Luisa and Robert Long on how to make independent research more fun (2023)

Host and Guest Introduction

• Kieran Harris introduces Luis Rodriguez as the new permanent member of the 80,000 Hours podcast team.
• The episode features Rob Long and Louisa, recorded independently but framed as a standalone discussion on independent research.
• Harris notes the conversation diverges from their previous episode on "artificial sentience" (Episode 146) to focus on research methodology and career psychology.

Quantifying Uncertainty and Modeling

• Rob Long describes using probability distributions for the "nuclear war report" to handle inputs ranging from 1 to the total number of existing weapons.
• Long utilized "bimodal" distributions to reflect uncertainty where outcomes cluster at extremes (few vs. many weapons used) rather than the middle.
• Long admits to using "uniform distributions" (equal probability) as a baseline when truly ignorant, refining them only when empirical data or theory suggested a "bump" in likelihood.
• The discussion contrasts the nuclear war project, which had concrete empirical inputs (e.g., smoke lofting physics), with "consciousness" research, where model uncertainty is significantly higher.
• Transparency is highlighted as a key strategy: publishing a "guesstimate model" allows others to adjust inputs, reducing the anxiety of being "wrong."
• Long found comfort in the ability to share reasoning and invite counter-arguments, shifting the focus from being the "final word" to facilitating a collaborative model.

Psychology of Research: Imposter Syndrome and Self-Belief

• Long characterizes his struggle as "imposter syndrome" stemming from high standards and a belief that he must solve complex philosophical problems (like AI consciousness) to be successful.
• The conversation distinguishes between "low self-confidence" (Louisa's position) and "high standards" (Rob's position), though both face self-doubt.
• Rob notes that the "rationality community" often fails to balance the belief that "I might solve this" with the necessity of publishing provisional, imperfect results.
• Both speakers argue that researchers should avoid the mindset that work cannot be published until the problem is fully solved, citing Ellie Hassenfeld and Holden Karnofsky's work on effective charity as a positive example of releasing preliminary findings.
• Rob suggests intrinsic motivation comes from the act of "getting less confused" about the world, rather than proving one's genius.

Operational Strategies for Independent Researchers

• Social Accountability: Louisa emphasizes that a "friend or manager" telling a researcher to "send that now" prevents year-long isolation and perfectionism.
• Co-working: Rob describes a workflow where he and co-authors write entirely in Zoom meetings, eliminating the isolation of independent drafting.
• Daily Management: Long advocates for daily "rubber ducking" sessions with a manager to track progress and address blocks immediately.
• AI Assistance: Both speakers use AI (ChatGPT) for brainstorming, "HR-like" encouragement, and generating interview questions, though they acknowledge its limitations compared to human emotional support.
• Environment Control: Long recommends the "Cold Turkey Blocker" app to restrict internet access to Google Docs only, minimizing distractions.
• Self-Diagnosis: The speakers agree that if a researcher consistently feels bad and unmotivated despite fixing their environment, it may indicate a fundamental lack of interest in the topic, signaling a need to change paths.

Career Decisions: Impact vs. Personal Fulfillment

• Rob reflects on a past belief that impact should completely override personal interest, a view influenced by the "don't follow your passion" advice from 80,000 Hours.
• He now acknowledges the limits of self-coercion: he cannot permanently "lock away" his need for intrinsic satisfaction or creative stimulation.
• The speakers propose a "negotiated decision" model where the part of the brain seeking impact and the part seeking interest reach a consensus, rather than one dominating the other.
• Rob notes that ignoring intrinsic interests led to job hopping and reduced long-term impact in his past career choices.
• Both speakers agree that a "negotiated consensus" prevents the internal conflict that arises from bullying the "resistant parts" of oneself.

Specific Career Mistakes and Lessons

• Long-term Focus: Rob identifies a weakness in maintaining focus over months or quarters, despite having strong day-to-day habits.
• Perfectionism as a Bottleneck: Rob's default setting was that a project is "not done until solved," leading to prolonged delays on solo projects.
• External Deadlines: Louisa helped Rob by setting hard delivery dates, contrasting his internal "never done" mindset with the external reality that work must be shared after 6–12 months.
• The "10% Lottery" Fallacy: Long notes that very few independent researchers naturally possess the discipline to finish and share work on time without external deadlines.
• Accountability Mechanisms: The speakers reference a past arrangement where they tied financial stakes to accountability goals to overcome procrastination.
• Normalizing Struggle: They conclude that struggling with these issues is the norm for researchers, and self-flagellation is unnecessary when external structures (like managers) are absent.

Forward-Looking Statements

• The episode suggests that future AI models may eventually serve as effective "rubber ducking" partners for brainstorming, provided they simulate genuine conversation.
• Rob Long directs listeners to his Twitter (@rgblong) and Substack ("Experience Machines") for continued research updates.
• The team encourages listeners to revisit Episode 146 for the full context on the AI sentience topic.