TL;DR: Lateral Workshop is a program in Berkeley from September 11–13 for mid-career and senior professionals who are interested in transitioning into AI safety work to meet people in the field and explore how they can fit in. If you have experience in another field and are interested in translating that experience into impact in AI safety, apply here by August 9th!
If AI safety is to keep scaling, we’ll need a lot of people that the field currently doesn’t supply well. People who know how to run 30+ person organizations. People who have deep connections in DC. People with specialized knowledge in cybersecurity or verification.
Professionals from outside AI safety have those skills, but it’s been difficult for them to make the switch. The professionals we have spoken with rarely lack relevant skills. More often, they struggle to understand how the field works, where they could contribute, and how to become credible candidates. Here are some of the specific barriers experienced professionals face while trying to transition:
Difficulty in understanding the key priorities of the field
Lack of networks within the field and poor understanding of where to start to build them
Lack of free time to invest in time-intensive upskilling and context building
Financial risk (getting a role is competitive, and the field is insular enough that rejections from roles are correlated)
Difficulty evaluating a young and unfamiliar field that lacks many of the conventional markers of professional maturity
Some professionals do navigate these barriers through persistent self study and deep attempts to understand the field. But requiring every candidate to independently reconstruct the field is slow and unnecessarily costly. AI safety has reduced similar friction for students by building clearer early-career pipelines; we think experienced professionals could benefit from a similar investment in talent pipelines and infrastructure.
Even after some contact with the field, it has been hard to hire experienced professionals:
In virtually all cases, the reason organizations must hire experienced talent in the first place is because the role has higher stakes. A good leader can make an organization go well, while a poor leader can devastate culture and sink an organization quickly.
Organizations are currently hesitant to take bets on people. This may be a good policy for individual organizations to follow, but can simultaneously be negative on the ecosystem level.
Hiring rounds index heavily on in-group signals that are correlated, but not necessarily causal to impact. These disadvantage people who come to AI safety from outside the main cultural context.
Professionals are sometimes poorly calibrated on which of their skills are useful—many of their specialized technical skills are less applicable to AI safety, but the tacit people-modeling knowledge they have is applicable.
Here are some excerpts from example roles requiring experienced talent:
“7–8+ years of senior communications experience … building or owning crisis communications frameworks, with an understanding of reputational risk management” – Head of Communications & PR, Apollo Research
“You’re a strong engineer with solid infra fundamentals. You can dig into unfamiliar systems, debug from logs, and identify and fix performance bottlenecks” – Member of Technical Staff, Evaluation Execution, METR
“Multiple years of relevant research or policy experience and a track record of producing high-quality work” – Research Fellow, GovAI
“Write a coalition letter and get twenty organizations to sign it, ranging from religious groups to workers’ rights advocates” – Policy Advisor, Encode
Pipeline legibility
A research-oriented person has a relatively defined ladder: BlueDot → ARENA → SPAR → MATS → junior researcher → senior researcher. For mid-career non-research talent, the paths look like this:
Founder: BlueDot → Catalyze or another incubator → hopefully get funding
Operations person: BlueDot → get hired, somehow
Generalist: BlueDot → do unspecified things for a year → Generator
There’s a missing intermediate step between introductory content and the programs that place people. For professionals, there are few clear steps after a BlueDot course.
It’s not just about content learning: for early-career people, university groups provide the grounds to transmit context, culture, and motivation, and there’s nothing similar for professionals. Successful mid-career transitions have happened by individuals who have forged their own paths, because nothing existed for them.
Past solutions
Workshops are able to significantly affect the career paths of early-career participants. Jacob estimates he got involved in AI safety 3 months earlier than he otherwise would have had he not attended an AISST/MAIA retreat. In The Case for AI Safety Capacity Building Work, Asya Bergal lists some testimonials:
Max Nadeau: “Going to the Impact Generator workshop. That was, like, extremely random. And, like, resulted in a major acceleration to my career.”
Rachel Weinberg: “Helping with Future Forum, which came from meeting Leilani at Impact Generator”
Gabriel Wu: “[Having] the opportunity to visit Constellation”
Aric Floyd: “After Icecone, I was pretty bought in on I should do something about this with my life.”
Ryan Kidd: “Icecone: brought me over from Australia; connected me with the top experts, funders in AI safety; empowered me to scale MATS, LISA.”
A handful of career advising options for experienced professionals are available: Successif, High Impact Professionals, SteadRise, and Consultants for Impact.
General events, such as ControlConf, have been a superb way to meet with people that would otherwise be hard to find. However, it’s rare to find resources explicitly directed towards experienced professionals so that they can understand the norms present and get the most value out of these events.
Our step towards a solution
After people first learn about AI safety, it’s often unclear what to do next. Even when it is clear, there exists little social proof to follow through with it for people outside of core AI safety spaces. Lateral Workshop is our solution: a three-day weekend for experienced professionals in Berkeley, California. We will hold 1:1s, small group discussions, and talks from AI safety experts. After the weekend, we will continue to provide support for attendees as they continue on their AI safety journey. Travel, meals, and accommodation for attendees will be covered.
The workshop will feature talks and discussions from individuals at organizations like AI Futures Project, Coefficient Giving, MATS, Kairos, BlueDot, MIRI, Macroscopic Ventures, Americans for Responsible Innovation, Constellation, FAR.AI, UK AISI, and more.
Lateral will seek to provide the following benefits:
Field Legibility: Bringing in speakers and guests who can speak about their role, their day-to-day work, and their organization as a whole. This can quickly de-risk the ecosystem for participants, and make it feel much more real. Bringing in high quality speakers with well-informed, well-thought out views on AI safety and why it matters.
Context-building: Bringing participants into frequent contact with existing AI safety employees, both during the workshop and at a mixer one evening of the weekend. The goal is not just increasing people’s networks but allowing them to better understand the core priorities of the field.
Financial Security: Making candidates aware of career transition grants and typical AI safety salaries, so they can better understand their options from a financial perspective.
Understanding Potential Contributions: Showing attendees that they could meaningfully contribute to impactful work in AI safety by exposing attendees to relevant roles and workshopping personal pathways to impact.
Excitement: Drumming up general excitement in potential talent, as in-person multi-day events have a proven track record of doing.
Accountability: Post-workshop virtual programming and check-ins keeps promising attendees accountable for completing the journey they set out on, increasing the likelihood that they will continue learning, working on personal projects, networking, and eventually land an impactful role in AI safety.
Who should apply
Applicants should have:
Working knowledge of AI safety and the ability to articulate their concerns around serious risks from AI. This could mean having completed a BlueDot course, being an avid LessWrong reader, following AI safety blogs, listening to the 80,000 Hours podcast, etc…
Clear signals of high competency in their current career
A desire to improve the world, and the understanding that this might be done in unintuitive ways
Examples of object-level skills:
Building functions and teams from scratch
Governmental experience, emergency management, national security
Cybersecurity, hardware
Communications: media relations, technology communications, campaigns
Law, finance, and operations
Headhunting and executive search
Convening: running large events, navigating politics of bringing multiple stakeholders together
Coalition building and advocacy
Leadership, executive support
If you are interested in attending Lateral Workshop, apply here by the end of the day on August 9. If you know someone who would be a great attendee, please refer them here. If you can’t make September 11–13 but want to be notified about future iterations, fill out our expression of interest form.
If you have other comments, questions, or concerns, leave them below or reach out at contact@lateralworkshop.org.
This project is being completed through the Generator Residency, a program to help upskill AI safety generalists and fill talent needs in the ecosystem.
Announcing Lateral Workshop for experienced professionals moving into AI safety
Link post
TL;DR: Lateral Workshop is a program in Berkeley from September 11–13 for mid-career and senior professionals who are interested in transitioning into AI safety work to meet people in the field and explore how they can fit in. If you have experience in another field and are interested in translating that experience into impact in AI safety, apply here by August 9th!
If AI safety is to keep scaling, we’ll need a lot of people that the field currently doesn’t supply well. People who know how to run 30+ person organizations. People who have deep connections in DC. People with specialized knowledge in cybersecurity or verification.
Professionals from outside AI safety have those skills, but it’s been difficult for them to make the switch. The professionals we have spoken with rarely lack relevant skills. More often, they struggle to understand how the field works, where they could contribute, and how to become credible candidates. Here are some of the specific barriers experienced professionals face while trying to transition:
Difficulty in understanding the key priorities of the field
Lack of networks within the field and poor understanding of where to start to build them
Lack of free time to invest in time-intensive upskilling and context building
Financial risk (getting a role is competitive, and the field is insular enough that rejections from roles are correlated)
Difficulty evaluating a young and unfamiliar field that lacks many of the conventional markers of professional maturity
Some professionals do navigate these barriers through persistent self study and deep attempts to understand the field. But requiring every candidate to independently reconstruct the field is slow and unnecessarily costly. AI safety has reduced similar friction for students by building clearer early-career pipelines; we think experienced professionals could benefit from a similar investment in talent pipelines and infrastructure.
Even after some contact with the field, it has been hard to hire experienced professionals:
In virtually all cases, the reason organizations must hire experienced talent in the first place is because the role has higher stakes. A good leader can make an organization go well, while a poor leader can devastate culture and sink an organization quickly.
Organizations are currently hesitant to take bets on people. This may be a good policy for individual organizations to follow, but can simultaneously be negative on the ecosystem level.
Hiring rounds index heavily on in-group signals that are correlated, but not necessarily causal to impact. These disadvantage people who come to AI safety from outside the main cultural context.
Professionals are sometimes poorly calibrated on which of their skills are useful—many of their specialized technical skills are less applicable to AI safety, but the tacit people-modeling knowledge they have is applicable.
Here are some excerpts from example roles requiring experienced talent:
“7–8+ years of senior communications experience … building or owning crisis communications frameworks, with an understanding of reputational risk management” – Head of Communications & PR, Apollo Research
“You’re a strong engineer with solid infra fundamentals. You can dig into unfamiliar systems, debug from logs, and identify and fix performance bottlenecks” – Member of Technical Staff, Evaluation Execution, METR
“Multiple years of relevant research or policy experience and a track record of producing high-quality work” – Research Fellow, GovAI
“Closer to 10 years [of experience] for a COO profile.” – Chief Operating Officer, Encode
“Own initiative areas end-to-end — from recruiting founders and designing organizations to commissioning products and running fellowships” – Principal, Project Development, Future of Life Foundation
“Write a coalition letter and get twenty organizations to sign it, ranging from religious groups to workers’ rights advocates” – Policy Advisor, Encode
Pipeline legibility
A research-oriented person has a relatively defined ladder: BlueDot → ARENA → SPAR → MATS → junior researcher → senior researcher. For mid-career non-research talent, the paths look like this:
Founder: BlueDot → Catalyze or another incubator → hopefully get funding
Operations person: BlueDot → get hired, somehow
Generalist: BlueDot → do unspecified things for a year → Generator
There’s a missing intermediate step between introductory content and the programs that place people. For professionals, there are few clear steps after a BlueDot course.
It’s not just about content learning: for early-career people, university groups provide the grounds to transmit context, culture, and motivation, and there’s nothing similar for professionals. Successful mid-career transitions have happened by individuals who have forged their own paths, because nothing existed for them.
Past solutions
Workshops are able to significantly affect the career paths of early-career participants. Jacob estimates he got involved in AI safety 3 months earlier than he otherwise would have had he not attended an AISST/MAIA retreat. In The Case for AI Safety Capacity Building Work, Asya Bergal lists some testimonials:
Max Nadeau: “Going to the Impact Generator workshop. That was, like, extremely random. And, like, resulted in a major acceleration to my career.”
Rachel Weinberg: “Helping with Future Forum, which came from meeting Leilani at Impact Generator”
Gabriel Wu: “[Having] the opportunity to visit Constellation”
Aric Floyd: “After Icecone, I was pretty bought in on I should do something about this with my life.”
Ryan Kidd: “Icecone: brought me over from Australia; connected me with the top experts, funders in AI safety; empowered me to scale MATS, LISA.”
A handful of career advising options for experienced professionals are available: Successif, High Impact Professionals, SteadRise, and Consultants for Impact.
General events, such as ControlConf, have been a superb way to meet with people that would otherwise be hard to find. However, it’s rare to find resources explicitly directed towards experienced professionals so that they can understand the norms present and get the most value out of these events.
Our step towards a solution
After people first learn about AI safety, it’s often unclear what to do next. Even when it is clear, there exists little social proof to follow through with it for people outside of core AI safety spaces. Lateral Workshop is our solution: a three-day weekend for experienced professionals in Berkeley, California. We will hold 1:1s, small group discussions, and talks from AI safety experts. After the weekend, we will continue to provide support for attendees as they continue on their AI safety journey. Travel, meals, and accommodation for attendees will be covered.
The workshop will feature talks and discussions from individuals at organizations like AI Futures Project, Coefficient Giving, MATS, Kairos, BlueDot, MIRI, Macroscopic Ventures, Americans for Responsible Innovation, Constellation, FAR.AI, UK AISI, and more.
Lateral will seek to provide the following benefits:
Field Legibility: Bringing in speakers and guests who can speak about their role, their day-to-day work, and their organization as a whole. This can quickly de-risk the ecosystem for participants, and make it feel much more real. Bringing in high quality speakers with well-informed, well-thought out views on AI safety and why it matters.
Context-building: Bringing participants into frequent contact with existing AI safety employees, both during the workshop and at a mixer one evening of the weekend. The goal is not just increasing people’s networks but allowing them to better understand the core priorities of the field.
Financial Security: Making candidates aware of career transition grants and typical AI safety salaries, so they can better understand their options from a financial perspective.
Understanding Potential Contributions: Showing attendees that they could meaningfully contribute to impactful work in AI safety by exposing attendees to relevant roles and workshopping personal pathways to impact.
Excitement: Drumming up general excitement in potential talent, as in-person multi-day events have a proven track record of doing.
Accountability: Post-workshop virtual programming and check-ins keeps promising attendees accountable for completing the journey they set out on, increasing the likelihood that they will continue learning, working on personal projects, networking, and eventually land an impactful role in AI safety.
Who should apply
Applicants should have:
Working knowledge of AI safety and the ability to articulate their concerns around serious risks from AI. This could mean having completed a BlueDot course, being an avid LessWrong reader, following AI safety blogs, listening to the 80,000 Hours podcast, etc…
Clear signals of high competency in their current career
A desire to improve the world, and the understanding that this might be done in unintuitive ways
Examples of object-level skills:
Building functions and teams from scratch
Governmental experience, emergency management, national security
Cybersecurity, hardware
Communications: media relations, technology communications, campaigns
Law, finance, and operations
Headhunting and executive search
Convening: running large events, navigating politics of bringing multiple stakeholders together
Coalition building and advocacy
Leadership, executive support
If you are interested in attending Lateral Workshop, apply here by the end of the day on August 9. If you know someone who would be a great attendee, please refer them here. If you can’t make September 11–13 but want to be notified about future iterations, fill out our expression of interest form.
If you have other comments, questions, or concerns, leave them below or reach out at contact@lateralworkshop.org.
This project is being completed through the Generator Residency, a program to help upskill AI safety generalists and fill talent needs in the ecosystem.