MAP-MAD Spectrum
Move shared technologies from destructive rivalry toward shared prosperity
- Difficulty
- Expert
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 95%
Gawdat places collective choices on a spectrum between MAD, mutually assured destruction, and MAP, mutually assured prosperity. The mechanism is game-theoretic: unchecked competition rewards each actor for accelerating, concentrating control, or externalising harm, even when the combined result is worse. A MAP intervention makes the shared upside or shared danger visible, creates a project or rule set no participant can achieve alone, and distributes access to the result. Gawdat cites nuclear agreements as an imperfect response to common destruction and CERN as an example of countries pooling scientific effort. He proposes a comparable global AI collaboration, but that proposal is advocacy rather than a demonstrated solution. Use the spectrum to design and test cooperation, including verification, holdout behaviour, power asymmetries, and benefit distribution.
Origin
Extracted from The Diary of a CEO
Core principles
- 01Rivalry can turn shared capability into shared risk
- 02Recognised common danger can create cooperation
- 03Large shared projects can distribute benefits
- 04Prosperity depends on governance and distribution, not capability alone
- 05Cooperation must account for holdouts and concentrated power
How to run it
- 1
Name the equilibrium
Describe what rational competitors do today and why their separate incentives create a harmful shared result.
Pro tip Write the payoff for accelerating, cooperating, and defecting.
Watch out Do not assume every conflict is caused by one incentive.
- 2
Define mutual prosperity
Specify the measurable outcome all participating populations should gain, not merely what institutions or owners receive.
Pro tip Include access, safety, and distribution metrics.
Watch out Aggregate growth can conceal concentrated gains and widespread losses.
- 3
Find the cooperation anchor
Identify infrastructure, standards, research, or safeguards that are more credible or valuable when built together.
Pro tip Use a shared problem that no actor can solve efficiently alone.
Watch out A central project can itself concentrate power.
- 4
Design verification and holdout responses
Create transparent commitments, monitoring, and proportionate responses when participants defect or stay outside the agreement.
Pro tip Make compliance observable where possible.
Watch out Rules without enforcement may reward defectors.
- 5
Audit distribution
Measure who receives the capability, wealth, safety, and decision rights produced by cooperation.
Pro tip Include affected people who are not formal participants.
Watch out Shared construction does not guarantee shared prosperity.
In the wild
Gawdat presents CERN as a project too complex for one nation to build alone. Countries contribute scientists, collaborate, and use the resulting knowledge, making it his example of mutually assured prosperity.
→ The example illustrates cooperation around shared infrastructure, without proving that the same governance would transfer cleanly to AI.
Gawdat advocates hosting scientists from around the world to build AI for global benefit rather than national or corporate competition. Bartlett finds the idea compelling while noting directive and unintended-consequence risks elsewhere in the discussion.
→ The proposal supplies a MAP direction but leaves governance, control, and enforcement unresolved.
Common mistakes
Calling growth shared prosperity
More aggregate wealth does not show that citizens, workers, or poorer countries receive the gains.
Ignoring the cooperation controller
A shared institution can reproduce concentration unless decision rights and accountability are designed explicitly.
Assuming goodwill defeats incentives
The framework requires verification and holdout planning, not hope that every actor will cooperate.
Is it for you?
Best for
Governments, industries, or consortia governing technology whose risks and benefits cross organisational boundaries.
Not ideal for
Small unilateral decisions where no meaningful shared-risk or coordination problem exists.
From the transcript
“I call that the MAP-MAD spectrum: mutually assured prosperity versus mutually assured destruction.”
“We need a CERN of AI.”
From the episode
Ex-Google Exec (Mo Gawdat) on AI: The Next 15 Years Will Be Hell Before We Get To Heaven… And Only These 5 Jobs Will Remain!