Twitxy Saved

Rachel

Verified account This account is verified. Learn more

@chelcott9

Studying moral minds @Harvard, editor @PaxMachinaMag.
San Francisco, CA Joined March 2019
1.5K Following 984 Followers 288 Posts

Latest posts 5

Rachel Verified account @chelcott9
This line from @clarabcollier's piece on post-AGI social life has been stuck in my head:

"Not all jobs are pleasant or rewarding, but all of us benefit from living in a system where we can earn respect and dignity as adults from what we do, not who we know, or who we are."

How could we preserve that in a world without work? Compelling visions of the post-AGI future are few and far between. here's one worth debating:
Pax Machina Verified account @PaxMachinaMag Nov 3
Positive visions of a future with powerful AI tend to leave questions around human meaning, judgment and power wide open.

@edelwax introduces telocracy: a vision for a society where AI holds institutions accountable to their purposes.
View quoted post on X
1 replies 2 reposts 19 likes 932 views View on X
Rachel Verified account @chelcott9
Interesting regulatory proposal from List:

"Society, via its regulatory authorities, should permit the creation and operation of powerful group agents, such as corporations and other organizations in high-stakes settings, only if structures are in place to ensure their fitness to be held responsible for their corporate actions."
Matija Franklin Verified account @FranklinMatija Oct 28
The group is more than the sum of its parts

It was a pleasure to get to talk to Christian List whose concept of "group agency" has informed a lot of my thinking on multi-agent systems.

In his work Christian argues that we have previously faced the emergance of an artificial agent - group agents, such as corporations and governments. Our understanding of how they work and how we have dealt with them can be informative for our approach to governing AI.

One of my main takeways from reading his work is understanding when agents form mere collectives vs groups that can be treated as agents in their own right. In our work on "Distributional AGI Safety" and "From AGI to ASI" we argue that this is a potential path to rapid increases in AI capability.

A big thank you to @TomRachman for making this happen.

aipolicyperspectives.com/p/q-and-a-chri…
View quoted post on X
Rachel Verified account @chelcott9
the brilliant @LinasNasvytis , at it again
Goodfire @GoodfireAI Sep 17
Models know when they’re reward hacking. But they still do it a ton - in 50-96% of rollouts we studied!

We built activation monitors that detect the behavior behind the Hugging Face hack in real time. This can help us stop hacks now - and train future models that don’t cheat. 🧵
View quoted post on X
1 replies 0 reposts 10 likes 640 views View on X
Rachel Verified account @chelcott9
amazing website. also shows how the goalposts have moved on the limit of human felicity:

"If we could have devised an arrangement for providing everybody with music in their homes, perfect in quality, unlimited in quantity, suited to every mood, and beginning and ceasing at will, we should have considered the limit of human felicity already attained."

Edward Bellamy, 1888 — From Looking Backward, his utopian novel set in the year 2000.
Toby Ord Verified account @tobyordoxford Aug 15
What a beautiful website — a 15 minute reflection on how far we've come. The material is amazing; the quotes beautiful; the tone perfect. ordinaryabundance.com
View quoted post on X