An Anatomy of Hype
Writing 310 years ago, the essayist Jonathan Swift had this to say on the art of political lying.
“It often happens, that if a lie be believed only for an hour, it hath done its work, and there is no further occasion for it. Falsehood flies, and truth comes limping after it, so that when men come to be undeceived, it is too late.”
Swift’s essay was about how clever, devious people use political lies to get their way, leaving a trail of damage behind them.
Enter a new political lie about artificial intelligence and NHS data.
It’s the usual breathless stuff. “The world is undergoing the fastest technological and scientific revolution in the whole history of human civilisation, with profound implications for the United Kingdom,” gush former prime minister Tony Blair and former leader of the Conservative Party William Hague in a new report .
“The challenge of responding to that revolution is so urgent, the danger of falling behind so great and the opportunities so exciting that we urged all political parties to make their response to it nothing short of creating a new national purpose. We set out how the UK could harness the power of technology and reimagine the state and public services by restructuring Whitehall, making better use of data”.
If ‘weapons of mass destruction’ was the political lie that achieved the objective of regime change (the overthrow of Saddam Hussein), here we have a new type of regime change in sight: deposing health data as a public and personal asset and its mass transfer to the control and benefit of Big Tech.
The hype machine
Hype is the main tool of the political lie. The trick with hype is to set an agenda by mesmerising people with AMAZING Opportunities!!! while deflecting their attention from on-the-ground realities until they can’t be avoided any more.
Here are some of the detailed tricks of skilful political lier.
- Present one small part — the part you are trying to promote — as if it is the whole picture, ignoring everything that doesn’t fit into your carefully crafted agenda.
- Exaggerate the benefits of what you are proposing to the nth degree.
- Skim over all its flaws, inconsistencies, risks and loopholes. (Make sure you mention them briefly to give the appearance that you have thought about them, but only to brush them aside as minor, technical details to be sorted out later.)
- Use small elements of truth to tell the big lie.
- Craft the way you present your arguments so that the conclusions you arrive at are inevitable, given the assumptions you’ve smuggled in along the way.
Tony Blair’s agenda for an NHS data landgrab is a good case study of hype in the service of a political lie.
Present a part as the whole
If you believe Tony Blair, the future of the NHS — indeed the competitiveness of the country as a whole (!) — depends on the UK taking a lead in AI (see ‘exaggeration’ below). Taking the lead in AI depends on amassing huge amounts of data to train this AI … which means that we simply have to hand over the nation’s health data to the Big Data experts — the Big Tech monopolies.
Tony Blair writes as if the insights generated by such AI will transform the provision of health and care services. This is nonsense.
Setting AI to work on big data sets does nothing to change those areas where treatments are already well developed and understood. It does nothing to improve the operational delivery of these treatments. In other words, it will not and cannot touch most of what the NHS actually does, day-to-day. Neverthless, it is being presented as if it is the answer to everything.
In reality, what the NHS needs to transform itself — to cut costs while improving quality — lies at the opposite end of the data spectrum. It lies in new data logistics capabilities: the ability to get exactly the right personal data to and from the right people at the right times, safely and efficiently, in a privacy protecting way.
Better data logistics enables better physical logistics — improving the ways in which work is organised; finding ways to reduce time and effort spent while improving outcomes. It’s about the practical, detailed work that takes place on the front line. Nothing to do with the statistical analysis of big sets of data.
But nowhere in his 72 page report does Mr Blair give this very real opportunity a single mention (or, for that matter, in any of his previous reports). After all, the whole point of the magician with his conjuring trick is to focus peoples’ attention on your trickery; to deflect their attention from what’s really going on behind the scenes.
Exaggerate the benefits
According to Tony Blair, Artificial Intelligence is “so urgent and important that how we respond is more likely than anything else to determine Britain’s future.” Wow! We had better drop everything else then! Focus everything we’ve got on it!
But this too is nonsense. AI will never water the Sahara, achieve world peace, feed the hungry or solve the challenge of Climate Change.
AI currently takes two main forms: machine learning techniques such as genetic algorithms and Large Language Processing models like ChatGPT.
The machine learning form of AI is very good at identifying patterns and correlations in huge amounts of data. That’s great. It can be a useful tool within a well managed research process.
But … er … that’s about it.
One drawback of this form of AI is that while it is very good at spotting correlations, a high proportion of them are meaningless false positives. To make things worse, spotting correlations has got nothing to do with understanding causation. Confusing correlation with causation can — and has already — set many damaging hares running.
Also, this form of AI is bound by the ancient ‘GIGO’ law of computing: Garbage In, Garbage Out. Machine learning outputs are only as good as the data they’ve been trained with (by humans), and very often this data is not very good. In fact, for almost every issue that a machine learning AI is set to work on, there is far more information out there in the world that is not contained in its training sets. This means the ‘insights’ it generates are always partial and contingent.
In addition, because of the way it works, there is no audit trail as to how it arrives at its conclusions (and it may arrive at different conclusions if it crunches the same set of data a second time). On top of this, no matter how good the analytics might be, what happens in the real world depends entirely on how any resulting actions are implemented. Which has got nothing to do with AI.
So machine learning isn’t half as transformational as people like Blair are making it out to be. Not by a very long shot.
The other Large Language Processing sort of AI (the one that drives ChatGPT and its rivals) works by brute number crunching. It analyses previous bits of text (that it has been fed with by humans) to calculate the most probable next letter in a word and the most probable next word in a sentence. It doesn’t understand the meaning of anything it spews out.
This sort of AI is quite good at providing summaries of conventional wisdoms (e.g. the words used in texts that have already been generated by humans).
But … er … that’s about it.
It has no awareness of ideas or information that are not in the data sets it has been trained with. It cannot use its (non-existent) understanding of content to produce any new ideas. Meanwhile, because it works only by brute number crunching without any idea of meaning, it can and does invent ‘facts’ and ‘statements’ that are simply untrue — what AI practitioners call ‘hallucinations’. So, yes, it can be useful for students wanting to cheat with essays or consultants wanting to streamline their report-writing. But change how the real world actually works? Not really!
AI is just another of a long line of tech bubbles. They all have a common pattern. The promoter’s trick is to dazzle people with promises of wondrous change tomorrow … to undertake a landgrab of today’s available resources. It’s worked many times before, and it’s being tried again here.
Skim over the flaws and risks
The trick here is to present something that is extremely difficult and fraught with risk as if it were safe and easy. This way, people aren’t put off by a realistic assessment of what’s really involved in making it happen.
Tony Blair’s papers are a masterpiece of such skating on thin ice. He tells us that we don’t have to worry about privacy because all the data will be anonymised. But he doesn’t provide any chapter or verse as to how this will actually work, or if (for example) this data could be de-anonymised as quickly and easily as it was anonymised.
Meanwhile, many of the promises he makes are based not on the use of anonymised data for statistical analysis but on extreme personalisation. “Over the coming decades,” he says, “AI systems will be created that can understand each individual’s composition and lifestyle, and harness that information to provide highly personalised lifestyle and treatment suggestions on demand, with little marginal cost for each additional patient.” How is that going to work if the data is anonymised?
Most of his proposals are aimed at giving ‘researchers’ access to data about people. But he never defines what a bona fide researcher is. Cambridge Analytica could have passed themselves off as ‘researchers’.
Research, he tells us, “serves a dual purpose; it is both a public service enabling technological advances and a commercial opportunity.” He makes a big fuss about the NHS being able to ‘sell’ our data to the private sector, thereby helping to fund the NHS. Really? What proportion of the NHS budget would it cover? And how exactly will this commercial opportunity be realised? By Big Tech companies using NHS data to generate new treatments, and then draining the NHS of cash in charges for access to these treatments?
Nowhere are such questions worked through. All skimming. All about making it sound both wonderful and easy. All skating on very thin ice indeed.
Use a small element of truth to tell a big like
For hype to be credible it has to contain an element of truth. People have to be able to recognise what it says and say to themselves “Yes! That makes sense!” As soon as you get their heads nodding you can use their agreement to lure them into the much bigger lie that lies behind your truth-hook.
Tony Blair’s reports use this trick often. For example, he writes that today’s fragmented system of data collection prevents researchers from “understanding how different features of a patient — their lifestyle, genetics, personality and physiology — relate to one another”. Absolutely spot on. According to the World Health Organisation, 60% of related factors to individual health and quality of life are correlated to lifestyle. And he’s right. Today’s fragmented approach to data collection and use makes it impossible to join the necessary data dots.
This is why individuals need personal data stores, so that they can collect all this data safely. So it can be used under their control, for their benefit.
But that, of course, is not what he suggests. The undeniable truth of data fragmentation is just a useful hook to sell his bigger lie. Fragmentation, he continues, “is limiting efforts to generalise beyond the subjects studied and holding back the transformative potential of AI.” (Back to presenting a small part of the picture as if it were the whole). Therefore, rather than addressing this fragmentation by empowering individuals with their own data, we should build a massive new centralised database … and hand it over to Big Tech.
Assumption smuggling
This entire campaign is based on a particular set of assumptions. A paradigm. This paradigm tells us that wealth can only be created by private companies making products which they sell in markets using money prices.
But the NHS is living proof that the ‘firms-products-prices-markets-profits’ paradigm is not the only way to create wealth and deliver value. That’s why the Big Tech companies and free market ideologues hate it so much. Because it doesn’t fit into their mental straitjacket and doesn’t serve their vested interests. It’s why all these reports’ recommendations are framed in terms of markets and monetisation: selling (not sharing) the data; ‘commercial opportunities’ (not public service); ‘competitiveness’ (who on earth is the NHS actually ‘competing’ with?)
The end result is downright insulting. “Countries that are home to frontier laboratories not only provide cutting-edge training but also exert a powerful pull for the next generation of top entrepreneurs, researchers and other globally mobile talent,” the report declares. So what about the millions of people who are not ‘top entrepreneurs, researchers and other globally mobile talent’? In other words, what about the people who actually deliver the service, day in and day out? Clearly, for the hype merchants they amount to nothing.
Conclusion
Tony Blair’s reports are not particularly unique or special. They are one of a genre, all of which display the same core characteristics. They display:
- unquestioning obedience to the mythology that monetisation and markets are the only way to create wealth
- a studied indifference to the day-to-day operational realities of actually providing services
- endless displacement activity that obsesses with The New! And The Cool!!! where the further they are away from day-to-day operational realities the better
- barely disguised contempt for the people who actually produce and use the services in question
- all for the promotion of the barely-concealed agenda of asset stripping to line the pockets of the already rich and powerful
All bandwagons flourish by presenting a part as if it were the whole, by exaggerating the part’s benefits and by making light of the difficulties. Bandwagoneers use this heady mix to manipulate people to their own ends — to spread political lies. Today, the message that ‘AI is our saviour’ is one such political lie.
We ended our last blog on the NHS data bandwagon by saying the people of this country deserve better than this. They do. They deserve and need to be equipped with their own data and data tools to make their lives better — not to have their data and tools to be handed over to parasitic monopolists.