Breaking into AI Safety: A 3-Month Crash Course
Practical steps to transition into AI safety roles, from foundational knowledge to networking strategies.
1Foundational knowledge in AI safety
The article emphasizes building foundational knowledge through 30 minutes of daily reading or listening, leveraging resources like 80,000 Hours' AI risk reading list and BlueDot courses. This approach helps newcomers grasp core problems and potential solutions, enabling meaningful conversations at AI safety conferences. The goal isn't deep technical expertise but understanding key concepts like the transformer architecture's significance.
2Developing AI safety 'context'
Beyond basic knowledge, the article stresses developing 'context' - forming opinions about AI progress timelines and pressing problems. This involves understanding different organizations' theories of change, like MIRI and CSET's approaches. Hiring managers value candidates who can discuss these topics thoughtfully, even if their views evolve later. This differs from other industries where such deep contextual understanding isn't as critical.
3Identifying in-demand skills
The article outlines several AI safety career paths: technical research, policy, operations, fieldbuilding, and communication. It recommends analyzing job postings and expert career paths to identify transferable skills. The author suggests starting with short experiments (weekend projects) to test fit for different areas, such as taking technical courses or writing about think tank reports.
4Leveraging fellowships and programs
AI safety offers numerous entry points through fellowships and incubators, with the field growing 3000% in eight years. Resources include 80,000 Hours' technical fellowship list and EmergingTechPolicy.org's database. Fellowships provide concrete projects, feedback, and referrals. The article advises applying even if unsure about competitiveness, as the process itself builds valuable skills.
5Strategic application approach
The author recommends dividing applications: 25% for dream roles (learning opportunities), 50% for core focus areas, and 25% for skill-building positions outside AI safety. Setting job alerts is crucial as some roles open briefly. The article notes many candidates underestimate their qualifications and miss opportunities by not applying.
6Effective networking strategies
While networking can feel transactional, the article frames it as research. Referrals save organizations time in hiring, and even casual conversations provide useful signals about candidates' understanding. The author suggests attending meetups, conferences (like EA Global), and relocating to hub cities (SF, London, DC) to access tacit knowledge within AI safety communities.
7Deliberate practice techniques
The article emphasizes deliberate practice: setting specific goals, working at comfort zone edges, and creating feedback loops. It suggests structured courses and side projects, noting AI's growing role in providing learning feedback. While novel contributions aren't immediately necessary, consistent skill-building through projects is crucial for demonstrating capabilities.
8Building a public portfolio
Surveying 40 AI safety organizations revealed strong emphasis on public work: blogs, projects, videos, and AI demonstrations. This visibility helps employers assess candidates' abilities. The article acknowledges the discomfort of public feedback but positions it as essential for career transition, contrasting with traditional industries where private work may suffice.
9Testing fit through small projects
The article recommends starting with time-bound experiments to efficiently explore options before narrowing focus. Examples include technical courses, blog writing, data dashboards, or event hosting. Both positive and negative outcomes (enjoyment vs. disinterest) provide valuable data for career direction decisions in the nascent AI safety field.
10Comparative advantage identification
A key step involves reflecting on existing skills that could transfer to AI safety roles. The article suggests using tools like AI career chatbots and examining admired experts' paths. It highlights the importance of direct conversations with practitioners, which can reveal more in 30 minutes than weeks of self-study about specific roles' realities.
11Intermediate skill-building roles
For those lacking direct experience, the article recommends transitional roles in adjacent fields like software engineering, policy staffing, or communications. These positions build relevant skills while maintaining focus on eventual AI safety transition. The approach acknowledges that side projects alone may not suffice without some professional foundation.
12Urgency and field growth
The article underscores AI safety's rapid expansion (3000% growth in 8 years) and urgency due to catastrophic risk potential. This creates unique opportunities for newcomers despite the field's technical nature. The structured 3-month approach balances quick entry with substantive preparation, differing from traditional career transitions that may require longer timelines.
Original article: 80,000 Hours — Breaking into AI Safety: A 3-Month Crash Course