CBPAI - Coalition for a Baruch Plan for AI

We're part of a new initiative called the "Coalition for a Baruch Plan for AI". The goal of this initiative is to set up an international treaty + organization. Inspired by the 1946 Baruch Plan for nuclear governance, adapting its principles to AI to prevent vetoes and geopolitical tensions.

PauseAI is a member of this coalition. Just like Trustless.ai (@Rufo Guerreschi) and @Tolga Bilge.

https://www.cbpai.org/

Masterplan PDF

How you can help

We're looking for volunteers to help with setting up the organization and communications.
Help out with communications (website + socials, most of which still need to be set up)

CBPAI

urgent-maroon•8/1/24, 11:25 AM

this I like very much. I need some time to read through the lengthy proposal as I don't have any extensive knowledge on the Baruch plan and why it failed to form a rounded opinion on this initiative but it is at least something very concrete and seemingly well written out from skimming, that we can start taking to political representatives and others

urgent-maroon•8/1/24, 11:29 AM

I'll name drop it in my call today anyway just to get awareness of it going and will return with opinions later

colossal-harlequin•8/2/24, 1:57 AM

Did you reach out directly to https://encodejustice.org ? They seem like such an obvious coalition member.

Encode JusticeAtiqur Rahman

Homepage - Encode Justice

Encode Justice is reimagining our generation’s collective AI future.

Ccolossal-harlequin Did you reach out directly to https://encodejustice.org ? They seem like such an...

hurt-tomatoOP•8/2/24, 7:26 AM

We definitely will!

Ccolossal-harlequin Did you reach out directly to https://encodejustice.org ? They seem like such an...

specific-silver•8/16/24, 2:26 PM

Yes, seems from this post that they would be aligned with fostering a Baruch Plan for AI:

https://encodejustice.org/youth-activists-call-on-world-leaders-to-set-ai-safeguards-by-2030/

Encode JusticeSneha Revanur

Youth call on world leaders to set AI safeguards by 2030 - Encode J...

Youth activists issued a sweeping call to action on Thursday for heightened AI protections.

hurt-tomatoOP•11/25/24, 12:10 PM

Live document

Google Docs

Case for a Coalition for a Baruch Plan for AI (v.1 live draft)

Case for a Coalition for a BaruchPlan for AI (v.1.0) By the Coalition for a Baruch Plan for AI (Live late draft. To be published on December 5th, 2024) Abstract This text makes a case for a Coalition for a Baruch Plan for AI. It presents a comprehensive proposal for the timely establishment...

skinny-azure•11/25/24, 1:34 PM

Is there anywhere we can submit feedback? I noticed a typo

Sskinny-azure Is there anywhere we can submit feedback? I noticed a typo

hurt-tomatoOP•11/25/24, 1:44 PM

I'll ask Rufo to open it up for suggestions

colossal-harlequin•11/26/24, 4:15 PM

Please do.

I like the look of this a lot: for me it's a good fit as an input for the incoming US administration to consider.

I realize others may have been contacted behind the scenes, but from the names of experts cited I'm not confident anyone in or near US National Security has yet seen it? And it seems likely any draft would benefit from comments from same.

So explicitly: I really think CBPAI should consider explicitly reaching out to the Harris brothers, edouardedouard and jeremie@jeremie@gladstone.aigladstone.ai

they are the nerds who got US gov interested in soliciting their Action Plan, helped OpenAI folk blow whistles, have gotten to talk to a bunch of backroom folk whose day job is "how do we reduce the risk of nuclear conflict today" and had time presenting AIXR issues to the Trump executive team.

he live draft shouldn't be widely shared, of course, but I plan to point these two at it - but contact from the authors would be more worthy of attention.

colossal-harlequin•11/26/24, 4:26 PM

A vote weighting based on population and GDP would ensure that powerful and rich nations, starting from superpowers would have a moderately larger share of power and upsides.

This is pragmatic but I wouldn't want to pretend it is fair. The "is" says those currently wealthy can indeed force decisions that concern risks to everybody, but my "ought" reckons they should not.

colossal-harlequin•11/26/24, 4:55 PM

The Astounding Opportunities of AI

Similarly, I get the reasons to promise wonders along with saving the world from destruction. I was a bit surprised there was no mention of e.g. narrow-not-general AI as a mechanism here but appreciate there's a wide variety of views on the tenability of solving technical alignment outside of a race.

I know we don't have to put "even though it might be safe, we can choose not to build sufficiently dangerous capability and sacrifice significant upside" front and center. Saying it loud here, though.

colossal-harlequin•11/26/24, 5:01 PM

which makes them ever less relevant as the AI race merges become more and more a dual-use race among superpowers.

Suspect "the AI race merges to become" was intended, and suggest "all AI races merge to become" is better still.

colossal-harlequin•11/26/24, 5:04 PM

It would also help substantially is all other states

s/ is / if /

colossal-harlequin•11/26/24, 5:10 PM

not only the most secure route, but alos the fastest

s/ alos / also /

colossal-harlequin•11/26/24, 5:46 PM

OK, think no point in me suggesting further corrections one-by-one. The writing appears to get rather less tight and polished as we get into the roadmap, etc. That's fine.

skinny-azure•11/26/24, 6:12 PM

I also have some suggestions, a coordinated place to submit them would be nice

Ccolossal-harlequin Please do. I like the look of this a lot: for me it's a good fit as an input fo...

colossal-harlequin•11/30/24, 3:33 PM

Was hoping for a "great" or a "no", but without those, I did mail the Harris bros on this. CBPAI cc-ed and @Joep Meindertsma bcc-ed.

colossal-harlequin•12/6/24, 3:11 PM

They are in contact. Acknowledgement from Edouard and a solicitation for their input from Rufo. Yay. https://discord.com/channels/1100491867675709580/1233038219650535484/1314609751471951892

CBPAI - Coalition for a Baruch Plan for AI

How you can help

Similar Threads

Similar Threads

Similar Threads