Help us seed AI Safety Brussels

post by gergogaspar (gergo-gaspar), ENAIS (enais-contact) · 2024-08-07T06:32:24.760Z · LW · GW · 2 comments

Contents

  TLDR:
  Potential priorities
    Running courses for professionals
    Seeding university groups
    Organising events, meetups and connecting the community
  Who is working on this at the moment
  On risks
  Conclusion
None
2 comments

crossposted on the EA forum: https://forum.effectivealtruism.org/posts/EwdgPJfeieCxe4cxa/help-us-seed-ai-safety-brussels

TLDR:

Brussels is a hotspot for AI policy and hosts several think tanks doing good work on AI safety. However, there is no established AI Safety group with paid staff that brings the community together or works on reaching out to students and professionals. For this reason, the European Network for AI Safety (ENAIS) and members from EA Belgium are teaming up to seed AI Safety Brussels.

This is a call for expressions of interest for potential collaborators to make this happen, with a wide range of ways to contribute, such as being a potential founder, policy/technical lead, volunteer, advisor, funder, etc. If you are interested in helping out in some capacity, please fill out this 2-minute form.

Potential priorities

I (Gergő) will caveat this by saying that the exact strategy (and name) of the organisation will be determined by the founding team and advisors. However, we think there are several potential pathways in which someone doing community building full or part-time could add a lot of value.

Running courses for professionals

The AIS community, as well as the EU AI Office, are bottlenecked by senior talent. Currently, there are only a few groups and organizations working to do outreach towards professionals.[1]

Brussels has a lot of senior people working on AIS. Understandably, they don't run courses to onboard others to AI Safety, as they are busy with object-level work. Conditional on seniority, if someone gets funded to start AIS Brussels, they could leverage the existing network and create an environment that is quite attractive for (policy) professionals to join, who are new to the field of AI Safety. By running courses similar to AI Safety Fundamentals by Bluedot, such a person (or team) could introduce AIS to hundreds of professionals per year and support them in their journey of upskilling and help them get into high-impact roles.

For clarification, by professional outreach, we don't mean actively reaching out to policy professionals working at e.g., the European Parliament to request meetings etc. The existing think tanks are in a better position to do this kind of work.

Seeding university groups

To our knowledge, Brussels has no AI Safety university groups at the moment. The founders could help seed such groups, by doing city-wide outreach [EA · GW] to students.

Organising events, meetups and connecting the community

We know from an AI policy organizer in Washington that gated events (such as invite-only dinners for people in AI policy) can add a lot of value. As far as we know people working in think tanks are well-connected, but perhaps the broader AIS community could benefit from more events and meetups. A paid organiser could support people who volunteer their time to make current meetups happen, as well as organise additional events if there is a sufficient need for them.

We’re currently asking our contacts in Brussels whether they feel like there are sufficient opportunities to network with peers, or whether more or different opportunities would be helpful. 

Who is working on this at the moment

I, Gergő Gáspár, co-director for the European Network for AI Safety (ENAIS) am currently spearheading this project. I have 4+ years of EA/AIS community-building experience by founding EA and AIS Hungary and providing intro courses to 300+ people. I could support the new hires and share best practices for running courses and events. 

Tom Dugnoille, software engineer and organiser for EA Brussels. He has been living in Brussels for 9 years and would be able to support the founding team in getting a sense of the local landscape.

Armand Bosquillon de Jenlis is a computer engineer and independent AI policy and strategy researcher. He has been living in Belgium for 31 years.

Agus Covarrubias, CEA's AI Safety group support lead, as well as the rest of the ENAIS staff supports us: Manuel Allgaier (ENAIS co-director, previously director of EA Germany, EA Berlin & EAGxBerlin 2022) and Teun van der Weij (MATS scholar). This work is overseen by the ENAIS board: Esben Kran (Co-director of Apart Research), Dušan D. Nešić (Operations Lead at PIBBSS, EA Serbia & AIS Hub Serbia founder) and Jonathan Claybrough (EffiSciences board member).

On risks

Starting any (field-building) organisation comes with downside risks. Given the importance of Brussels in the AI policy landscape, these risks are significantly higher compared to other AIS city/national groups. We assess these risks carefully and are actively seeking the advice of experienced community builders and (local) AI policy professionals. If you are interested in helping us minimize downside risks please fill out our form and indicate that you are open to red-teaming our strategy.

We would be especially interested in hearing from you if you thought carefully about community-building in this context and think that starting such an organisation would be net-negative. Please email me at gergo@enais.co or leave us anonymous feedback here.

Conclusion

We think there is likely low-hanging fruit that organisers working in a professional capacity could capitalise on, conditional on the right strategy. If you are interested in helping us make AIS Brussels happen, please fill out this form.

If you would like to keep up to date on this project and other field-building-related news in Europe, sign up for the ENAIS newsletter for field builders.

  1. ^

    These are Successif, High Impact Professionals, Halcyon Futures, Far AI, Bluedot, and Arkose. 80,000 Hours is also doing some work on this front, as well as some national groups such as EA Sweden and AIS Hub Serbia.  My claim is that even if all of these organisations had 2x the capacity that they currently have, professionals outreach would still be neglected.

2 comments

Comments sorted by top scores.

comment by Chris_Leong · 2024-08-08T00:55:17.230Z · LW(p) · GW(p)

I previously ran AI Safety Australia and New Zealand. Happy to chat and share my thoughts if that’s useful!

Replies from: gergo-gaspar
comment by gergogaspar (gergo-gaspar) · 2024-08-09T09:24:15.076Z · LW(p) · GW(p)

That would be awesome, thanks for offering! Will reach out next week!