Thinking tool: Use leverage against weakness

Headshot of Adam Jones

Adam Jones

An effective strategy is to concentrate your relative strength (leverage) against the problem/competitor/opposition's relative weakness.

Leverage might be:

  • inherent strengths you might have
  • a reframing of the problem or some insight into what actually matters
  • concentrated application of effort where it is most effective

Weaknesses might be:

  • things that the opponent is bad at (often it is useful to think about the opponent's strengths and what they have committed or traded off to become strong in that area)
  • a bottleneck in a system

In particular, it is often a mistake to overly focus on eliminating your weaknesses. While it's useful to mitigate some risks, you need to focus on using strengths to win. Good Strategy Bad Strategy gives David and Goliath as an example: David declines Saul's armour - which would only have made him a smaller, slower Goliath - and instead brings a sling, turning the fight into a contest of speed and range against a giant built for close combat.

A vast landscape: on the right, Goliath in heavy armour with spear and shield, towering in the foreground with an army behind him; on the far left, a small David in a plain tunic, already whirling his sling. The whole valley sits between them

The US's strategy in the Cold War offers another good example. Here the US realised:

  • Their strength was technology (from a broadly tech-literate population, capital-intensive economy, dynamic civilian tech sector)
  • The Soviets' weakness was a hypersensitivity to certain threats (such as people overflying their air space) and lack of cost-effective defence technology

The strategy was then to use their technology lead to invest in threats that forced the Soviets to shell out far more resources to counter, and where countering them did not increase Soviet offence capabilities. For example, improving missile accuracy to cause Soviets to rebuild bunkers and overspend on air defence, and improving the quietness of submarines to force the Soviets to pour money into anti-submarine detection. This also required a mindset shift from thinking about 1-5 year timescales to thinking maybe about 10+ year timescales.

Looking down on rows of identical unopened drinks cans packed tightly together, filling the frame

Another example is Crown Cork & Seal, a can manufacturer. Few giant customers (Campbell's, Coca-Cola) with enormous buying power meant margins were razor thin, and existing manufacturers already had well-integrated plants. So instead, Crown went after short-run, hard-to-handle orders and optimised for reducing setup costs rather than ongoing costs. Big producers couldn't really compete in this market, as for their strength (long-run efficiency at scale) they had huge setup costs, making short runs unattractive for them.

This framework can apply to global problems too.

On the problem of catastrophic risk from AI systems, a "weakness" of the problem is that dangerous capabilities are far cheaper to demonstrate than to rule out. And a relative strength is that there is a funding ecosystem for building and running evals or doing elicitation research. Therefore, a good strategy for a body like UK AISI is to develop highly accurate and sensitive evals that can demonstrate what the risks are, which then incentivises much larger actions to be taken.

This is part of my thinking tools series. Also consider: