license :: install :: contribute :: issues :: cite :: contact
Imagine that, very quickly, we can find simple models that let us succinctly descrine the world around us. In that world, models are easy to grow and maintain. Such models can be easily understood, so we can use them to find what matters most (or least). Further we audit such models to find and fixes sources of biases in their conclusions. Lastly, such models are not resource intensive to build or deploy so anyone can build them and everyone everywhere can use them.
Do you live in that work? Have you checked? Before you waste time and money (and much CPU) to build complex and opaque AI models, have you first tried something simpler?
For example, all the AI methods explored here are quick to code and fast to run. If you applied them to your problems then, at the very least, you will get a baseline system which you could use to show the superiority of any other other, seemingly more sophisticated, method. And at the very most, if these simple methods work well, then you have avoided all that extra modeling.
Ultimately it is an experimental issue whether or not simplicity-first works for some new problem. Hence, we call this book "Is AI Simple?" and we offer you the following challenge. We invite you to apply our methods to your own problems, then tell us how it all goes. In this way, with time, we will build up lists of problems that should (or should not) be tackled using simpler AI.
This book is in two parts. Firstly, some case studies are presented where "simplicity-first" performed very well indeed. Secondly, there are 15 coding challenges aimed at teaching programmers how to write simpler AI systems:
- These challenges could be used as a guided self-study or as the basis of a graduate class on programming AI systems. Each challenge comes with just enough theory to understand the problem, as well as numerous suggestions on how this code might be improved.
- Each challenge comes with a fully working example, which you need to your favorite language. You have mastered the challenge when your unit tests produce similar output to our unit tests.
For reasons of simplicity and portability (especially to IOT devices), we code our examples in LUA-- which you can just treat as a executable specification (so you will not need to code in LUA to use this book).
Less. Plz.
-- timm
Less, but better.
-- Deiter Rams
Truth is shorter than fiction.
-- Irving Cohen
...pursue ultimate simplicity... -- Marie Kondō
It's easy to be heavy; hard to be light.
-- G.K. Chesterton
Simplicity is the ultimate sophistication.
-- William James
Or, rather, let us be more simple and less vain.
-- Jean-Jacques Rousseau
Half of mathematics is the art of saving space.
-- Mokokoma Mokhonoana
Our life is frittered away by detail. Simplify, simplify.
-- Henry David Thoreau
Don't use a big word where a diminutive one will suffice.
-- Anon
One must learn to be simple, anyone can manage to be complex.
-- Amit Kalantri
One day I will find the right words, and they will be simple.
-- Jack Kerouac
The art of being wise is the art of knowing what to overlook.
-- William James
For I am a bear of very little brain and long words bother me.
-- Winnie the Pooh
Life only puzzles us because we're not simple enough to understand it. -- Marty Rubin
Truth is ever to be found in simplicity, and not in the multiplicity and confusion of things.
-- Issac Newton
As complexity rises, precise statements lose meaning, and meaningful statements lose precision.
-- Lotfi Zadeh
The computing scientists main challenge is not to get confused by the complexities of (their) own making.
-- Edsger Dijkstra
Simplicity is a great virtue but it requires hard work to achieve it and education to appreciate it. And to make matters worse: complexity sells better.
― Edsger Dijkstra
- The goal is not models, but insight. Every answer should propose N better questions for round i+1. Find and cluster space of possible solutions. Report a summary of each cluster.
When I first meet simplicity. 2 pages, references, code.
- Greg Gay: BORE