Following a thought-provoking Institute of Directors’ panel discussion, Bailiwick Data Protection Commissioner Emma Martins reflects on the critical importance of foresight in technological development.
It was a privilege to be part of the
IoD midterm event last week. The theme was ‘Rise of the Machines – what does AI mean for business and boards in 2023?’.
The keynote speaker was Dr Alex Connock, Director of the University of Oxford Postgraduate Diploma in Artificial Intelligence for Business at Saïd Business School.
We had the opportunity to explore what artificial intelligence is, the role it plays in our lives as well as where it might be taking us in the future.
My fellow panellists brought a wealth of experience and perspective and that, I think, is the key. You are not alone if you think that artificial intelligence is only for technologists. You are not alone, but you are wrong. I want to try to explain why.
The use of ChatGPT as a ‘fourth panel member’ at the event was intriguing. For each question posed to us, ChatGPT was asked for its take.
One of the questions I had, was “What are the potential unintended consequences of AI?”
ChatGPT’s reply (in summary) was: bias, job displacement, and privacy concerns.
My list (in summary) was: human downgrading, bias and discrimination, and failures in ethics and accountability.
So there were some similarities in our responses which is not unexpected. ChatGPT is ‘simply’ scraping information that’s already online and there is sufficient commentary out there about the social and economic impacts of AI to pull together a comprehensive response.
I want to raise two specific things around ChatGPT’s answers: discrimination and intention.
Firstly, I want to give a real example of how technology can discriminate.
A touch-free soap dispenser was developed a few years ago. It assessed when a pair of hands were placed underneath it and dispensed soap into them. What on earth has that got to do with AI you may ask? Well, the dispenser was pretty sophisticated from a technological perspective. It used near-infrared technology, which sends out invisible light from an infrared LED bulb for hands to reflect the light back to a sensor which triggers the soap to dispense.
The problem was, it seems the sensor had been tested using only white people’s hands, so they didn’t work when non-white people used them.
This is a single example of technology reaching the market with discrimination baked in. There are so many more, and all of them are completely avoidable. All that needs to be done is for a conscious choice to be made from the outset to take deliberate steps to develop technology that works for everyone.
So, to the second problem I had, that of intention.
In the case of the soap dispenser, the developers clearly did not intentionally set out to build a ‘state of the art racist soap dispenser’. But just because there was no bad intention, does that get them off the hook?
I would argue (strongly) that it does not. Because there is a difference between unintended and unforeseen. If the developers had, at the outset, thought more carefully about their users they could have taken very simple steps to ensure better quality, more inclusive inputs. In later testing stages they had another opportunity to identify unforeseen consequences if they had been diligent enough to stress-test the product until it revealed its problems.
With many AI chatbots, including ChatGPT, being released to market in Beta (testing mode) there is a real danger of stress-testing happening whilst the product is already in the market. These tools are available to anyone who wishes to use them, for good or otherwise. In January 2023 ChatGPT’s own creator, OpenAI, published a report on how ‘large language models’ (like ChatGPT) might be misused for disinformation purposes. Their report highlights that: the number of propagandists will grow and diversify; the scale of disinformation campaigns will increase; campaigns will become more effective; novel tactics will emerge; and propaganda will be less obvious. You don’t have to be a technologist to realise that this has the potential to be a major problem for us all. To its credit, by publishing its report OpenAI is being transparent about the risks, and it does suggest mitigations. But it feels illogical, like a car manufacturer putting a car without brakes on the market and then suggesting that people install their own.
We tend to think of AI questions/issues as being the preserve of technologists. Those of us with only a vague grasp of IT are often not seen as relevant in discussions.
But the way technology is integrated into our lives is a question of what it is to be human. This article touches on just some of the issues that affect us individually and collectively. This is not something that affects other people or other times. It affects us all and affects us now.
Do not mistake this as a pushback on all technology or innovation. There are extraordinary things that technology does well, and will be able to do in the future, that makes our lives better. What it is, is a call for a better understanding of the role we must all play in shaping the word we live in and for us to take more responsibility in shaping it in a way that is based on human values.
Going back to the soap dispenser. What was needed on day 1 was not a technologist. What was needed was a broader perspective – a perspective grounded in an understanding of the human impact of technology and one which was able to think through all consequences (whether unforeseen, unintended, or otherwise).
Profound technological changes, like widespread adoption of AI, are too important and too impactful to be done badly. The importance of true foresight, which is arguably only possible when a diverse group of thoughtful humans come together with no agenda of their own, cannot be overstated. It requires imagination, creativity, insight, intelligence and diligence. These are things, I would argue, that only a diverse group of human beings can bring to the table.
&
Indulge