This story is unavailable.

There seem to be two problems with this argument:

(i) if having some friendly goals in the mix is enough to ensure pretty good outcomes, we may be more robust by taking an aggregation over many goals, rather than a single best-guess;

(ii) there could even perhaps be a ‘wisdom of the crowds’ effect, where the aggregate view is a more accurate predictor than even the best single utility function (you could get around this by building lots of diverse views and aggregation into your local attempt, but you get a lot of the work for free in the distributed case).

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.