I write in response to Alan Nichol’s letter about what he calls a “fundamental flaw” in Amazon’s artificial intelligence strategy (January 22).

I fully agree a “human in the loop” on all AI decision-making processes is a must, ensuring responsible strategies and upholding moral and ethical judgment.

However his claim that Amazon’s approach leaves “little hope for predicting a dozen [decisions] in a row without error” is a misleading overstatement that must be challenged. An accuracy rate of 90 per cent per decision does not mean out of every 10 decisions, one will be inaccurate. AI is not cumulative. I also remind readers that human decision-making is inherently and naturally flawed by bias, context and our own cultural and environmental references. The “structured business logic” in major banks he refers to is itself an automated process, and often how machine learning begins.

So let us not resort to alarmist claims that AI gives us “little hope” of ongoing accuracy, but instead ensure that any autonomous decision-making is people-monitored and managed, with strong governance in place.

Written by Noorin Virani.

Read the letter here.