Interview
#11 - Dr Spencer Greenberg on speeding up social science 10-fold & why plenty of startups cause harm
- Plans include publishing a full transcript with supporting article links, facilitating podcast subscriptions via the "80,000 hours" search term, and providing links to free coaching applications.
- Future content plans involve a study on charity cost perceptions published days prior, an upcoming interview with Tristan Harris, and papers debunking mass automation as the primary cause of current unemployment.
- Spencer Greenberg expects to release the "Uplift" app on a waitlist at uplift.us soon, featuring one to two interactive 30-to-40-minute weekly sessions with homework assignments designed to reduce depression by 50 percent over a month with benefits maintained at six weeks and six months.
- The "Uplift" follow-up study on a six-month basis is expected to confirm that initial benefits were maintained, while the "Guided Track" platform aims to expand from web-only to iPhone and Android deployment and evolve into a language enabling non-programmers to build behavior change tools.
- New programs on clearerthinking.org are imminent, including a "How Rational Are You?" test expected to be shared on Facebook and a "Common Misconceptions" test designed to analyze user confidence levels, with most results indicating average overconfidence.
- The "SparkWave" initiative currently maintains approximately 11 products in development and aims to recruit CEOs for business and marketing roles using actual work tests rather than resumes, operating as a research lab focused on scalable solutions for developing countries like India.
- Recruitment strategies will target specific demographics through "Task Recruiter" to overcome general population skew, leveraging online methods to achieve larger sample sizes at lower costs than traditional lab recruitment.
- Strategic plans involve maintaining 5 to 15 simultaneous projects to facilitate cross-disciplinary solutions, with interpersonal involvement cited as the primary method for preventing project abandonment.
- Machine learning is projected to continue growing in popularity and utility over the next 5 to 10 years, though the field faces risks of being both over-hyped for unsuitable applications and under-utilized for complex problem-solving.
- The effective altruism community is expected to broaden its portfolio beyond current focus areas like existential threats and global health to include mental health and specialized research in nanotechnology and biomedical fields.
- Critical risks identified include the potential for AI monopolies, superior prediction capabilities, and military advantages, with a specific expectation that a 3% probability of massive negative consequences warrants serious attention despite low perceived public risk.
- Estimates for human extinction probability are expected to be 100,000 times higher than the general public's 0.0001% belief, with a median effective altruist estimate of a 10% probability of an event wiping out 10% of the population in the next 50 years.
- Research methodology expectations involve iterative studies consisting of 5 to 7 or 14 iterations rather than monolithic ones, with a proposal to withhold 20% of data in a "vault" to enforce honesty and reduce false positives.
- Academic publishing is predicted to suffer from misaligned incentives favoring top journals, leading to lower replication rates and higher false positives compared to second or third-tier journals, necessitating tools that make releasing data the default action.
- Public perception of charity costs is expected to be skewed by the anchoring effect, with most believing saving a life costs a few dollars rather than the estimated $3,000 to $7,000 range, while also holding contradictory views on the morality of animal suffering.
- Analysis of animal welfare beliefs indicates 99% of people believe animals can suffer, yet only 53% consider it wrong to hurt them for food, a discrepancy expected to drive future behavioral changes regarding factory farming and similar practices.