AI Manifesto Logo

Manifesto for applied Artificial Intelligence development

We are uncovering better ways of developing Artificial Intelligence applications by doing it and helping others do it. Through this work, we have come to value:

  • Engineers, Experts, and Product Owners over Data Scientists
  • Customer-Driven Solutions over Data-Driven Problem Solving
  • Outputs That Matter over Inputs We Happen To Have
  • Software Interoperability over Tuning Algorithms
  • Explainability and Accountability over Accuracy

While we value the items on the right, the items on the left generate better solutions, more meaningful impact, and decreased confusion.

David "Gonzo" Gonzalez & Ben Taylor
© 2019, the above authors

We do not track you. We will never share or sell your data.

Next steps after signing the manifesto

5 second effort
Share the #AIManifesto with your friends and colleagues
5 minute effort
Share your plans to implement one or more of the values next year
$5 effort
Pick-up an #AIManifesto sticker to spread the word and help us host this. Get AI-Manifesto.org Swag

Share the values of the manifesto

1

Engineers, Experts, and Product Owners over Data Scientists

Illustration

2

Customer-Driven Solutions over Data-Driven Problem Solving

Illustration

3

Outputs That Matter over Inputs We Happen To Have

Illustration

4

Software Interoperability over Tuning Algorithms

Illustration

5

Explainability and Accountability over Accuracy

Illustration

Principles behind the manifesto

  1. Our highest priority is to satisfy the customer through early and continuous delivery of automation, forecasts, or predictions.
  2. We build projects around existing bottlenecks with defined KPIs even when it makes compiling a dataset difficult.
  3. Engineers must work together daily with subject matter experts and at regular intervals with product owners.
  4. Subject matter experts must decide what data the AI needs to be trained on and empowered to recognize and address bias.
  5. Operations must have clear reporting on AI performance and impact over time.
  6. Applied AI development must participate in software development processes.
  7. Consumption of AI outputs that improve specific KPIs are the primary measures of progress.
  8. Operations must not be bottlenecked by AI research.