Sharing — Thinking Fast and Slow, by Daniel Kahneman

How System 1 & System 2 Related to AI?

Sik-Ho Tsang
4 min readOct 26, 2022
Thinking Fast and Slow, by Daniel Kahneman

Today, I would like to share a book, Thinking Fast and Slow, by Professor Daniel Kahneman. This book is published in 2011 and has already got over 40000 Google citations right now. (Sik-Ho Tsang @ Medium)

In this book, he mentioned how we think fast and slow using System 1 and System 2 in our brains, how these systems affect human decision making.

But How System 1 & System 2 Can Be Related to AI?

Daniel Kahneman was awarded the 2002 Nobel Memorial Prize in Economic Sciences (shared with Vernon L. Smith). He is an influential living psychologist. He merges psychology and economics. His empirical findings challenge the assumption of human rationality prevailing in modern economic theory.

System 1 is Automatic and Impulsive

Daniel Kahneman receives the Nobel Memorial prize in Economic Sciences from King Carl Gustaf of Sweden in Stockholm, 2002 (Image from Interview Daniel Kahneman: ‘Clearly AI is going to win. How people are going to adjust is a fascinating problem’ in 2021)
  • It’s the system you use when someone sketchy enters the train and you instinctively turn towards the door.
  • It’s crucial to our survival.

System 2 is Very Conscious, Aware and Considerate

TED Talk: The riddle of experience vs. memory in 2010
  • This system is at work when you’re meeting a friend and trying to spot them in a huge crowd of people, as it helps you recall how they look and filter out all these other people.
  • It helps you exert self-control and deliberately focus your attention.

How System 1 Confuses System 2?

Thinking, Fast and Slow | Daniel Kahneman | Talks at Google in 2011
  • However, these 2 systems don’t just perfectly alternate or work together. They often fight over who’s in charge and this conflict determines how you act and behave.
  • Below are 2 examples (from 4-Minute Reading):

A baseball bat and a ball cost $1.10. The bat costs $1 more than the ball. How much does the ball cost?

  • If your instant and initial answer is $0.10, system 1 just tricked you as the brain is lazy.
  • Once you spent a minute or two actually thinking about it, you’ll see that the ball must cost $0.05. Then, if the bat costs $1 more, it comes out to $1.05, which, combined, gives you $1.10.

Imagine these 2 scenarios:

1. You’re given $1,000. Then you have the choice between receiving another, fixed $500, or taking a 50% gamble to win another $1,000.

2. You’re given $2,000. Then you have the choice between losing $500, fixed, or taking a gamble with a 50% chance of losing another $1,000.

Which choice would you make for each one?

  • This is related to prospect theory. If you’re like most people, you would rather take the safe $500 in scenario 1. This is because we’re a lot more afraid to lose what we already have, as we are keen on getting more.
  • One thing in the book to tell is that your behavior always affected by 2 systems in your mind.

But how they are related to AI?

But How System 1 & System 2 Are Related to AI?

  • (This section is beyond the book.)
  • The below video podcast, interviewing Daniel Kahneman in recent year of 2020.
  • One of the points he mentioned is that the current AI does not have the ability of reasoning, which has the same comment as Professors Bengio, Yann, and Hinton.
Daniel Kahneman: Thinking Fast and Slow, Deep Learning, and AI | Lex Fridman Podcast #65 in 2020 (Image captured from video podcast)
  • On the other hand, in Deep Learning for AI, which is an ACM magazine article written by Professors Bengio, Yann, and Hinton. They also mention the relationship of System 1/System 2 and AI:
  • They believe that the current deep learning is most successful at perception tasks only and generally what are called system 1 tasks.
  • Using deep learning for system 2 tasks that require a deliberate sequence of steps is an exciting area that is still in its infancy.
  • Our implicit (system 1) processing abilities allow us to guess potentially good or dangerous futures, when planning or reasoning. This raises the question of how system 1 networks could guide search and planning at the higher (system 2) level.

To achieve system 2 abilities, such as reasoning, being able to factorize knowledge into pieces which can easily recombined in a sequence of computational steps, and being able to manipulate abstract variables, types, and instances. We would like to design neural networks which can do all these things.

Thinking, Fast and Slow will transform the way you think about thinking.

References

[2011 Book] [Thinking Fast and Slow]
Thinking Fast and Slow

[4-Minute Reading] [Thinking Fast and Slow]
https://fourminutebooks.com/thinking-fast-and-slow-summary/

[2020 Podcast] [Thinking Fast and Slow, Deep Learning, and AI]
Daniel Kahneman: Thinking Fast and Slow, Deep Learning, and AI | Lex Fridman Podcast #65

0. Legends

2011 [Thinking Fast and Slow] 2017 [AI Reshapes World] 2019 [New Heights with ANN] 2021 [Deep Learning for AI] 2022 [Small is the New Big]

My Other Previous Paper Readings

--

--

Sik-Ho Tsang

PhD, Researcher. I share what I learn. :) Linktree: https://linktr.ee/shtsang for Twitter, LinkedIn, etc.