Engineering the AI applications (Part 4): When to use AI/ML based on (un)certainty of inputs and outputs?

In this short post, I would like to share a few lessons that we have learnt while developing the AI applications.

One question that comes often in designing AI based applications is

"When to use AI/ML based on (un)certainty of inputs and outputs?"

Conditions under which decisions are made:

1. Humans take decisions
2. Machines assist humans in taking decisions
3. Machines make decisions (autonomous)

Humans taking decisions - semi AI systems

1. If humans can take decisions with certainty, and the consequences are certain, measured and deterministic then AI need not be used.
2. If humans can take decisions with certainty, and the consequences are uncertain then AI can be used to predict or estimate the consequences (optimization problem).
3. If humans can't take decisions with certainty, then AI can be used to choose the best (optimization problem) alternatives or choices available.
4. If humans can't take decisions with certainty, then AI can be used to choose the best (optimization problem) alternatives or choices available; AI can be used to predict or estimate the consequences (optimization problem).

Machines assist humans in taking decisions - semi AI systems

1. If machines can process the available data for a given context and take decisions with certainty, and the consequences are certain, measured and deterministic then AI need not be used. Typically procedural programming languages will help solving such problems.
2. If machines can process available data for a given context that results in uncertainty under which decisions to be made, then AI can be used to choose the best (optimization problem) alternatives or choices available.
3. If machines can process available data for a given context that results in uncertainty under which decisions to be made, then AI can be used to choose the best (optimization problem) alternatives or choices available (optimization problem). AI is useful to to predict or estimate the consequences (optimization problem).

Machines make decisions (autonomous) - full AI systems

1. If machines can process available data for a given context that results in uncertainty under which decisions to be made, then AI can be used to choose the best (optimization problem) alternatives or choices available (optimization problem).
2. If machines can process available data for a given context that results in uncertainty under which decisions to be made, then AI can be used to choose the best (optimization problem) alternatives or choices available (optimization problem). AI is useful to to predict or estimate the consequences (optimization problem).

Summary

Affiliations

I write posts for? Vidcentum Technologies where I am founder director. Some of the articles are published in my blogs / articles / posts as well as on the company timeline / posts.

Vidcentum Technologies is a software and data company. It provides AI/IT/OT related services.