Search This Blog

Book of Why

 Book of Why

Judea Pearl


There are 3 levels of causation


Association: basic level - See - Observe

Intervention: Intervene to make a change and see the effect - “If-Then “

Counterfactual Thinking: Top level - posing “What if”


The current trend in AI is based on the idea that:

Data itself will guide us to the right answers to causal questions-

I am skeptical of this trend

As Data is profoundly dumb without 

Cause and Effects! ….Judea Pearl


Correlation may not mean causation! 

All that deep learning can do is Fit function to Data - Correlation only!

Rather what we need is: a causal structure of the possibilities

That then you can use to go from đŸ˜‚Correlation to causation!