From a ‘deranged’ provocateur to IBM’s failed AI superproject: the… – Gavi, the Vaccine Alliance

Just over a decade ago, artificial intelligence (AI) made one of its showier forays into the public’s consciousness when IBM’s Watson computer appeared on the American quiz show Jeopardy! The studio audience was made up of IBM employees, and Watson’s exhibition performance against two of the show’s most successful contestants was televised to a national viewership across three evenings. In the particular end, the machine triumphed comfortably.

One associated with Watson’s opponents Ken Jennings , who went on in order to make a career on the back of his gameshow prowess, showed grace – or was it deference? – in defeat, jotting down this commentary to accompany his final answer: “I, for one, welcome our new computer overlords. ”

In fact, their phrase had been poached from another United states television mainstay, The Simpsons. Jennings’ wry pop culture reference signalled Watson’s reception less as computer overlord and more as technological curio. But that was not how IBM saw this. On the back again of this very public success, in 2011 IBM turned Watson toward 1 of the most lucrative but untapped industries for AI: healthcare.

What followed over the particular next decade was a series of ups and downs – but mostly downs – that exemplified the promise, but also the numerous shortcomings, of applying AI to health care. The Watson health odyssey finally ended in 2022 when it was sold off “for parts” .

There is much to learn from this story about why AI and healthcare seemed so well-suited, and why that will potential has proved therefore difficult in order to realise. Yet first we need to revisit the particular controversial origins of data use in this field, long before electronic computers were invented, plus meet a single of the American pioneers, Ernest Amory Codman – an elite by birth, the surgeon simply by training, and a provocateur by nature.

Data’s role in the birth of modern medicine

While the utility associated with data in a general way had already been clear for several centuries, its collection plus use on a massive scale was a feature of the 19th century. By the 1850s, collecting census information had become commonplace. Its use was not merely descriptive; it formed a way to make determinations about how to govern.

The nineteenth century marked the first time that, as US systems expert Shawn Martin explains, “managers felt the need to tie the information that society collected to things like performance [and] productivity”. This applied to public health because well, where “big data” played a critical role within establishing relationships between populations, their habits and environment (both at home and work), and disease.

A well-known example will be John Snow’s discovery from the source associated with a cholera outbreak in London’s Soho neighbourhood within 1854. Now considered one particular of epidemiology’s founding fathers, Snow canvassed door in order to door asking whether the families within had cholera. His analysis came chiefly in the particular re-organisation of the data he collected – the plotting on the map – such that will a pattern might emerge. This ultimately established not really just the extent of the outbreak but additionally its source, the Broad Street water pump .

For Boston-born Codman, an outspoken medical reformer working at the particular beginning associated with the 20th century, such use of information to understand disease was up there since “one of the greatest moments in medicine”.

Though Codman has been involved within many data-driven reforms during his controversial career, one of the most successful was the particular Registry associated with Bone Sarcoma , which he established in 1920. His goal was to collect and analyse all of the cases of bone cancer (or suspected bone cancer) from across the US, and to use these to establish diagnostic criteria, therapeutic effectiveness and a standardised nomenclature.

There were a few rules for this registry. Individual doctors that contributed had to send x-rays, case reports and, if possible, tissue samples for examination by the registry’s consulting pathologists plus Codman himself. This would ensure both the accuracy and uniformity of pathological analysis. The particular effort was a success which usually grew over time: by 1954, when the particular American College of Surgeons sought the new home for the registry, it contained a good impressive 2, 400 complete, cross-referenced cases.

On the particular face associated with it, Codman’s decision to focus on bone cancer was baffling. It had been neither a pressing nor a common concern for doctors throughout the ALL OF US. But the disease’s relative rarity was 1 reason he chose this. Codman felt the amount of data received through his nationwide request would not be overwhelming with regard to his small team of researchers in order to analyse.

Perhaps more importantly, he knew that studying bone malignancy would raise the ire of far fewer associated with his colleagues than a more common disease might. In the clinical atmosphere in which expertise was understood as a combination of long experience with a dash of intuition – the physician’s “art” – Codman’s touting of data as a better way to obtain knowledge about a disease and its treatment was already being met along with vociferous opposition.

It didn’t help that he tended to be inflammatory and provocative within the pursuit associated with his data-driven goals. At a medical meeting in Boston within 1915, this individual launched a surprise attack on his fellow practitioners. In the particular middle of this staid affair, Codman unveiled an 8ft cartoon lampooning his colleagues regarding their apathy toward health care reform plus, as he or she saw it, their wilful ignorance associated with the limitations of the profession. As one (former) friend put it in the particular event’s aftermath, Codman’s only hope was that people would take the “charitable” view and consider him not an enemy from the profession yet merely “mentally deranged”.

Codman’s 8ft cartoon lampooned medical practices in the early 20th century. From The Shoulder by E.A. Codman
Codman’s 8ft cartoon lampooned medical practices in the early twentieth century.
From The Shoulder by E. A. Codman

Undeterred, Codman continued this pugnacious approach to their pioneering work. In a 1922 letter to the prestigious Boston Medical and Surgical Journal, he complained that the surgeons of Massachusetts have been particularly unhelpful to his registry. He explained that he had – politely – asked the particular 5, 494 physicians in the state in order to “drop him a postal stating whether or not he understood of the case” so that Codman could acquire “the best statistics ever obtained around the frequency of the disease”. To his chagrin, he had received just 19 responses in nearly two years. Needling the journal’s editors and readers simultaneously, he asked:

Is this because your Journal is not read? … [Or] because of the indifference from the medical profession as to whether the frequency of bone sarcoma is usually known or not?

Codman proposed a questionnaire that would allow the journal in order to see if the problem has been its lack of readership, or his colleagues’ “inertia, procrastination, disapproval, resistance or disinterest”. A subsequent editorial in response to Codman’s proposal was surprisingly magnanimous:

Whether we will it delete word, we are obliged to be irritated, amused or even instructed, according to our temperaments, by Dr Codman. Our advice is to be instructed.

An end to elitism?

Despite the establishment’s resistance, submissions to Codman’s registry began to grow such that will by 1924, he had enough material in order to make preliminary comments regarding bone cancer. For a single thing, he previously succeeded in standardising the particular much-contested matter of the proper nomenclature for the illness. This, this individual exulted, had been so significant that it should be likened to the particular “rising associated with the sun”.

The registry also offered up many pieces of “impersonal proof”, as Codman called their data-driven findings, of the rightness of certain theories that individual physicians had promoted. Claims, for example , that combined treatments associated with “surgery, mixed toxins plus radium” had been more effective than treatments that relied on any of these types of alone were borne out by the information.

Have you read?

The registry, because Codman’s colleague Joseph Colt Bloodgood place it , “excited great interest” among professionals, and not just because it had “influenced the entire medical world to pay a lot more attention to bone tissue tumours”. More importantly, this provided a new model for how to do healthcare work. Another admiring friend responded to Bloodgood:

The work of the registry [is] one of the outstanding American contributions in order to surgical pathology. As a method of study, it shows the necessity of really wide experience before the surgeon is definitely capable associated with handling intelligently cases of the disease . [It] can be impossible for any single individual to claim finality of this sort.

This emphasis on “very wide experience” over the experience of “any single individual” points to another critical reason to prefer data, based on Codman. His goal in changing the method by which medical knowledge was made had not been just in order to get better results. By seeking to undo the image of medicine as a good “art” that will depended upon the wisdom of a select group of preternaturally talented individuals, Codman also threatened to undo the class-ridden reality that underlay this general public veneer.

As the efficiency engineer Frank Gilbreth implied inside a 1913 article in the particular American Magazine , if it was true that medicine required no specific intrinsic gifts (monetary or otherwise), then absolutely anybody – whatever their class, race or even background – could do it, including “bricklayers, shovellers and dock-wallopers” who had been currently shut out of this kind of “high-brow” occupations.

Codman has been even more pointed. If data was used to evaluate the outcomes associated with his physician colleagues, he or she insisted, it would show that this quality of doctors and hospitals had been generally poor. He sniped that they excelled primarily in “making dying men think they are getting better, concealing the gravity associated with serious diseases, and exaggerating the importance of minor illnesses to suit the occasion”.

“Nepotism, pull and politics” were the particular order of the day in medication, Codman wrote in one particular of his most scathing takedowns of his co-workers at the Massachusetts General Hospital. Yet he made themselves the centrepiece of the critique, conceding that will his entrance to Harvard Medical School had come on the back of “friends and relatives among the well-to-do”. The particular only difference, he suggested, was that he was willing to own up in order to it, plus to subject himself and his work towards the scrutiny associated with data.

Data’s unflattering see of medicine

Codman was not the only person having a come-to-Jesus moment with data more than this period. In the 1920s, the United states social science researchers Robert and Helen Lynd gathered data within the little US town of Muncie, Indiana, as a way of creating a picture of the “averaged American” .

By the particular 1930s, the similarly-minded Mass Observation project took off in Britain, intending to collect data about everyday life so as to create an “anthropology of ourselves”. Crucially, both reflected the particular thinking that furthermore drove Codman: the right way to know something – a people, the disease – was to produce what seemed a suitably representative average. And this particular meant the amalgamation associated with often quite diverse and wide-ranging characteristics and their compression into a single, standard, efficient unit.

The turn from describing representative averages to learning from these averages is certainly probably greatest articulated within the work of pollsters, whose door-to-door interrogations were aimed at helping a nation to know itself simply by statistics. Within 1948, inspired by their own failure in order to correctly predict the outcome associated with the US presidential election – among the most famous psephological errors in the nation’s history – pollsters such as George Gallup plus Elmo Roper began to rethink their particular analytic methods, spinning away from quota sampling and towards random sampling.

The 1948 election was one of the most famous psephological errors in US history. Clifford K. Berryman/Wikimedia
The 1948 election has been one of the most famous psephological errors within US history.
Credit:   Clifford K. Berryman/Wikimedia

At the same time, thanks primarily in order to its military applications , the technology of computing began to gather pace. And the growing fascination along with knowing the globe via information combined with the unparalleled ability associated with computers to crunch this appeared a match made in heaven.

In the late-in-life preface to their 1934 data-driven magnum opus within the anatomy from the shoulder, Codman experienced comforted themself with the thought that he was a man ahead of his time. And indeed, just the few years after their death in 1940, statistical analysis started to pick upward steam within medicine.

Over the next two decades, figures this kind of as Sir Ronald Fisher , the particular geneticist plus statistician remembered for suggesting randomisation as an antidote in order to bias, great English compatriot Sir Austin Bradford Hill , who else demonstrated the connection between smoking and lung cancer, also pushed forward the integration of record analysis in to medicine.

However , it might take numerous more many years for word to finally leak away that, by data’s measure, both the methodologies of medical research and much associated with medicine itself was ineffective. In a movement led in part simply by outspoken Scottish epidemiologist Archie Cochrane , this unflattering statistical look at of medication finally really saw the light of day in the particular 1960s and 70s.

Cochrane went therefore far since to say that will medicine had been based on “ a level of guesswork” so great that any return to health after a medical intervention was more a “tribute to the sheer survival power of the minds and bodies” of patients than anything else. Aghast at the revelations embedded in Cochrane’s 1972 book, Random Reflections on Health Services , the Guardian journalist Ann Shearer published :

Isn’t it … a lot more than fair to ask what on Earth we – and more particularly, the particular medical They – have been doing all these types of years to let the wellness machine develop with such a lack of quality control?

The answer dates back in order to Codman’s bone fragments cancer registry half a century earlier. The particular medical establishment on each sides of the Atlantic had been avoiding with all their might the scrutiny that will data would certainly bring.

Computers finally obtain medical currency

Despite their own increasing ubiquity in the particular 1970s plus 80s, computers had still only haltingly joined the medical mainstream. Though the smattering associated with AI applications began to appear in healthcare in the particular 1970s, it was just in the 1990s that computers actually started to acquire some healthcare currency.

In a page borrowed straight through Codman’s time, the pioneering American biomedical informatician Edward Shortliffe noted in 1993 how the future of AI in medicine depended on the realisation that “the practice of medicine is inherently an information-management task”.

In the particular US, the Institute associated with Medicine and the President’s Information Technology Advisory Council released reviews highlighting the particular failures of medicine to fully embrace information technology. By 2004, a newly appointed national coordinator intended for health information technology was charged along with the herculean task associated with establishing a good electronic medical record for all Americans by 2014.

An IBM System 360 computer in 1969. USDA Forest Service via Wikimedia Commons
An IBM System 360 computer within 1969.
Credit:   USDA Forest Service via Wikimedia Commons

This explosion of interest in bringing computers directly into healthcare produced it an enticing and potentially profitable area to get investment. So it is no surprise that IBM celebrated Watson’s winning turn on Jeopardy! in 2011 by putting it to work on an oncology-focused programme with multiple US-based clinical partners selected on the basis of their access to healthcare data.

The idea was laudable. Watson would perform what machine learning algorithms do best: mining the particular massive amounts of data these institutions had at their disposal, searching pertaining to patterns that will would help to improve treatment. However the complexity associated with cancer as well as the frustratingly unique responses of patients to it, yoked together by information systems that were sometimes incomplete and sometimes incompatible with each other or with machine learning’s methods a lot more generally, limited Watson’s ability to be useful.

One sorry example has been Watson’s Oncology Expert Advisor , a collaboration using the MD Anderson Cancer Center in Houston, Texas. This particular had begun its existence as the “bedside diagnostic tool” that pored through patient records, scientific literature and doctors’ notes in order in order to make real-time treatment recommendations. Unfortunately, Watson couldn’t “read” the doctors’ notes. While good at mining the scientific books, it couldn’t apply these large-scale discussions towards the specifics from the people in front of this. By 2017, the task had been shelved .

Elsewhere, at New York City’s famed Memorial Sloan Kettering Cancer Middle, clinicians found a more elaborate – and infinitely more problematic – method forward. Rather than relying on the retrospective data that is machine learning’s usual fodder, clinicians invented new “synthetic” cases which were, by virtue of getting been invented, definitely less messy and more total than any kind of real data could be.

The project re-litigated the “data v expertise” debate associated with Codman’s period – once more within Codman’s favour – since this developed data got built straight into it the particular specifics of cancer treatment as comprehended by a small group of clinicians at a single hospital. Bias, in other words, had been programmed directly in, plus those engaged in training the system knew it.

Viewing historical patient information as too narrow, they rationalised that replacing this along with data that will reflected their own collective encounter, intuition and judgment could build into Watson For Oncology the latest and finest treatments. Of course , this did not work any better in the early 21st century than it acquired in the particular early 20th.

Furthermore, while these physicians sidestepped the problem associated with real data’s impenetrable messiness, treatment options available at a wealthy hospital in Manhattan were much removed from those available in the particular other localities that Watson was meant to serve. The particular contrast was perhaps starkest when Watson was introduced to other parts of the world , only to find the particular treatment regimens it recommended either didn’t exist or were not really in keeping with the local plus national infrastructures governing how healthcare has been done there.

Even in the US, the particular consensus, as one unnamed physician within Florida reported back to IBM, had been that Watson was the “piece of shit” . Most of the time, this either told clinicians what they already knew or offered up advice which was incompatible with local conditions or the specifics associated with a patient’s illness. At best, it provided up a snapshot of the views of a select few clinicians at a moment in time, now reified because “facts” that ought in order to apply uniformly and everywhere they went.

Many of the elegies written to mark Watson’s selling-off within 2022, having failed to create good upon its promise in health care, attributed its downfall in order to the same kind of overpromise and under-delivery that will has spelled the finish for many health technology start-ups.

Some maintained the fact that scaling-up of Watson from gameshow savant to oncological wunderkind might have been effective with more time. Perhaps. But in last year, time was of the particular essence. In order to capitalise for the goodwill toward Watson plus IBM that Jeopardy! had created, to be the trailblazer in to the lucrative but technologically backward world associated with healthcare, experienced meant striking first and fast.

Watson’s high-profile failure highlights a good overlooked barrier to contemporary, data-driven healthcare. In the encounters along with real, human patients, Watson stirred upward the exact same anxieties that will Codman got encountered – difficult questions about exactly what it is exactly that medication produces: care, and the human being touch that comes with it; or even cure, plus the info management tasks that play a critical role here?

A 2019 study of US patient perspectives associated with AI’s part in health care gave these types of concerns some statistical shape. Though a few felt optimistic about AI’s potential to improve healthcare, a vast majority gave voice in order to fundamental misgivings about relinquishing medicine to machine learning algorithms that could not explain the logic they employed to reach their diagnosis. Surely the absence of the physician’s judgment would increase the risk of misdiagnosis?

The persistence of this worry offers quite often resulted in caveating the function of device learning along with reassurances that will humans are usually still in charge. In a 2020 report in the InnerEye project, meant for example, which used retrospective data to identify tumours on patient scans, Yvonne Rimmer, a clinical oncologist in Addenbrooke’s Medical center in Cambridge, addressed this particular concern:

It’s important for patients to know that the AI is helping me in my professional function. It’s not replacing me in the process. I doublecheck everything the AI does, and can change it if We need to.

Data’s uncertain role in the long term of healthcare

Today, whether a doctor gives you your diagnosis or you get it from a computer, that diagnosis is not primarily based over the intuition, view or experience of either doctor or individual. It’s driven by data that provides made our cultures of mainstream treatment relatively more uniform plus of a higher standard. Just as Codman foresaw, the introduction of information in medicine has furthermore forced the greater degree of transparency, both in terms associated with methodologies and effectiveness.

Nevertheless , the a lot more important – and potentially intractable – problem with this modern strategy to health is its lack of representation. Because the Sloan Kettering dalliance with Watson began in order to show, datasets are not the particular “impersonal proofs” that Codman took them to become.

Even under less egregiously subjective problems, data undeniably replicates plus concretises the biases associated with society itself. As MIT computer scientist Marzyeh Ghassemi explains, data offers the particular “ sheen of objectivity ” whilst replicating the ethnic, racial, gender and age biases of institutionalised medicine. Thus the tools, tests and techniques that are based on this particular data are also not impartial.

Ghassemi highlights the inaccuracy of pulse oximeters, often calibrated upon light-skinned individuals, for all those with darker skin. Others might note the outcry on the gender bias within cardiology , spelled out especially in higher mortality rates for women who have heart attacks.

The list goes on and on. Remember the human genome project, that big data triumph which has, according to the ALL OF US National Institutes of Wellness website , “accelerated the particular study associated with human biology and improved the practice of medicine”? It almost exclusively drew upon genetic studies of white Europeans. According to Esteban Burchard in the University of California, San Francisco:

96% associated with genetic studies have already been done on people with European origin, even though Europeans make up less than 12% of the world’s population … The particular human genome project should have been called the European genome project.

A lack of consultant data offers implications designed for big information projects over the board – not least for precision medicine , which will be widely touted as the antidote to the problems of impersonal, algorithm-driven health care.

Precision or “personalised” medication seeks in order to address one of the essential perceived drawbacks of data-based medicine by locating finer-grained commonalities between smaller and smaller sized subsets from the population. By focusing upon data from a genetic and cellular level, it may yet counter the criticism that the particular data-driven method of recent decades is usually too blunt and insensitive a tool, such that “even the most frequently prescribed drugs for that the majority of common circumstances have very limited efficacy”, according to computational biologist Chloe-Agathe Azencott .

Yet personalised medication still feeds on the same depersonalised data since medicine more generally, so it too is definitely handicapped simply by data’s biases. And even if it can step beyond the problems associated with biased data – plus, indeed, organizations – the particular question of its role in the future of our own everyday healthcare does not end there.

Even taking the utopian view that personalised medicine might make possible treatments as individual because we are, pharmaceutical companies won’t create these remedies unless they are profitable. And that requires either prices so high that will only the wealthiest of us could afford all of them, or a market therefore big that these businesses can “ achieve the requisite return on investment ”. Truly individualised care is not really on the table.

If the goal in healthcare is to help a lot more people by being more representative, more inclusive and more attentive to person difference within the medical everyday associated with diagnosis and treatment, large data isn’t going to help us out. In least not as things currently stand.

For the story of healthcare information to date has directed us squarely within the other direction, towards homogenisation plus standardisation since medical goals. Laudable as the rationales for such a focus just for medicine have been at different moments in our history, our expectations for your potential for machine studying to enable all associated with us to live longer, healthier lives remain something of the pipe dream. Right now it is nevertheless us humans, not our own computer conspirtors, who hold most sway over the individual wellness outcomes.

The Conversation


Caitjan Gainty , Senior Lecturer in the History of Science, Technology and Medicine, King’s College London

Doctor Caitjan Gainty is a winner of The Conversation’s Sir Paul Curran award for academic communication

The Conversation

This article can be republished from The Conversation under the Creative Commons license. Read the original article .

This article is part of Discussion Insights
The Insights team generates long-form journalism derived from interdisciplinary study. The team is working with academics through different backgrounds who possess been involved in projects targeted at tackling societal and scientific challenges.

Disclosure statement

Caitjan Gainty does not work with regard to, consult, personal shares in or receive funding from any company or organisation that could benefit from this article, and provides disclosed no relevant affiliations beyond their particular academic appointment.


King’s University London   provides funding because a member from the Conversation UK.

Leave a Reply

Your email address will not be published. Required fields are marked *