Worldviews, altruism, and embracing variance (with Emmett Shear)

Jun 12, 2024 Episode Page ↗
Overview

Spencer Greenberg speaks with Emmett Scheer about intersubjective concepts, distributed systems, optimism bias, and leading successful tech startups, exploring his nuanced worldview and practical lessons from entrepreneurship.

At a Glance
29 Insights
1h 20m Duration
25 Topics
5 Concepts

Deep Dive Analysis

Emmett's Worldview: Complexity and Nuance

Ideology vs. Worldview: Key Distinctions

Standard Silicon Valley Worldview Stereotype

Personal Journey: Understanding Money's Role

Personal Journey: Re-evaluating Democracy's Purpose

Intersubjective Concepts as Coordination Mechanisms

The Essential Role of Middle Management

Democracy as a Safety Valve, Not Policy Aggregator

Proposal: No Confidence Voting for Legislators

Critique of Utilitarianism and Global Solutions

Importance of Local Knowledge and Humility in Aid

Optimism Bias and Active Inference Explained

Why People Consistently Miss Daily Goals

Variance: When It's Good, When It's Bad

Entrepreneurship's Biggest Challenge: Problem Identification

Startup Near-Death Experiences and Rapid Change

The Problem with General Startup Advice

Effective Learning from Users and Customers

Riding Technology Waves vs. Blind Trend Following

Y Combinator's Evolution and Enduring Strengths

Experience as OpenAI Interim CEO

Reflecting on the OpenAI Board's Decisions

Critique of the Effective Altruism Movement

San Francisco: A City Embracing Variance

Approaching AI Dangers: Tools vs. Self-Improvement

Worldview vs. Ideology

Worldviews are complex, messy gestalts that are absorbed over time and stretch across multiple domains, reflecting how someone truly engages with the world. Ideologies, in contrast, are fundamentally simpler, single-lens frameworks that offer a more compressed and often less nuanced perspective.

Intersubjective Concepts

These are concepts like money or democracy that derive their meaning and function from collective human belief and agreement. They are coordination mechanisms where information and value reside within the network of people, rather than being inherent in the individual components themselves.

Coordination Mechanisms

Systems such as money, democracy, or management structures that enable multiple independent agents to cohere and move in a common direction. They solve the problem of how individual components can act effectively without full global information, creating good global results from local coordination.

Optimism Bias (Active Inference)

In the context of active inference, agents anticipate their own survival and desired future states (like maintaining homeostasis) and take actions to make those expectations come true, as they dislike being surprised. This bias is crucial for motivation and persistence, as failing to anticipate positive outcomes leads to learned helplessness and inaction.

Variance in Outcomes

The degree of deviation from an average outcome. While sometimes undesirable (e.g., in precision engineering), variance can be beneficial in situations with power-law returns (like startup investing) where the greatest gains are at the extreme positive end. Exploring novelty inherently generates both failures (negative variance) and disproportionate successes (positive variance).

?
What is the difference between an ideology and a worldview?

An ideology is a simpler, single-lens framework, whereas a worldview is a complex, messy gestalt that encompasses a broad array of topics and is absorbed through extensive interaction, reflecting someone's general approach to things.

?
Why is middle management important in a company?

Middle management is crucial for propagating information both top-down (fragmenting vision) and bottom-up (abstracting details), reconciling gaps between strategic vision and ground-level reality, and enabling coordination in large organizations that cannot be run as pure democracies.

?
What is the true purpose of democracy?

Democracy primarily serves as a safety release valve, allowing citizens to remove bad leaders when conditions become sufficiently poor. It is not designed to aggregate people's opinions about specific policies, as most citizens lack well-informed opinions on a broad range of issues.

?
Why do people consistently get less done than they expect each day?

People often set goals (like productivity or healthy habits) that are above their current point of homeostasis, meaning they are constantly fighting against a gradient pulling them back. This leads to consistent underachievement because they are aiming for a higher state and will typically fall short rather than overshoot.

?
What is the best way for startup founders to learn from users/customers?

Founders should approach user conversations with an open mind, seeking to deeply understand users' problems and their world *before* formulating a solution. This 'learning' approach is more effective than merely 'validating' a pre-conceived idea, which cannot make the idea better.

?
Is it a good strategy for startups to follow new technology trends (e.g., AI, blockchain)?

Yes, riding major technology waves (like the internet or mobile phones) is often a good strategy because it indicates a large, emerging opportunity. However, this differs from blindly following second-order trends, which involves copying what everyone else is doing and increases the risk of failure.

?
What was it like for Emmett Shear to be interim CEO of OpenAI for 2.5 days?

It was a very high-stress but exciting experience, where he focused on understanding the chaotic situation, identifying OpenAI's best interests, and navigating towards the best reachable outcome, which was ultimately unwinding the initial board decision.

?
What is Emmett Shear's main critique of the effective altruism movement?

His main critique is against the idea that one can rationally determine and rank globally 'most important' actions or charities in a context-free way. He argues this approach underweights uncertainty, the dramatic need for humility, and the importance of local, embodied knowledge.

?
How should society handle potential future dangers from artificial intelligence?

Society should separate AI dangers into two categories: for AI as a tool, wait for actual harms to arise before regulating; for AI approaching recursive self-improvement, establish a 'sensor network' to track progress and be prepared to slow down or stop development if that critical threshold is neared without clear understanding of its dynamics.

1. Prioritize Local Impact

Focus on making your personal community and immediate connections good, then invest in your neighborhood or city, and only expand efforts to state, country, and global levels once local spheres are thriving due to the dramatic need for humility and groundedness in local information.

2. Anticipate Success Actively

Cultivate a deep, rich anticipation of success to drive action, skill acquisition, and hard work, as this belief is causally linked to achievement because you expect the effort to pay off.

3. Learn Before Ideating

To succeed in problem-solving (especially startups), engage with users in an open-ended way to deeply understand their problems and build an accurate model of their needs before developing solutions, rather than validating pre-conceived ideas.

4. Identify Worthwhile Problems

In entrepreneurship, recognize that the harder and more critical task is figuring out what problems are truly worth solving, rather than merely solving problems efficiently.

5. Cultivate Optimism Bias

Develop an optimism bias by anticipating positive outcomes (e.g., survival, comfort) to motivate action towards achieving those expectations, rather than passively expecting discomfort or failure.

6. Embed Before Helping

When assisting a community, spend months immersed within it, talking, learning, proposing small interventions, and doing things by hand first to gain deep, embodied knowledge before building solutions.

7. Utilize Management Structure

Implement a management structure in organizations to facilitate both top-down vision fragmentation and bottom-up information consolidation, ensuring effective communication and reconciliation of gaps between strategy and ground-level facts.

8. Find a Co-Founder

If at all possible, find a co-founder to start your company with, as this vastly increases your chance of success in entrepreneurship.

9. Ride New Tech Waves

Leverage new, impactful technologies as a strategic approach for startup success, as historically, many large winning companies have emerged by riding significant waves like the internet or mobile phones.

10. Prioritize Company’s Best Interest

As a CEO, your fundamental mandate is to be a steward of the company’s best interest, requiring you to quickly understand the situation and determine what actions serve that interest.

11. Gather Diverse Perspectives

To rapidly make sense of a chaotic situation, talk to a wide range of people including those within the organization, those close to it, and anyone with a new or informed perspective.

12. Prevent Burnout with Impact

Prevent burnout by actively anticipating that your work has meaning and will drive change in the world, as failing to do so makes sustained motivation almost impossible.

13. Set Ambitious Goals

To improve, set goals that are at a higher level of abstraction and above your current point of homeostasis, understanding that you will consistently aim for and likely fall slightly short of this higher target, which is the nature of willpower and growth.

14. Evaluate Charitable Impact

When giving money to do good, care about the impact of your donations and do your best to estimate it quantitatively or qualitatively, as effectiveness matters.

15. Contextualize Personal Impact

Determine what is most important for you to work on based on your personal context, including where you can have the best impact, your location, knowledge, and connections, rather than seeking a universal ‘most important’ problem.

16. Eliminate Known Bad Practices

Identify and eliminate predictably detrimental actions (e.g., not fully committing to a startup) to reduce negative variance without sacrificing potential high-end positive outcomes, as these actions are predictably detrimental.

While noticing a real technology trend is good, avoid copying what everyone else is doing within that trend, as this ‘second-order trend following’ is a good way to fail because the median startup dies.

18. Act on Unique Opportunities

Seek out and act quickly on opportunistic charitable work where you have trusted personal connections, an asymmetric information advantage, and can provide support with less oversight for maximum leveraged impact.

19. Pledge Fixed Charitable Giving

Consider taking a pledge (e.g., 10% of income) for charitable giving, as this can be a sufficient contribution and alleviate guilt about not doing enough.

20. Don’t Limit Measurable Good

Recognize that while funding measurable, low-variance charitable work is good, most high-impact good is not easily quantifiable or repeatable, and don’t limit your efforts solely to what is easily measurable.

21. Deepen Understanding Cycle

Approach complex concepts by first believing for simple reasons, then doubting for complicated reasons, and finally believing again for more sophisticated and nuanced reasons to achieve a deeper understanding.

22. Assess Desired Variance

Understand that variance is not inherently good or bad; assess situations to determine if you need to minimize variance (e.g., precision tasks) or embrace it for potential upside (e.g., innovation, power law returns).

23. Reframe Democracy’s Purpose

Understand democracy primarily as a safety release valve that allows for the removal of bad leaders when things get bad enough, rather than a primary mechanism for aggregating policy opinions or setting direction.

24. Regulate Tech by Actual Harm

For new technologies like AI that create powerful tools, adopt a regulatory approach that waits for actual harms to arise before acting to prevent them, rather than guessing what harms might occur.

25. Monitor Recursive AI Progress

Pay close attention to AI development, especially for signs of recursive self-improvement, by registering progress and verifying how close projects are to this critical threshold, to be prepared to slow down or stop if necessary.

26. Reform Legislative Elections

Advocate for a system where incumbents win by default unless a no-confidence vote removes them, leading to an open race, and oppose term limits for legislators as they empower bureaucracy and parties at the expense of individual elected officials.

27. Absorb Complex Worldviews

To truly understand a person’s worldview, spend a good deal of time with them in person or absorb their general approach by reading their extensive writings across a broad array of topics.

28. Cultivate Startup Persistence

Be persistent in your startup endeavors, as giving up will likely lead to failure; persistence is a generally true and good piece of advice for increasing your chances of success.

29. Attend E-Learning Webinar

Sign up for the free ‘How to Build Your Own Clearer Thinking Style e-Learning Tool’ webinar on June 19th at 1 p.m. Eastern Time to learn secrets of creating effective e-learning modules and get questions answered live.

Worldviews are big, complicated, messy gestalts matching the world itself, which is this big, highly detailed thing.

Emmett Shear

Democracy's point is when it gets bad enough, you can evict the bum, right? Like that's, that's why democracy is good is like, ultimately like it allows you to remove the bad leaders.

Emmett Shear

The dangerous thing is when you go in and you inject a bunch of energy into a system where you're not getting reciprocal energy back out because you think it's going to help.

Emmett Shear

Burnout is when you start failing to anticipate that your work has any meaning or that will have an impact.

Emmett Shear

The easy part was solving the problems and the hard part was figuring out what problems were worth solving.

Emmett Shear

There is no good general startup advice. You will not get from us general startup advice. You will get from us startup advice for your startup.

Emmett Shear

The world is full of unknowns, the idea that you'll have the correct answer is like, no, you won't, you don't even want to have the correct answer.

Emmett Shear
2005
Y Combinator first class year Year Emmett Shear was part of the first Y Combinator batch.
2006
Justin.tv co-founding year Year Emmett Shear co-founded Justin.tv.
2011
Twitch spin-off year Year Twitch spun off from Justin.tv.
2011
Y Combinator part-time partner start year Year Emmett Shear became a part-time partner at Y Combinator.
2.5 days
OpenAI interim CEO duration Duration Emmett Shear served as interim CEO at OpenAI in November 2023.
1%
People getting more done than expected daily Percentage of people in a poll who reported getting more productive stuff done than expected in a day.
72%
People getting less done than expected daily Percentage of people in a poll who reported getting less productive stuff done than expected in a day.
20-25
Startups per YC group Approximate number of startups a YC group partner works with collectively (or ~15 with visiting group partners).
Over 10,000
YC alumni network size Current estimated number of alumni in the Y Combinator network.
10%
Suggested charitable giving percentage Percentage suggested by 'giving what we can pledge' for charitable contributions.