Summary

A supermind is a group acting together in a way that seems intelligent. Nearly every meaningful problem is solved by nested groups, from brains to companies to markets. Their collective intelligence depends on individual ability, the ability to work well together, and cognitive diversity; social perceptiveness and evenly distributed contributions matter more than satisfaction or motivation. Groups coordinate through hierarchies, democracies, markets, communities, and ecosystems, each with different costs and benefits. IT can make superminds smarter by involving more people and enabling new forms of organization. Smarter institutions combine liquid democracy, prediction markets, reputation systems, crowdsourcing, and hyper-specialized work. The long-run endpoint is a global mind whose goals reflect its most powerful superminds – making governance, human-computer synthesis, and the distribution of power central design problems.

“In the long run, there is a direction – an arrow in human history – in which technologies arise that allow richer forms on non-zero-sum iterations – that is, interactions that all participants are better off for having interacted.”

“If materialism is true, the United States is probably conscious.”

Key Takeaways

Notes

Introduction

• Coordinated problem solving advantages:
  1. Luck often matters
  2. Different people know different things
  3. Different people think differently
• Nearly every problem is solved by groups of people (even a sandwich via supply chain)
• Most of brain is devoted so social intelligence - groups evolutionarily preferred
• Information technology, industry
• Can measure group intelligence and consciousness
• Computer groups who were most effective had interaction patterns matching conscious human brains
• 4 species of superminds:
  • Hierarchies: authority
  • Democracies: voting
  • Markets: voluntary exchange
  • Community: informal consensus + shared norms
• Ecosystems: decisions made based on who has most power + ability to reproduce
• Should create groups of collectively intelligent humans + AIs

1: What Are Superminds?

1: Recognizing Superminds

• Smith: markets’ invisible hand
• Supermind: a group of individuals acting together in a way that seems intelligent
• Many layers / nesting dolls
  • Eg. human brain → store → market
• Observer must define evaluation in order to assess intelligence
• Varying definitions of intelligence
• Specialized intelligence: the ability to achieve specific goals effectively in a given environment
• General intelligence: the ability to achieve a wide range of goals effectively in different environments
  • ~ versatility, adaptability
• 4 components:
  1. A group
  2. Some actions
  3. Some interconnections b/w actions
  4. Some goals + evals

2: Collective Intelligence

• G score intelligence is 30-60% correlated with any task performance
• Eg. study: more correlated with job success than trials, references, interviews, academic achievement
• Types of group tasks:
  1. Generating
  2. Choosing
  3. Negotiating
  4. Executing
• Same with groups: group intelligence predicts ~45% of misc task performance
• Collective intelligence: ↑
• Individual mean + max intelligence correlated
• Not correlated:
  • Satisfaction with group
  • Motivation to see group succeed
  • Comfortableness in group
• 3 significant factors:
  • Average social perceptiveness of group members
    • Social intelligence: ↑ measured through face + eyes emotion test
  • Evenly distributed contributions
  • Proportion of women
• EQ carries over from in-person to over text
• Cognitive diversity important, too
• 3 cognitive styles:
  1. Verbalizers: reasoning with words
     • Humanities
  2. Object visualizers: overall properties of images
     • Visual arts
  3. Spatial visualizers: analyzing part by part (eg. arch blueprint)
     • Engineering
• Best combo: not too homo to hetero - right balance
• 3 key determinants:
  1. Individual ability
  2. Ability to work well together
  3. Cognitive diversity
• Thomas Edison → GE: only company included in 1896 → current Dow Jones
• GE, Apple have high collective general intelligence

2: How Can Computers Help Make Superminds Smarter?

3: How Will People Work with Computers?

• AI: tools → assistants → peers → managers
• Tools increase specialized intelligence
• Primitive peers, eg. online bots
• Primitive managers, eg. stoplight
  • Mechanical Turk
• Human + agent coordination

4: General Intelligence

• AI: intelligence exhibited by machines
• “I don’t have a lot of patience with this argument”
• Progress in AI is notoriously hard to predict
  • Has seemed about 15-25 years away for the last 60 years
• Approaches
  • Common sense rules
  • Big data
  • ML
• IBM Watson’s society of agents
• “Humans in the loop” → “computers in the group”

5: How Can Groups of People and Computers Think More Intelligently?

• Imagine “perfect intelligence” ideal state
• 5 cognitive processes of intelligence:
  • Decision
  • Create possibilities
  • Information gathering
    • Sensory
    • Memory
  • Learn from experience

3: How Can Superminds Make Smarter Decisions?

6: Smarter Hierarchies

• Hierarchy: where people in authority make decisions that their subordinates are required to follow
• Google is a highly automated hierarchy
• Motivation, creativity, flexibly becoming more important → decentralized decision making
• As-hocracies: ↑
  • Downside: informal power derived from reputations (community)

7: Smarter Democracies

• Democracy: decisions are made by a vote of its members
• Two methods:
  • Direct: impractical
  • Representative:
• Liquid democracy: ↑ best of both
  • Use online systems
  • Eg. delegate to your or other’s AI agents
• Can use tech to do 3 things:
  1. More fine-grained delegation
  2. More accurate predictions of what will be true using data
  3. More intelligent decisions be separating the two functions more clearly

8: Smarter Markets

• Market: decisions made by participants mutually agreeing to trade resources with one another
• Aggregate of decisions determine allocation of resources
• First supermind where no participant sees the whole problem for which the group is trying to make the decision
• Completely blind to prices consumers do not have to pay
  • Solution: gov’t oversees market
• Markets are useful for making predictions
• Prediction markets let you express your opinion more precisely
  • Includes bots as well
• Markets have a general intelligence

9: Smarter Communities

• Community: decisions are made by informal consensus or according to shared norms, both of which are enforced by reputations and access to resources
• Eg. Wikipedia
• IT fragments some communities and unites others
• Idea: online public disputation
  • Can lay out the structure of arguments
• Cyber-socialism
  • Reputation-based
  • Tradeoff is loss of privacy
  • Eg. China social credit system

10: Smarter Ecosystems

• Ecosystem: a supermind with no coordinated decision making
• Bowling Alone
  • Markets are controlling more time and communities less
• Ecosystems decide who survives
• Primary focus is often replication (via memetics)
• Superminds have a will of their own, but generally follow will of the individuals
• In the long run, people can generally choose their supermind, thus ↑
• Eg. the market for movies > the market for communities
• The principle of evolutionary utilitarianism: ecosystems generally in the long run generally try to provide the greatest good for the greatest number of people

11: Which Superminds Are Best for Which Decisions?

• Net benefit: benefit - cost (to maximize)
• First time all 5 have been systematically compared

1. Cost of group decision-making:
   • Democracies: high
   • Markets + hierarchies + communities: medium
   • Ecosystems: low

• Markets could be higher or lower based on transaction costs (legal etc.)

2. Benefits of group decision-making:
   • Ecosystems: low
   • Market: medium- (must be positive sum)
   • Communities: medium
   • Hierarchies + democracies: high

• 2 benefits ↑:
  1. Cooperation:
     • Less conflict
     • Tragedy of the commons
     • Reciprocal altruism
  2. Bigness:
     • Power of size
     • Economies of scale
     • Scope
     • Specialization

3. Distribution of benefits: (utilitarian)
   • Hierarchies + ecosystems: low
   • Democracies + communities: med
   • Markets: high

• Pareto optimal: can’t make someone better or worse off by trading
• Markets most prevalent now because they most take advantage of economies of scale and specialization
  • Why? New technology
• Governments managing markets:
  • Enforce contracts
  • Truth in advertising
  • Health + safety standards
  • Services for natural monopolies (eg. roads and national safety)
  • Income redistribution
  • Research for collective benefit
• How IT affects distribution of “”:
  1. Larger groups
  2. Likely to ↓ cost of decision-making
• Thus, markets and democracies will rise

4: How Can Superminds Create More Intelligently?

12: Bigger Is (Often) Smarter

• 2 ways new tech can make superminds smarter:
  1. Involving more individuals
  2. New ways of organizing
• Eg. crowdsourcing
• Brute force effect: ↑
• Wisdom of the crowds
  • Median of independent guesses
• Study: those further away from the field are more likely to solve crowdsourced problems

13: How Can We Work Together in New Ways?

• 3 aspects of specialization:
  1. Dividing the work in new ways
  2. Assigning tasks in new ways
  3. Coordinating interdependencies among tasks in new ways
• Eg. coding micro-tasks, Mechanical Turk
• Likely to be much more hyper-specialized online work
• Eg. better job matching
• Eg. ebay for jobs (Mercor?)
• Dependencies:
  • Flow
  • Sharing
  • Fit

5: How Else Can Superminds Think More Intelligently?

14: Smarter Sensing

• Big data / IoT
• Privacy + info as property
• Information sharing issues
• Prediction markets

15: Smarter Remembering

• Eg. medical community supermind

16: Smarter Learning

• Collective learning in ecosystem of superminds
• Explore-exploit tradeoff

6: How Can Superminds Help Solve Our Problems?

17: Corporate Strategic Planning

• 3 types of differentiation:
  1. Cost leadership
  2. Differentiation
  3. Focus (customer segment)
• Idea funnels
• Prediction markets, voting, etc.
• Evaluating disruptive ideas (from VCs)
• Spreadsheets, simulations, Bayesian networks

18: Climate Change

• Social norms won’t fix
• Hierarchical gov’t regulation too coarse
• Markets don’t account for externalities
  • Solution: carbon tax - but hard to pass regulation
• Global gov’t (eg. UN)

19: Risks of Artificial Intelligence

• Burden of proof is high for AI replacing jobs unlike prev tech revolutions
• 3 unique human capabilities:
  1. General intelligence
  2. Interpersonal skills
  3. Certain physical skills
• Social > cognitive intelligence
• Depends on elasticity of supply, demand, and substitutions
• Entertainment (EAP)
• Status symbols (BS jobs)
• Learning on the job
• Income redistribution
• FoLI + ASI
• Supermind needed: hierarchies + governing democracy
• Need global governance (better than UN)
• 3 actions now:
  1. Legal responsibility for action by automated systems
  2. AI attacks as war crimes
  3. Human computer synthesis

7: Where Are We Headed?

20: Hello, Internet, Are You Awake?

• Can groups be conscious?
• Components:
  1. Awareness: reacts to stimuli
  2. Self awareness: reacts to changes in itself
  3. Goal-directed behavior: intentional action to achieve goals
  4. Integrated information:
  5. Experience: “if there is something it is like to be that entity”
• Apple as a conscious entity
• “If materialism is true, the United States is probably conscious”
• Do markets count if the supermind never articulates this goal
  • “Thus, in this sense some superminds are conscious and others probably aren’t”
• Need objective, systematic ways of measuring consciousness
• Integrated information theory: consciousness associated with integrated information (Phi) - information the system generates that’s greater than the sum of its parts
• Eg. conscious perception of anything (eg. digital photograph pictures) generates integrated information in that system
• Phi for the Internet is growing - “it’s waking up”

21: The Global Mind

• “In the long run, there is a direction - an arrow in human history - in which technologies arise that allow richer forms on non-zero-sum iterations - that is, interactions that all participants are better off for having interacted.”
• The global mind / the global brain: ↑ endpoint of evolution on this earth
• Started with biology
• (Basically Homo Deus dataism)
• What does the global mind want? It’s an ecosystem, so whatever the most powerful superminds in it want
• Gaia
• Introspection
• “Is developing such a mind our destiny? I think it is.”