Does Astra count as a frontier-scale model with recurrence?
A user urged frontier labs to jointly commit not to train frontier-scale recurrent or otherwise unmonitorable language models. A reply questioned where Astra fits.
TLDR
On September 2, a user warned against training a frontier-scale recurrent—or otherwise unmonitorable—language model and called for frontier labs to coordinate on a commitment never to do so. They said that day had not arrived. A September 15 reply challenged that assertion, asking: “how is Astra not a frontier scale model with recurrence?”
Combined views
—
5 Sources, first seen ago
