TL;DR
- The Coming Wave argues that AI and synthetic biology represent a technological wave whose containment—preventing catastrophic misuse while preserving benefits—is the defining governance challenge of this century.
- Suleyman, a co-founder of DeepMind, draws on insider experience to argue that this wave is categorically different from past technology shifts: it is self-improving, proliferating rapidly, and arriving faster than governments, institutions, or societies can adapt.
- The book frames containment not as halting the technology but as navigating a narrow and difficult path between catastrophic misuse and stagnation—a challenge requiring unprecedented coordination between governments, companies, and civil society.
Source Info
- Title: The Coming Wave: Technology, Power, and the Twenty-First Century’s Greatest Dilemma
- Author: Mustafa Suleyman, Michael Bhaskar
- Publication Date: 2023
- Themes:
- AI safety and containment
- Synthetic biology
- Geopolitics and power
- Nation-state stability
- Technology governance
- Existential and catastrophic risk
Key Ideas
- The coming wave is characterized by its breadth (it crosses AI, biology, energy, and robotics simultaneously), its speed (progress compounds faster than institutions can respond), and its asymmetry (small actors can wield capabilities previously available only to nation-states).
- Containment is Suleyman’s central challenge: not stopping AI, but ensuring its development does not lead to authoritarian consolidation, nation-state collapse, or catastrophic misuse by non-state actors.
- The technology is already too advanced to stop, and the window for shaping how it lands—through regulation, norms, technical safety measures, and international agreements—is narrowing quickly.
Chapter Summaries
-
Prologue
- Main Idea: Suleyman opens by establishing the stakes: a transformative technology wave is coming, and the question is whether humanity can contain it.
- Key Points:
- Every major technology wave has had unintended consequences.
- What is different now is the speed, breadth, and autonomous nature of the technology.
- Suleyman frames himself as an insider who believes in the technology’s promise and fears its risks in equal measure.
- Takeaway: The book is a warning from someone who helped build the wave—which makes it different from outside criticism.
-
Chapter 1: The Containment Problem
- Main Idea: The core dilemma is that powerful technology tends to proliferate and resist control, even when the consequences of proliferation are serious.
- Key Points:
- Technologies from nuclear weapons to the internet have demonstrated that containment is difficult but not impossible.
- AI and synthetic biology pose unique containment challenges because they are cheap to replicate, improve autonomously, and are already widely distributed.
- Containment is not zero-or-one: partial, imperfect containment is still worth fighting for.
- Defined Terms:
- Containment: The effort to limit the most harmful applications and proliferation of a powerful technology without eliminating its benefits.
- Takeaway: The difficulty of containment is not an argument against attempting it—it is an argument for starting now.
-
Chapter 2: The Acceleration
- Main Idea: AI progress is accelerating faster than most outside observers appreciate, driven by compounding improvements in data, compute, and algorithmic efficiency.
- Key Points:
- Suleyman traces the exponential trajectory from early AI systems to current foundation models.
- The pace of improvement has surprised even the researchers building the systems.
- Most institutional planning still assumes linear technology change—a dangerous miscalibration.
- Defined Terms:
- Acceleration: The compounding speed of AI capability improvement, driven by simultaneous advances in hardware, data, and methods.
- Takeaway: Planning for AI governance based on today’s capabilities is like planning flood defenses based on yesterday’s water levels.
-
Chapter 3: The Technology of Intelligence
- Main Idea: Large-scale AI systems are approaching and in some domains exceeding human-level performance, with implications that go far beyond any single application.
- Key Points:
- Transformer architectures and foundation models represent a qualitative shift, not just a quantitative improvement.
- AI is now capable of reasoning, coding, scientific hypothesis generation, and persuasion—tasks once considered uniquely human.
- The near-term trajectory points toward systems with agentic, autonomous capabilities.
- Defined Terms:
- Foundation model: A large AI model trained on broad data that can be adapted to many downstream tasks.
- Agentic AI: AI systems that can pursue goals autonomously over extended time horizons without step-by-step human direction.
- Takeaway: The technology of intelligence is no longer a distant prospect—it is an emerging present reality.
-
Chapter 4: The Technology of Life
- Main Idea: Synthetic biology is advancing in parallel with AI and poses its own wave of capabilities—including the potential to engineer pathogens and life itself.
- Key Points:
- Gene editing, protein folding prediction, and DNA synthesis are all becoming cheaper and more accessible.
- The combination of AI and synthetic biology amplifies both the benefits and the risks of each.
- The barrier to creating dangerous biological agents is falling rapidly.
- Defined Terms:
- Synthetic biology: The design and engineering of biological components and organisms using computational and molecular tools.
- Dual use: Technologies or knowledge that can serve both beneficial and harmful purposes.
- Takeaway: Synthetic biology is not a distant threat—it is a present-tense containment problem requiring the same urgency as AI.
-
Chapter 5: The Four Features of the Coming Wave
- Main Idea: The coming wave has four distinctive characteristics that make it more dangerous than previous technology waves.
- Key Points:
- Asymmetry: small actors can access capabilities previously available only to nation-states.
- Hyper-evolution: the technology improves itself faster than humans can track.
- Omni-use: AI and synthetic biology apply across virtually every domain simultaneously.
- Autonomy: the systems can act independently of human direction.
- Defined Terms:
- Asymmetry: The capacity of small or non-state actors to wield disproportionate power through advanced technology.
- Omni-use: The ability of a technology to transform many different domains simultaneously rather than one sector at a time.
- Takeaway: These four features, taken together, make the coming wave qualitatively different from the industrial revolution, nuclear age, or internet era.
-
Chapter 6: The Dilemma
- Main Idea: The central dilemma is that the incentives to develop and deploy the technology are overwhelming, and no single actor can solve the containment problem alone.
- Key Points:
- Companies face competitive pressure to ship; governments face pressure to not fall behind rivals.
- Individuals who refuse to develop AI simply cede influence to those who will.
- International coordination is hard because nations distrust one another and the technology moves faster than treaties.
- Defined Terms:
- Proliferation dilemma: The situation in which the rational self-interest of each actor leads to collective outcomes that harm all actors.
- Takeaway: The containment problem will not be solved by good intentions alone—it requires structural solutions that change the incentives.
-
Chapter 7: The Authoritarian Dream
- Main Idea: Advanced AI dramatically increases the capacity of authoritarian regimes to surveil, control, and direct populations—threatening liberal democracy at a global scale.
- Key Points:
- AI enables surveillance at scale, predictive policing, and information control that previous authoritarian governments could only dream of.
- The technological advantage may be shifting toward concentrated state power rather than distributed individual freedom.
- The race between authoritarian and democratic uses of AI is already underway.
- Defined Terms:
- AI-enabled authoritarianism: The use of AI systems by governments to consolidate political control through surveillance, censorship, and behavioral prediction.
- Takeaway: If containment fails in the governance domain, the most likely outcome is not chaos—it is a new form of stable authoritarianism.
-
Chapter 8: The Nation-State Under Pressure
- Main Idea: AI threatens the capacity of existing nation-states to maintain order, collect taxes, regulate markets, and monopolize force.
- Key Points:
- Economic disruption from AI may destabilize labor markets faster than safety nets can adapt.
- AI-enabled non-state actors (criminal organizations, terrorist groups, rogue companies) gain leverage against governments.
- The social contract underlying democratic governance depends on state competence—which AI disruption may erode.
- Takeaway: Even stable democracies face institutional stress from AI disruption; the question is whether institutions can adapt fast enough.
-
Chapter 9: Three Hard Problems
- Main Idea: Suleyman identifies three particularly difficult sub-problems in containment: technical safety, international coordination, and political economy.
- Key Points:
- Technical alignment is unsolved and may not be solved before systems become dangerously capable.
- International agreements require trust between competing powers who have little of it.
- The political economy of AI—massive profits, diffuse harms—disfavors regulation.
- Takeaway: Each hard problem is solvable in theory; solving them simultaneously under time pressure is the actual challenge.
-
Chapter 10: Grand Bargains
- Main Idea: Effective containment will require grand bargains—between companies and governments, between nations, and between the present and the future.
- Key Points:
- Technology companies must accept binding safety obligations in exchange for continued access and legitimacy.
- Nations must accept verification regimes and shared standards, even under conditions of mutual distrust.
- Democratic societies must accept some limits on AI development in exchange for protection from the worst outcomes.
- Defined Terms:
- Grand bargain: A large-scale negotiated agreement in which multiple parties accept costs in exchange for shared benefits they could not achieve alone.
- Takeaway: Grand bargains are historically rare and difficult—but the alternative to attempting them is accepting uncontrolled proliferation.
-
Chapter 11: A Narrow Path
- Main Idea: The conclusion frames containment as a narrow but navigable path that requires urgent, coordinated action across technical, political, and cultural domains.
- Key Points:
- The path is narrow because too little regulation leaves catastrophic risks unaddressed, and too much halts beneficial development.
- Suleyman argues for mandatory safety testing, international oversight bodies, and technical standards for dangerous capabilities.
- The window for navigating the path is open—but will not remain open indefinitely.
- Takeaway: Containment is hard, urgent, and possible—but only if the people and institutions with power to act treat it as their first priority.