OWP liVe REPORT

54

prove more valuable than humans in the decision-making process.

How machine learning works

How can artificial intelligence help? Let’s look at how machine learning works.

Robots to the rescue?

Machine learning offers a means to enhance your decision-making by processing large amounts of information according to strict criteria. This is something humans cannot do as well. However, in situations where you do not know your goal or have no immediate feedback, benefits from machine learning are narrower, e.g. providing recommendations to inform human judgment. “This is where machine learning can help in testing hypotheses,” said Professor Chevallier. “We can use machine learning to play devil’s advocate.” “Machine learning and AI tools will mainly help you fix noise – they are still going to struggle with biases. Wherever you have a noise problem, feel free to use machine learning. Where you have a bias problem, take a step back,” said Joshi.

There are three main types of machine learning:

1 Supervised learning: Humans provide many labelled examples of what patterns we want AI to learn. 2 Unsupervised learning: This uses deep learning to both conduct feature recognition, recognize patterns, extract and classify, without any labelled inputs from humans. 3 Reinforcement learning: AI mimics human learning by following a series of instructions and changing approach based on feedback until it achieves an optimal policy. To make sense of any application of machine learning in decision-making, it will also help to break down the types of decisions we might face depending on the context surrounding those decisions: Do you know precisely what you want to achieve and how to do it? Do you receive timely feedback allowing you to assess whether you are still on track? Once you have answered those questions, and knowing how AI works, we can look at ways in which technology might

AMIT JOSHI

MAUDE LAVANCHY

ARNAUD CHEVALLIER