Some notes about Dominic Cummings

• it appears to me that there is only one way to get a rational and balanced discussion. Directors should hire a second advisor to make the case against the proposed acquisition, with its fee contingent on the deal not going through

  • https://dominiccummings.com/2014/10/30/the-hollow-men-ii-some-reflections-on-westminster-and-whitehall-dysfunction/
  • https://dominiccummings.com/2019/06/26/on-the-referendum-33-high-performance-government-cognitive-technologies-michael-nielsen-bret-victor-seeing-rooms/

• it’s about getting physical, spatial training data for our biological neural networks

• without this the data our neural networks are trained on are the words of others, i.e. their attempts to make persuasive arguments

• BUT the most important things we would like to model are complex systems

  • can we actually model them at all?
  • maybe language and metaphor are the best tools we have

• to develop not “gut feel” but the right kind of gut feel, i.e. gut feel based on having an intrinsic understanding of the system rather than gut feel based on not wanting to let your drinking buddy down

• I believe that we can learn to reason about things without understanding them!

  • By developing intuition
  • But it’s important to develop the correct intuition
  • So we don’t fall for fallacies, the supernatural, etc.

• How can we unlock the reasoning ability of thousands of people

• who are currently stuck regarding reasoning about large datasets because they can’t program?

• A lot of Victor’s stuff is really ambitious.

  • A lot of these ideas are really, really hard
  • What’s the MVP?
  • Or alternatively, what’s the 2% of the work that gets 98% of the benefit?

• I’m much more excited by the idea of checklists than Bret Victor’s stuff

• Victor is clearly a visual thinker, but what about people with other thinking styles?

• One reason Victor’s stuff hasn’t come to pass is because the market for consumer applications is far larger (1000x?) than tools