t is said that there will be a twentyfold increase in data (180 zettabytes) from now to 2025. In this time artificial intelligence will grow tenfold every year and be worth $380 billion (£301 billion) in five years. Some 90% of this will come from companies and enterprises.
The ethical consequences of artificial intelligence are therefore the most pressing problem in the business world. That was why some 42 countries met last week to try to create a world first corporate governance framework for it.
The promise of AI is that it will imbue machines with the ability to spot patterns from data and make decisions better and faster than humans. But the technology has outrun the regulation of it, exposing challenges including biased AI decisions which disadvantage certain people, and wider misinformation.
“For the first time in history, America and like-minded democracies of the world, will commit to common AI principles reflecting our shared values and priorities,” said Michael Kratsios, the United States representative of the OECD which led the initiative and whose country has signed the accord.
It was significant, however, that while the UK also signed up, China did not. The leaders in AI are the US and China, and according to the FT, Google, Microsoft and their like are stressing self regulation and rapid technological development as key. China in contrast has created a push for AI development, which it describes as a “new focal point of international competition”.
The UK, which is quite good at AI but tiny in comparison to America and China, is pioneering an ethical approach. Its Centre of Data Ethics and Innovation was the first in the world. And in a further initiative today, the Institute of Business Ethics has published a paper by Peter Montagnon which attempts to integrate AI into mainstream corporate governance principles.
In her foreword, Philippa Foster Back, the IBE director says the potential for AI to do good is huge, but so is the downside.
Machine learning can lead to excessive power for those who control the data. “This raises genuine concerns about loss of security and abuse — for example through intrusion into privacy, exploitation of vulnerability and unfair treatment of individuals when systems are biased,” she says.
“The introduction of AI therefore needs to be accompanied by strong and carefully considered ethical principles… Those who consider and respond to the ethical challenges of AI, and are true to their values, are more likely to be trusted. And those who are trusted are more likely to survive and prosper in the long run.
“This, in turn, suggests a business or economic model that is quite distinct from that commonly found in large countries that are pushing hard on the technology button, like the US and China.”
Montagnon says some boards say AI is just too difficult; others that boards have already lost control. Still more say it is better to wait and see what technologies emerge.
But as someone said, while companies may forget Facebook, Facebook will not forget them. Montagnon says most technologies have a human dimension and it is this which the board should focus on.
At the moment AI is not going to make qualitative judgments that machines can implement. Instead AI is about augmenting human activity, not thinking but doing. It cannot think as a human would. Boards in turn want AI technology to allow them to make the company more efficient.
Montagnon says there are nine points companies should consider. Making sure a human is in charge is one, sharing the benefits of AI among the stakeholders is another. Ensuring accountability is a third, meaning companies should be wary of AI-enabled machine learning going too far. Should driverless cars break the speed limit or go on the pavement to avoid crashes, for example.
Avoiding bias, treating customers, employees and contractors fairly, are four, five and six; keeping data safe, and dealing with breaches if they do happen are seven and eight; and number nine is whether the ethical code can help.
Here the answer is yes, but up to a point. It is no good just having overarching principles and it is no good if the company is too detailed in its approach. It is good if the ethics are granular and relevant.
The real question though is whether AI will dominate business or whether companies will still be able to keep some semblance of control.