Interview
#1 - Miles Brundage on the world's desperate need for AI strategists and policy experts
- AI policy is expected to evolve from high-level framing to specific proposals and formal models over the next few years, driven by a steady increase in the skill floor as technologies diffuse to individuals with fewer resources.
- Job opportunities in AI strategy and policy are projected to expand significantly within the next few years, with a potential slackening in competitiveness as organizations like DeepMind and OpenAI develop larger teams focused on safety topics.
- Governments are predicted to view maintaining a leading AI edge as a critical national security issue, creating a risk of accelerated arms race dynamics if state actors ramp up capabilities without balancing safety and performance tradeoffs.
- Advanced autonomous drones are anticipated to enable groups such as terrorists or state actors to carry out damage on a larger scale, while mass surveillance risks remain ambiguous regarding whether states or corporations pose the greater threat.
- The AI development landscape is likely to remain concentrated among top organizations due to the concentration of computing power and talent, making it highly improbable for a random individual to out-compete them, though a slowing down among a small number of actors is possible.
- Future risks may be straightforward extensions of current issues or involve qualitatively different challenges, with uncertainty prevailing regarding the specific nature of threats arising from continued AI progress.
- Career pathways for influencing AI policy include roles in military and intelligence bureaucracies to build connections and gain system knowledge, as well as positions in government or non-profits where the agenda is not yet fully defined.
- Organizations such as the Center for the Study of Existential Risk and the Leverhulme Center for Future of Intelligence represent part of a larger, rapidly evolving range of opportunities, with many new entities emerging within the last one to five years.
- Collaboration between technical experts and policy strategists is identified as the most exciting area of work, requiring qualitative analysis and relationship building rather than solely technical problem-solving.
- Specific entities like DeepMind are actively hiring for AI policy roles, while academic programs at institutions like Oxford offer diverse disciplinary backgrounds ranging from human social dimensions to machine learning and cybersecurity to support these efforts.
- Individuals with non-technical backgrounds are expected to find that AI policy benefits from their expertise, provided they are willing to collaborate with complementary skill sets, though there is a risk of discrediting causes by being overly alarmist without sufficient expertise.
- While government involvement is crucial for framing issues and convening discussions, rushing into state-led leadership could negatively impact arms race dynamics, and the exact right location for solving immediate AI policy problems remains unclear due to early institutional development stages.