Five predictions for Machine Learning and AI in 2021

Paul Clough, Professor of Search & Analytics at The University of Sheffield and Head of Data Science at Peak Indicators, looks ahead

I’m in the business of helping organisations make better use of their data and good predictions about the future. But the story of 2020 should perhaps make anyone a little nervous about future-gazing. Still, here goes: my five predictions for key developments in Machine Learning and Artificial Intelligence (AI) in 2021.   

There will be an ever-louder debate on AI ethics

The big questions for 2021 are, ‘How can you identify and reduce AI biases?’ and ‘Which framework for ethical AI should I adopt?’

Several factors will make these questions more and more insistent in 2021.

• In 2020, the UK Government set up the centre for Data, Ethics and Innovation. It provides lots of useful resources and will really start to have impact in 2021.
• AI is getting more complex – with advances in Deep Learning, which mimics the workings of the brain, as well as with more and more powerful tools coming out, then making AI more transparent is vital.

AI is spreading – and as adoption increases, more and more people are asking the questions, ‘Who’s in control?’ and ‘Who’s accountable?’

Machine Learning development will be increasingly automated

Most ML projects still involve many hundreds of hours of human time, especially preparing data, feature engineering, and optimising models. Automated platforms, tools and libraries makes ML simpler, more accessible and automate repetitive and tedious tasks. I expect to see more and more AI projects with less and less need for human input across all stages of development in 2021.

HR will make even more use of AI

The workplace is changing radically, especially due to Covid-19. How we work and where we do so both look very different now than they did in January 2020. These changes may well require greater use of AI tools in HR, for instance in on-boarding, employee management and optimising online collaborative working.

In late 2020, Peak Indicators launched Ada, the world’s first artificial-intelligence digital assistant, designed to improve how HR professionals, data analysts and line managers get analytical insights from HR data. I expect AI to really take off in this field in 2021.

More and more people will ask if AI is bad for the planet

What role should AI companies play in the climate crisis? AI can be hungry for computational power. That requires electricity, and electricity often means fossil fuels and carbon-dioxide emissions.

Some models – especially large language ones – use staggering amounts of data. That means they consume vast amounts of power. Training one large language model – using a neural architecture search method – produced the same amount of CO2 as the lifetime running of 5 average US cars.

It is not a message everyone wants to hear – computer scientist Timnit Gebru was recently pushed out of Google for raising it. But the debate will get hotter in 2021.

AI tools will develop a friendlier face

Using AI tools used to be like doing brain surgery – a matter best left to serious experts. But high-powered AI now has ever-more-user-friendly interfaces. Services such as Alexa and Google Home will become more and more powerful, and there will be a focus on making the user experience naturalistic, easy and satisfying.

Amazon, for example, has already created a personality for Alexa, making it more and more like interacting with a human being, and not an enormously complex set of algorithms. We can expect more of that in 2021, across both consumer and business AI applications.