[Governance] Craig Mundie and Tom Friedman on global AI governance - "Make a wiki"

[Joly MacFie]

"So we're just collectively going to build one mechanism. And people feed
in all the collective set of rules in a well-defined process, right? And
basically it, you think of this as essentially a wrapper around an existing
AI. So it sees the inputs and it sees the attempted outputs, and then it
makes a decision as to whether it's going to permit that output. If it
does, it signs it. If it doesn't, it doesn't, right? And then the world
moves along.

The rules are just basically, the way I think of this in very simplistic
terms, is what you want to have is a giant curated wiki. And for each
societal element who has basically, is a participating, call it country for
starters, the country figures out for that thing who their proxies are to
take their rule sets and put them in on their page of the wiki—says US
rules, China rules, India rules, Europe rules, or French rules, or whatever
they're right. And they don't have to negotiate with anybody, they just put
them in. And all they expect is that when an AI is operating essentially in
the virtual sovereign environment of that country, it applies their rules
in that context. And so you've chunked this thing out into a bunch of
separate activities, each engineered for purpose and then you essentially
compose a system out of them."

https://www.patreon.com/posts/148797684/



-- 
--------------------------------------
Joly MacFie  +12185659365
--------------------------------------
-
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.igcaucus.org/pipermail/governance/attachments/20260122/a8ee2c59/attachment.htm>