Framing Outreach

<aside> 🗣️ Outreach should be thought of as an invitation, not as persuasion. In other words, we’re enabling students who seem intellectually curious in AI safety to join our community, not convincing them they have to.

</aside>

• This means we shouldn’t be bothered when someone just isn’t interested in the ideas. If we make our invitation, and they don’t find it appealing that’s totally okay. There will be many people that do.
• Our job as group organizers is just to know our stuff, and present the core ideas of AI safety accurately. There’s no need to twist AI safety to “get” someone to join

Note: this doesn’t mean you shouldn’t put effort into making your club a desirable place to be. Generally being friendly, finding persuasive arguments for why AI safety should be taken seriously, and avoiding unnecessarily polarizing discussions is not at odds with our principles.

Keep your goals in mind

Most AI safety university groups’ ultimate goal is something like “get as much talent into AI safety.” We think that aggressive, super broad outreach is not the best way to achieve this goal. Outreach is important, but we think establishing excellent group culture, and inspiring impactful career change probably does more good. We’d rather see a relatively smaller group of dozens of students dedicated to reducing risks from transformative AI than hundreds with a vague and incomplete notion of what AI safety is and little interest in joining the field.

Don’t reinvent the wheel

Many groups have already found successful outreach strategies and tactics.

• Focusing on the most efficient outreach strategies.
  • Most groups should join the activities fair, run faculty directed outreach, and post in group chats before spending time on physical outreach and social media pushes.
  • A step by step walkthrough of an outreach campaign can be found in GDocs
• Using resources and templates
  • AI safety Groups are more than willing to share their resources. Before starting a new outreach project, look online (e.g., this guide) or ask in the AIS Groups Slack for existing resources.

Do outreach early

Early outreach reaches students before they become overwhelmed with school or join other clubs. An outreach push which starts the week before school and ends two weeks into school is considerably more effective than later outreach.

<aside> 🕢 Top organizers organize their outreach around a “crunch time” which starts the day freshmen move in and ends a couple weeks into the school year. We recommend new groups adopt this practice.

</aside>

We recommend devoting a significant fraction of your capacity to outreach during this period. Then switch your focus to non-outreach projects.

If you’re able to execute a strong outreach push very early in the semester, consider moving your Fellowship application deadline to within the first two weeks of school. This forces people to commit before things get busy.

Take an experimental approach

The information in this guide is a summary of what works best on average for AI Safety Groups in the US and UK. It’s perfectly reasonable that some of our broadest recommendations don’t apply for your school.

So we’d generally encourage two things: taking your context into account by making judgement calls about whether you think some strategy or some piece of advice will work in your particular context; and taking an experimental approach by trying out a few different things, tracking how well they’re doing, and then iterating.

<aside> 📒 A simple and highly effective way to track data is to include a “How did you hear about our group” question on your fellowship application.

</aside>