The Coming Wave
DeepMind co-founder on containment; the rival lab's view.
Summary
Suleyman co-founded DeepMind and later Inflection AI, and argues that artificial intelligence and synthetic biology together constitute a wave of technology that will be extraordinarily powerful, will spread rapidly and cheaply, and cannot be contained by any currently available mechanism.
His central problem is what he calls the containment problem. He argues that these technologies have four features making containment unusually difficult: asymmetry, meaning small actors can produce large effects; hyper-evolution, meaning development is faster than institutional response; omni-use, meaning the same capability serves beneficial and harmful purposes and cannot be separated; and autonomy, meaning systems increasingly act without direct human instruction.
He argues that the incentives running against containment are overwhelming, including geopolitical competition, commercial return, research prestige and genuine benefit, and that historically no technology of substantial value has been successfully suppressed. His proposed response has ten components including technical safety research, audits, choke points at the level of hardware and compute, developer accountability, business model changes, government capability building, international treaties, a culture of restraint within the research community, public participation, and coherence across all of them. He is explicit that he does not consider any of them sufficient.
The book is written by someone with substantial commercial interests in the technology he is warning about, which he acknowledges. Critics argue that the framing serves incumbent interests by favouring regulation that established firms can afford, and that the treatment of near-term harms is thin relative to the speculative risks. The containment framing and the four features are the analytical contribution.
Key takeaways 8
- Artificial intelligence and synthetic biology together form a wave whose containment is harder than that of any previous technology.
- Asymmetry means small actors can produce effects previously requiring state resources, which changes who must be constrained.
- Omni-use means the same capability serves beneficial and harmful purposes and the two cannot be separated at the technical level.
- Development is faster than institutional response, which means regulation addresses the previous generation of capability.
- No technology of substantial value has been successfully suppressed historically, which sets the base rate against containment.
- Hardware and compute represent the most practical choke points, since they are physical, concentrated and countable.
- Suleyman proposes ten components for containment and states explicitly that none of them is sufficient.
- The author holds substantial commercial interests in the technology he is warning about, which he acknowledges and which shapes the argument.
Who should read it
Investors and founders in artificial intelligence and biotechnology who want the containment framing from someone inside the industry, with his interests in view.
Who recommended it 3 · 3 direct
“Everything is about to be changed by AI in a radical, scary way”
Similar books 6
Chosen by shared topics, the entries that hold them, and the people who recommended both.
The struggle to make machine learning systems do what people actually intend.
Recommended by: Sam Altman
The OpenAI-DeepMind race that set the shape of Nvidia's customer base.
Recommended by: Andrew Wilkinson
How Hinton, LeCun and Bengio kept neural networks alive until deep learning took off.
The rationalists around LessWrong and their fear of artificial intelligence risk.
Where it sits in the reference 3
Topics