Embed­ded World-Models

(A longer text-based ver­sion of this post is also avail­able on MIRI’s blog here, and the bib­li­o­graphy for the whole se­quence can be found here)

(Edit: This post had 15 slides ad­ded on Saturday 10th Novem­ber.)

The next post in this se­quence, ‘Embed­ded Agency’, will come out on Sunday, Novem­ber 4th.

To­mor­row’s AI Align­ment Forum se­quences post will be ‘The easy goal in­fer­ence prob­lem is still hard’ by Paul Chris­ti­ano, in the se­quence ‘Value Learn­ing’.