Black Basil Technologies

AI Manifesto

Manifesto for AI Product Development

It began with a simple idea: to craft guiding principles for AI product development. Inspired by the renowned Agile Manifesto, I swiftly translated this notion into words and captured my thoughts in this very post. My goal? To infuse the essence of the Agile Manifesto into our own AI manifesto, keeping it clear and accessible to all.

This is just the beginning, and I welcome contributions from individuals of all backgrounds to enhance and refine this draft.
Together, we can make it exceptional.

We’re discovering smarter methods for creating AI products, both through our own endeavors and by assisting others in the process.

In our journey, we’ve realized the importance of:

  • Ethical AI over Unethical AI
  • No AI over AI
  • Simple AI model over complex AI model
  • Lightweight models over heavy models
  • Data quality over Data quantity
  • High variety over low variety data
  • Explainable and transparent AI over black box AI
In essence, while we acknowledge the importance of the items on the right, our greater emphasis is placed on prioritising the items on the left.

Principles Behind The AI Manifesto

We adhere to these principles:

  1. Our highest priority is to build AI ethically for the betterment of the human race.  Continuously track what goes into model training with Data Governance setup.

  1. Build products which solve business problems and create value even without AI. AI products without any value should be avoided.

  1. Start with simpler model architecture and evolve gradually

  1. Build AI models frequently with fresh data and user feedback

  1. Build smaller models with less hyperparameters. Training and running AI is expensive, use the resources wisely. 

  1. Use high quality data even if you have to sacrifice on quantity. Build data pipelines and capture the data quality metrics and filter before using that data.

  1. Use high variety in data to represent the wider data points. Capture the metrics and filter the balanced data with equal weight to each attribute

  1. Build models which can be explained by looking at the features used in training the model. The signals should be interpretable and should correlate with the model predictions. Build new features using domain specific knowledge and train on them.

As we embark on this journey to shape the future of AI product development, let us remember that our collective efforts hold the power to drive meaningful change. With each contribution, we move closer to realizing a vision of ethical, transparent, and impactful AI solutions. Together, let’s continue to push the boundaries of innovation and create a future where AI serves humanity’s greatest interests. Join us in shaping the AI Manifesto, an initiative driven by
Rajat Singhal, Founder of Black Basil Technologies, , and together, let’s make history.

Connect with us to be part of initiative: