When you hear the term ‘responsible AI’, it’s understandable that this is to be interpreted as the model itself being responsible for the consequences of its actions. However, if AI is indeed artificial, surely that means it cannot assume responsibility? In this context, we refer to the overarching responsibility of AI from the very early stages of creation right up to deployment - so whose responsibility is it for the models to act safely and for the greater good of society?
Many options spring to mind: the government, businesses and independent regulators. Or should it be someone else entirely?
Whilst the development of AI is creating countless opportunities to improve the lives of people across the globe in industries such as education, healthcare, accessibility, finance and more, it’s also raising new challenges around the best way to ensure these models are fair and transparent, whilst being private and secure. This calls for regulation to ensure that companies, researchers, and enterprises alike are following the correct steps for developing responsible AI.
Expectation vs. Reality
The phrase ‘artificial intelligence’ holds big expectations. When companies announce that they are using artificial intelligence, often this comes along with a promise or intention to solve a problem either internally, for their customers, or for a bigger social impact. In the early stages of adoption, this is often not the case as companies are faced with teething problems, buggy algorithms and integration issues. In order to meet the expectations of society it’s important for businesses to manage these by creating a realistic, strong and transparent AI strategy, which includes governance structures and practices needed for “responsible AI" and company-wide AI adoption.
In a recent article, Charles Radclyffe at Forbes reinforces the importance of AI being employed for the benefit of society by explaining that ‘the first step towards responsible AI needs to be about people and not strategy.’
AI for Good vs. Responsible AI
AI can only be ‘good’ if it is responsible - the two go hand in hand. It’s worth mentioning however, that AI for good has another area of core focus. Whilst responsible AI hones in on the ways in which individuals and companies should be using AI, AI for good zooms out looking at the bigger picture of how AI can be applied to solve global challenges like reducing poverty, accessible healthcare, increasing sustainability, climate change, better education, future of food and much more.
Increasing the trust of AI for the general public is incredibly important. If consumers are comfortable that the information they’re giving over will be treated without risk of being compromised they’re far more likely to be loyal customers who trust the service. If, for example, a company is known for having biased algorithms or for building models that are too complicated for humans to understand, there will be limited trust for consumers who will struggle to get on board. As PwC explained in a recent article, ‘AI needs to be explainable so algorithms are transparent or at least auditable. It builds a control framework to catch problems before they start. It deploys teams and processes to spot bias in data and models, and to monitor ways malicious actors could “trick” algorithms. It considers AI’s impact on privacy, human rights, employment, and the environment. Leaders in responsible enterprise AI also participate in public-private partnerships and self-regulatory organisations, to help define standards worldwide.’
It’s all very well putting strategy first, but if people are not at the front of developers minds, they will find themselves in challenging positions having created blackbox and untrustworthy AI.
How can we help?
If you’re keen to learn more about both responsible AI and applying AI for social good, join RE•WORK at the upcoming summits:
• AI for Good Summit, San Francisco, 20 - 21 June, confirmed speakers include Carlos Filipe Gaitan Ospina, ClimateAI; Erin Kenneally, U.S. Dept of Homeland Security; Girmaw Abebe Tadesse, University of Oxford; Kay Firth-Butterfield, World Economic Forum. View confirmed speakers here.
• Responsible AI Summit, Montreal, 24 - 25 October, previous speakers include Yoshua Bengio, Universite de Montreal; Natacha Mainville, Google Brain; Brendan Frey, Deep Genomics; Daphne Koller, Calico. Speakers will be announced soon. Suggest a speaker here.
Get in touch: if you’d like any more information about the summits, email John at [email protected]