Matt Shumer: Government incentives could fund AI alignment without starving model training

mattshumer_ · x · 2026-09-10

In a debate with Matthew Berman and Theo, Berman argued that allocating more resources to alignment research means fewer resources for training the next model. Matt Shumer pushed back: government financial incentives for alignment research could expand the total resource pool, making alignment and model training less of a zero-sum tradeoff.

Original post →

More from AGI Musings

AGI Musings channel →