We help people who care about AI safety find a co-founder and start something together.
The people are there, and so is the money. What's missing is people starting projects. Most apply for jobs instead.
Many skilled people want to work on AI safety.
Funders are looking for good projects to back.
Too few people start something. They wait for a job offer instead.
Knows AI safety research or policy.
Applies to the same few jobs as hundreds of others. Most get rejected.
Knows how to start and run an organization.
Wants to help, but doesn't know the field well enough to see what to start.
Together they could start something. But they rarely meet.
For example BlueDot and Generator.
The problem: they're hard to get into. Thousands apply for about 10 places. Most of the people who don't get in never start anything.
Meetups, conferences and talks.
The problem: people stay in their own circle. Policy people talk to policy people. Technical people talk to technical people.
If experts and builders could easily find each other, many more AI safety organizations would get started.
We're doing research through interviews and forms to understand why people aren't starting more AI safety organizations.
We match operations people with experts in AI safety. Apply and we'll try to find you a match.
We also use the answers to learn what helps and what stops people from starting AI safety organizations.
Events in London and Brussels where experts and builders meet to find someone to start a project with.
We asked. We interviewed 16 people at an AI safety bootcamp, and 124 people filled in our matching form. Here's what they told us.
Every one of the 16 people we interviewed had at least one idea. Three are already running a program or organization. Six never got further than notes or conversations.
In the matching form, 99 of 124 put starting or joining an early-stage organization at or near the top of their list. Only one ruled it out.
16 interviewees
The most common reason was not having the right partner. Someone was missing, left for a job, or wasn't available.
Money and self-doubt came next. Self-doubt shows up early, before people have a plan. Partner and money problems show up at every stage.
Most people named more than one reason, 2.6 each on average.
out of 16 interviewees
The interviewees with ideas struggled to find operations and fundraising people. Yet 56 of the 124 form respondents list operations as one of their skills, several with decades of experience running teams and organizations.
Only 65 of 124 respondents have a concrete idea. Many would rather join than lead. Fundraising is the rarest skill: only 22 people have it.
Both groups tell the same story, as the chart shows.
share of each group that mentioned it
| Theme | Interviews (of 16) | Form (of 124) |
|---|
Every interviewee said yes to an in-person event. What they asked for was someone to make the right introductions.
The most common wish was for someone to filter first.
Talk to the three people that could be interested, and then we see if it's a match.
People want to test working together before committing.
I want to test the relationship over a bit of time … trying out a little project and seeing if we work together.
Money is the biggest worry for people who can't take a financial risk. Matches should come with a route to funding, an incubator or a stipend.
One interviewee started an organization only after someone suggested it. Another went ahead after encouragement from a funder.
So that's what we're building: curated matches, and events where experts and builders meet.
About this data. These are small, early samples. All 16 interviewees attended the same bootcamp, so they are more engaged than most. The 124 form respondents chose to fill in a matching form, and all responded between 26 October and 2 November. Treat the numbers as signals, not proof.