AI safety undergrad weighs dropping out over short AGI timelines

In a detailed LessWrong post, AI safety student 'hersheys' publicly weighs dropping out of university to work full-time on alignment and field-building. They frame the decision not as a binary but as a spectrum of academic engagement, noting they've already slipped from ~65% in their first semester to ~40% in the second, and are now considering dropping to just one class (15%) or zero. Key arguments include a belief in rapid AI takeoff (<2 years), school's bottleneck on their bandwidth, academia's failure to adapt curricula for a post-ASI world, and strong peer pressure from respected colleagues who recommend leaving.

Their post-dropout plans are concrete: organizing Georgia Tech AISI full-time, providing operational support to early-stage organizations like Second Look Research, founding an AI Lab Watch initiative to audit frontier lab safety practices, and scaling talent pipelines through workshops like DCMC 2.0. They also probe key assumptions — how fast transformative AI will arrive and whether current education will retain value. The post captures a broader phenomenon: a growing cohort of AI-safety-focused students who see direct field work as more impactful than formal degrees, and who are making critical career decisions based on aggressive timeline estimates.