A 7-step plan to have impact in AI safety – from career advisor Matt Beard

Field Context and Timeline

• The AI safety field has grown by approximately 3,000% over the last eight years.
• Matt, a former advisor at 80,000 Hours, asserts that an efficient and driven individual can pivot into AI safety within three months.
• Hiring managers and experts endorse a standardized set of steps for entry, though efficiency depends on prior background.

Foundational Knowledge Requirements

• Candidates should complete dozens of hours of foundational reading via daily habits (e.g., 30 minutes daily) or reading groups.
• Essential resources include AI risk essential reading lists, problem profiles, blue dot courses, and foundational papers on the alignment problem.
• Goal: Achieve the ability to hold a 30-minute conversation at a conference demonstrating understanding of core problems and potential solutions.
• Specific Knowledge Targets
  • Articulate a view on the speed of AI progress over the next few years.
  • Identify which specific problems are most pressing based on that progress timeline.
  • Form opinions on the theories of change for various AI safety organizations.
• Candidates must be prepared to discuss their reasoning for their views; deference to experts is acceptable, but forming and owning one's own views is required.
• Hiring managers often reject candidates who fail to communicate sufficient "AI safety context" or mission alignment.

Skill Identification and Testing

• Broad contribution categories include technical AI safety, AI policy, strategic operations, scaling, field building, and communication.
• Career reviews and job postings should be analyzed to identify in-demand skills.
• Experimental Approach
  • Candidates should run short, time-bounded "cheap tests" (e.g., weekend projects) to empirically determine fit and interest.
  • Examples of experiments include taking a technical course, writing a blog post about a think tank report, building a data dashboard, or hosting a local event.
  • Feedback from these experiments should be used to narrow down a career path.

Application Strategy

• The field features a high density of incubators and fellowship programs compared to other industries.
• Fellowship Resources
  • 80,000 Hours maintains specific lists for technical AI safety and AI policy fellowships.
  • AISafety.com and EmergingTechPolicy.org offer additional databases.
• Fellowships provide quick feedback, concrete project experience, and referrals.
• Application Tactics
  • Set up alerts on the 80,000 Hours job board; some roles close within weeks.
  • Divide applications into categories: "dream jobs" (for learning), "core focus," and "career capital" roles outside immediate AI safety.
  • Submit expressions of interest to top organizations if relevant experience exists.

Networking Protocols

• Networking is framed as a "research project" to mitigate the inefficiency of talent markets and resume screening.
• Referrals and informal signals (e.g., coffee chats) can significantly speed up the initial application review stage.
• Methods of Engagement
  • Attend local meetups, online forums (LessWrong, EA forum), Twitter, and events.
  • Send succinct, engaging cold emails regarding a person's work.
  • Relocate to hubs like San Francisco, London, or DC, or attend major conferences like Effective Altruism Global.
• Networking provides tacit knowledge and helps test personal fit for specific roles faster than self-study.

Deliberate Practice and Portfolio Development

• Candidates facing barriers should engage in deliberate practice with specific goals and feedback loops.
• Learning should be accelerated by "learning by doing" using AI models for feedback and design.
• Side Projects
  • Projects need not be novel contributions but must demonstrate skill building and initiative.
  • If direct experience is lacking, candidates should consider intermediate jobs in software engineering, operations, or advocacy to build capital.

Public Work and Feedback Loops

• 80,000 Hours surveyed nearly 40 AI safety organizations, which strongly emphasize publishing outputs (blogs, posts, code, videos).
• Public work serves as proof of competence, proactivity, and "personal agency" to hiring managers who cannot see a candidate's internal thought process.
• Feedback Mechanism
  • Publishing work tightens feedback loops, allowing candidates to identify skill gaps or communication issues months earlier than in private study.
  • Negative feedback is treated as valuable data for pivoting or refining skills.

Autonomy and Initiative

• Candidates should not wait for permission to perform the work they want to do.
• Case Study
  • A candidate named Luca was hired by 80,000 Hours after building a side project (AI Safety Talent Network) that addressed a market gap; the team discovered him via his public work.
• Proactively solving problems or building tools can lead to full-time offers even without prior formal employment in the sector.

Psychological Considerations

• Imposter Syndrome
  • Advice: Compare oneself only within the right reference class (peers, not senior experts) and verify empirical progress.
• Burnout and Sustainability
  • Sustainable effort (e.g., 30 minutes daily) is more effective than frenzied bursts followed by quitting.
• Stress Management
  • Rejection is common in this competitive field; candidates are encouraged to be patient and kind to themselves while acknowledging the high stakes of the work.

Resource Recommendations

• Candidates are directed to the 80,000 Hours Advising page, the free AI career chatbot, and options to apply for personalized advising sessions.