Last Thursday I attended, as is my wont, the annual Bowman Lecture at the University of Glasgow, a lecture series devoted to the public understanding of statistical methods in their application to various sciences, and indeed to sundry walks of life. Last year the lecture was given by the head of MI5. (I suppose it must be all right now to say so: it was after all a public lecture.) I remember Prof Bowman saying, “How are we going to follow this?” Well, this year the speaker was Dr Chris Wiggins, Chief Data Scientist of the New York Times, also an assistant professor at Columbia University. He gave a talk entitled How Data Happened: A history from the Age of Reason to the Age of AI. The talk was, in effect, a precis of a book Dr Wiggins co-wrote with the historian Matthew L. Jones, entitled, How Data Happened: a history from the age of reason to the age of algorithms. That immediately begs the question, is there a close association between “AI” and “algorithms”? Do Turing Machines think algorithmically? If this, do that? I suppose they do, in the sense that everything is reduced to binary code. We, Homo sapiens, don’t think algorithmically, unless we are forced to do so by our on-line managers and work supervisors. That way lies mental breakdown, and madness.
I greatly enjoyed the Bowman Lecture. That came to me as something of a surprise, because I didn’t feel particularly kindly disposed towards my idea of Artificial Intelligence. Recently, our erstwhile PM Sir Tony Blair came on the airwaves to wax enthusiastic about the potential for AI to transform the National Health Service. I wrote to The Herald, twice. I think I may have mentioned it in this blog. I well remember the way, around a quarter of a century ago, the predecessor to AI, Information Technology (IT), “transformed” the Health Service. When the sharp suits closed in on the NHS with their automated systems, I remember thinking at the time that this was not about improving patient outcomes, it was all about power. Clinicians were not looking for a procedure or methodology with which to solve a clinical problem. Rather Information Technologists were looking for a field of human activity – health, education, policing – in which a computerised system could be applied, or, perhaps, inflicted.
But I was disarmed by Chris Wiggins, because as it seemed to me, he had no axe to grind. He wasn’t peddling a system that might benefit an organisation. Rather he was recounting the relationship between humanity and data – or enumerated facts – that has developed over the last 250 years. He didn’t even seem to be particularly impressed by the notion that numerical data added weight to an argument. Data can be used as a tool or weapon, to argue what is true. It can be a tool for rearranging, or defending, power. Data can be created and curated. The use, or abuse, of “Big Data” becomes an unstable game that can be played among states, companies, and people generally.
What is “Artificial Intelligence”? Are these super-computers really “Intelligent”? It turned out that this was a question of more interest to the audience than to the lecturer, who was more focused on what these machines can, of themselves, achieve. After all, Alan Turing devised a computer that could break the Nazi Enigma Code, without pondering whether or not his Colossus could “think”. So Dr Wiggins was less interested in the philosophical questions posed by AI, referred to by an audience member as “epistemological”, as to what these systems might be capable of. Similarly, he wasn’t preoccupied by the notion that AI might present humanity with an existential threat. Would AI, asked the Vice-Principal, signal the death of scientific creativity? Dr Wiggins thought not. Was AI a good thing? All he would say was that he thought it was here to stay.
Here to stay indeed. It’s back on the front page of today’s Herald. AI hopes for patients with heart failure. This describes a pilot study led by the University of Dundee, working with an AI company to develop software to scan patient records and patient investigations, such as echocardiography, in order to identify specific patients who would benefit from specific treatments. AI tech, says the follow-up article on Page 8, could “revolutionise” care for heart failure patients.
But you see, medicine is not remotely like that. You don’t start with an investigation and then go looking for a patient. You start with a patient, and you start with a history. As a doctor, you don’t just collect data. You step into the patient’s shoes. You almost go into a trance. For a moment, you become the patient. You try to experience yourself, something of the patient’s experience, and you try to formulate a plan of management that is utterly unique to a specific individual. It comes at a cost.
One thing is clear; the outcome achieved by the automated perusal of vast amounts of data will only ever be as good as the quality, and validity, of the data itself. Data in medicine, for example, must always be taken with a pinch of salt. Death certificate data is particularly suspect. As to cause of death, in the absence of a post mortem, the doctor usually takes a punt, primarily aimed at facilitating funeral arrangements on behalf of the bereaved. There is a letter in today’s Herald about Lib Dem MSP Liam McArthur’s Assisted Dying Bill, shortly to reappear in Holyrood, which instructs doctors to issue death certificates recording cause of death as the terminal illness of the deceased, and not the lethal potion that has been taken. The cause of death is misrepresented. The terminal illness did not kill the patient; and in the normal course of events, it might never have killed the patient. This is just one more example of the recording of bad data. It needs to be resisted.
The oft quoted cliché attributed to Mark Twain about lies, damned lies, and statistics, can be taken two ways, either as short-hand for a rapidly accelerating spiral of disinformation, or as a portrayal of a contrast – on the one hand lies and damned lies, and on the other, data as an exemplar of truth. Even more obscure than Mark Twain, Robert Burns once said that “Facts are chiels that winna ding”, which I think means that truths will resonate with one another, and not be dissonant. Yet we live in an age of cognitive dissonance, humbug, and fake news. Pontius Pilate’s question, “What is truth?” is the question of our time.
jamescalumcampbell 