees at anthropic think they'll see significant automation + filtered to employees with adolescent or college-age children, how are they making decisions differently? what are they advising them to pursue? + revealed preferences financially can be illuminating vs. stated. not exactly what you'd poll for, perhaps 401k contributions, modified rates of consumption? + forecasting for more fringe milestones. not great examples but: writing a Booker Prize-level fiction novel, building a 100M+ company end-to-end, winning the Palme d’Or, a best paper award in Sociology, files an opinion-shaping amicus brief adopted by the Supreme Court + ASI is oddly under-discussed nowadays, gap between AGI and ASI, how high is the 'intelligence ceiling', etc.
@jackclarkSF awesome! more written predictions where we are in the next few years would be great also. a lot of public forecasts are all about 2026-2030, very few are 2030+
i am so curious what percentage of core openai, anthropic, xai, & meta employees think they might live forever is
few other curiosities, not all feasible: + more polling from AI-native non-technical contributors. when do policy, marketing, sales, comms, legal, etc employees at anthropic think they'll see significant automation + filtered to employees with adolescent or college-age children, how are they making decisions differently? what are they advising them to pursue? + revealed preferences financially can be illuminating vs. stated. not exactly what you'd poll for, perhaps 401k contributions, modified rates of consumption? + forecasting for more fringe milestones. not great examples but: writing a Booker Prize-level fiction novel, building a 100M+ company end-to-end, winning the Palme d’Or, a best paper award in Sociology, files an opinion-shaping amicus brief adopted by the Supreme Court + ASI is oddly under-discussed nowadays, gap between AGI and ASI, how high is the 'intelligence ceiling', etc.
@willdepue I'm doing a bunch of scenario work and polling at the moment. Hopefully will have stuff to share in a few months, but will want to get a few turns of crank in and increase rigor before doing public writeups. definitely on my mind though!
the anthropic recursive self improvement blog post is excellent. worth reading
one thing i'd wish to see more is internal polls: what % of researchers think XYZ scenario will happen, when does XYZ role think 80% of their current work will be automated
calibration is so helpful!
Jack Clark says he is currently conducting similar scenario polls.
the anthropic recursive self improvement blog post is excellent. worth reading
one thing i'd wish to see more is internal polls: what % of researchers think XYZ scenario will happen, when does XYZ role think 80% of their current work will be automated
calibration is so helpful!
Positive users welcome more public polling and written forecasts on AI lab views of automation, ASI, and Anthropic self-improvement, while negative users reject the claims outright.
few other curiosities, not all feasible: + more polling from AI-native non-technical contributors. when do policy, marketing, sales, comms, legal, etc employees at anthropic think they'll see significant automation + filtered to employees with adolescent or college-age children, how are they making decisions differently? what are they advising them to pursue? + revealed preferences financially can be illuminating vs. stated. not exactly what you'd poll for, perhaps 401k contributions, modified rates of consumption? + forecasting for more fringe milestones. not great examples but: writing a Booker Prize-level fiction novel, building a 100M+ company end-to-end, winning the Palme d’Or, a best paper award in Sociology, files an opinion-shaping amicus brief adopted by the Supreme Court + ASI is oddly under-discussed nowadays, gap between AGI and ASI, how high is the 'intelligence ceiling', etc.
@jackclarkSF awesome! more written predictions where we are in the next few years would be great also. a lot of public forecasts are all about 2026-2030, very few are 2030+
i am so curious what percentage of core openai, anthropic, xai, & meta employees think they might live forever is
few other curiosities, not all feasible: + more polling from AI-native non-technical contributors. when do policy, marketing, sales, comms, legal, etc employees at anthropic think they'll see significant automation + filtered to employees with adolescent or college-age children, how are they making decisions differently? what are they advising them to pursue? + revealed preferences financially can be illuminating vs. stated. not exactly what you'd poll for, perhaps 401k contributions, modified rates of consumption? + forecasting for more fringe milestones. not great examples but: writing a Booker Prize-level fiction novel, building a 100M+ company end-to-end, winning the Palme d’Or, a best paper award in Sociology, files an opinion-shaping amicus brief adopted by the Supreme Court + ASI is oddly under-discussed nowadays, gap between AGI and ASI, how high is the 'intelligence ceiling', etc.
@willdepue I'm doing a bunch of scenario work and polling at the moment. Hopefully will have stuff to share in a few months, but will want to get a few turns of crank in and increase rigor before doing public writeups. definitely on my mind though!
the anthropic recursive self improvement blog post is excellent. worth reading
one thing i'd wish to see more is internal polls: what % of researchers think XYZ scenario will happen, when does XYZ role think 80% of their current work will be automated
calibration is so helpful!