Home Care In Oxford, IA

Everyone is worried about Artificial Intelligence. From writers in Hollywood to computer programmers, recent advances in technology are causing concern about what Generative AI is going to mean for the future of work, our society and the wider world. Is there nothing machines will not be able to do?

We have spent a decade researching the impacts of AI. Ten years ago, we wrote a paper estimating that some 47% of US-based jobs could be automated in principle, as AI and mobile robotics expanded the scope of tasks that computers can do.

Our estimates were based on the premise that, while computers might eventually be able to do most tasks, humans would continue to hold the comparative advantage in three key domains: creativity, complex social interactions, and interaction with unstructured environments (such as your home).

However, it is important to acknowledge there has been meaningful progress in these domains, with Large Language Models (LLMs), such as GPT4, capable of producing human-like text responses to a very wide range of queries. In the age of Generative AI, a machine might even write your love letters.

Yet if GPT4 does write your love letters, your in-person dates will become even more important. The bottom line is, as virtual social interactions are increasingly aided by algorithms, the premium on in-person interactions, which cannot be replicated by machines, will become even greater.

As virtual social interactions are increasingly aided by algorithms, the premium on in-person interactions, which cannot be replicated by machines, will become even greater

Second, although AI can produce a letter in the style of Shakespeare, this is only because Shakespeare’s works already exist, and on which an AI can be trained. AI is generally good at tasks which have clear data and a clear goal, such as maximizing the score in a video game, or the similarity to the language of Shakespeare. But if you want to create something genuinely new, rather than rehashing existing ideas, for what should you optimise? Answering the question of the true goal is where much human creativity resides.

Third, as we noted in our 2013 paper, there are many jobs that can be automated, but Generative AI – a subfield of the broader field of AI – is not yet an automation technology. It needs prompting from a human, and it needs a human to select, fact-check and edit the output.

Finally, Generative AI generates content that mirrors the quality of its training data – ‘garbage in results in garbage out’. And these algorithms require training on expansive datasets, such as extensive segments of the internet, as opposed to smaller, refined datasets developed by experts. Consequently, LLMs are inclined to create text that aligns with the average, rather than the extraordinary, portions of the internet. Average input yields average output.

If you want to create something genuinely new, rather than rehashing existing ideas, for what should you optimise? Answering the question of the true goal is where much human creativity resides

What implications does this hold for the future of employment? For one thing, the latest generation of AI will persist in necessitating human intervention. What is more, workers with less specialised skills stand to gain disproportionately, as they can now generate content that aligns with the ‘average’ benchmark.

What implications does this hold for the future of employment? For one thing, the latest generation of AI will persist in necessitating human intervention. What is more, workers with less specialised skills stand to gain disproportionately

Could the hurdles outlined above be overcome shortly, paving the way for widespread automation also of creative and social tasks? In the absence of a major breakthrough, we think it is unlikely. Firstly, the data already ingested by LLMs is likely to comprise a considerable fraction of the internet, making it unlikely that training data can be significantly expanded to power further progress. Furthermore, there are legitimate grounds to expect a surge of inferior AI-crafted content on the Web, progressively degrading its quality as a source of training data.

Second, while we have become accustomed to Moore's Law - the observational law that declares the quantity of transistors in an integrated circuit (IC) doubles approximately every two years - many anticipate this trend will lose momentum, owing to physical constraints, around 2025.

Third, it is estimated that the energy to create GPT4 cost a large fraction of its $100 million training cost – even before the price of energy went up. With the climate challenge looming large, there are questions over whether this approach can continue.

What is needed, in other words, is AI that is capable of learning from smaller, curated datasets, drawing upon expert samples, rather than the average population. But when such innovation will come is notoriously hard to predict. What we can do is create better incentives for data-saving innovation.

Consider this: around the turn of the 20th century, there was a genuine contest - would electric vehicles or the combustion engine prevail in the burgeoning car industry? At first, both contenders were on a par, but massive oil discoveries tipped the balance in favour of the combustion engine. Now imagine that we leveraged a tax on oil back then: we might have shifted the balance in favour of the electric car, sparing us plenty of carbon emissions. In similar fashion, a tax on data would create incentives for innovation to make AI less data-intensive.

Going forward, as we have argued elsewhere, many jobs will be automated, but not because of the latest wave of Generative AI. In the absence of major breakthroughs, we expect the bottlenecks we outlined in our 2013 paper to continue to constrain automation possibilities for the foreseeable future.

This opinion piece reflects the views of the author, and does not necessarily reflect the position of the Oxford Martin School or the University of Oxford. Any errors or omissions are those of the author.

Related Programmes

Future of Work Oxford Martin Programme on the Future of Work

An AI tool that can predict 10-year risk of deadly heart attacks, could transform treatment for patients who undergo CT scans to investigate chest pain, according to British Heart Foundation-funded research presented today at the American Heart Association’s Scientific Sessions in Philadelphia.

In the first real-world trial of the AI tool, it was found to improve treatment for up to 45 per cent of patients. The AI technology could potentially save the lives of thousands with chest pain, who may not have been identified as at risk of a heart attack, and therefore may not have received appropriate treatment to lower their risk. With the technology also found to be cost-effective, the researchers hope it could change the management of patients who are referred for chest pain investigations, across the NHS.

Every year in the UK around 350,000 people have a cardiac CT scan – the standard test to identify any narrowings or blockages in the coronary arteries. In around three quarters of cases, there is no clear sign of significant narrowings, so patients are often reassured and discharged. Unfortunately, many of these people will die from a heart attack in future, because small, undetectable narrowings may break up if they are inflamed, blocking the arteries. Until recently, it was not possible to identify these patients at risk.

In the new study, funded by the British Heart Foundation, Professor Charalambos Antoniades and his team at the University of Oxford's Radcliffe Department of Medicine, analysed data from over 40,000 people undergoing routine cardiac CT scans at eight UK hospitals. Participants were followed up for a median of 2.7 years. While those with significant coronary artery narrowings were more likely to have serious cardiac events or death, twice as many patients with no significant narrowings experienced heart attacks and cardiac deaths.

The team then used a new AI tool, trained using information on changes in the fat around inflamed arteries – which can indicate the risk of events such as heart attacks – as well as information on narrowings of the arteries and other clinical risk factors. Further testing on an additional 3,393 patients over 7.7 years revealed it could independently and accurately predict risk of cardiac events.

Among those with no obstructions to their arteries, those with the highest levels of inflammation in their blood vessels had a more than 10-fold higher risk of cardiac death compared to those with lower levels of inflammation.

In a world-first pilot, the team provided AI-generated risk scores to clinicians for 744 consecutive patients, and found that in up to 45 per cent of cases, clinicians altered patients’ treatment plans, indicating that this AI tool could be hugely valuable in guiding and informing how patients with chest pain are managed, ensuring early identification and preventative treatment of those at highest risk.

Analysis comparing the use of the AI tool to standard care revealed it was highly cost-effective for the NHS. In addition, the researchers estimate that implementing this technology in the NHS could lead to over 20 per cent fewer heart attacks and 8 per cent fewer cardiac deaths and strokes, among those having the test. With the technology required to power the AI tool already commissioned by NHS England for a pilot programme in five NHS hospitals, the researchers are hopeful that it could soon be rolled out across the UK.

Professor Charalambos Antoniades, BHF Chair of Cardiovascular Medicine and Director of the Acute Multidisciplinary Imaging & Interventional Centre, at the University of Oxford, said: 'Our study found that some patients presenting in hospital with chest pain – who are often reassured and sent back home – are at high risk of having a heart attack in the next decade, even in the absence of any sign of disease in their heart arteries. Here we demonstrated that providing an accurate picture of risk to clinicians can, alter and potentially improve the course of treatment for many heart patients.

'We hope that this AI tool will soon be implemented across the NHS, helping prevent thousands of avoidable deaths from heart attacks every year in the UK.'

Professor Sir Nilesh Samani, Medical Director at the British Heart Foundation, said: 'This research shows the valuable role AI-based technology can play in better identifying those patients most at risk of future heart attacks and thereby help clinicians make better treatment decisions for their patients.

'Too many people are needlessly dying from heart attacks each year. It is vital we harness the potential of AI to guide patient treatment, as well as ensuring that the NHS is equipped to support its use. We hope that this technology will ultimately be rolled out across the NHS, and help to save the lives of thousands each year who may otherwise be left untreated.'

Professor Antoniades’s research is also supported by the National Institute for Health and Care Research Oxford Biomedical Research Centre.

Carl-Benedikt Frey is the Dieter Schwarz Associate Professor of AI & Work at the Oxford Internet Institute and a Fellow of Mansfield College, University of Oxford. He is also Director of the Future of Work Programme and Oxford Martin Citi Fellow at the Oxford Martin School.

After studying economics, history, and management at Lund University, Frey completed his PhD at the Max Planck Institute for Innovation and Competition in 2011. He subsequently joined the Oxford Martin School where he founded the programme on the Future of Work. Between 2012 and 2014, he was teaching at the Department of Economic History at Lund University.

In 2012, Frey became an Economics Associate of Nuffield College and Senior Fellow at the Institute for New Economic Thinking, both University of Oxford. He remains an Associate Fellow of the Department of Economic History at Lund University, and a Fellow of the Royal Society of Arts. In 2019, he joined the World Economic Forum’s Global Future Council on the New Economic Agenda, as well as the Bretton Woods Committee. And between 2020 and 2022, he was a member of the Global Partnership on Artificial Intelligence (GPAI) – a multistakeholder initiative to guide the responsible development and use of AI, hosted by the OECD.

In 2013, Frey co-authored The Future of Employment: How Susceptible Are Jobs to Computerization. With over 12,000 citations, the study’s methodology has been used by President Barack Obama’s Council of Economic Advisors, the Bank of England, the World Bank, as well as the popular automation risk-prediction tool of the BBC. In 2019, the paper also featured on the Last Week Tonight Show with John Oliver.

Frey has served as an advisor and consultant to international organisations, think tanks, government, and business, including the G20, the OECD, the European Commission, the United Nations, and several Fortune 500 companies. He is also an op-ed contributor to the Financial Times, Foreign Affairs, Scientific American, and the Wall Street Journal, where he has written on the economics of artificial intelligence, the history of technology, the future of cities, and remote work.

His academic work has featured in over 100 media outlets, including The Economist, New York Times, Time Magazine, The New Yorker, Le Monde, and Frankfurter Allgemeine Zeitung. In addition, he has frequently appeared international broadcast media such as CNN, BBC, PBS News Hour, Al Jazeera, and Sky News.

His most recent book, The Technology Trap: Capital, Labor, and Power in the Age of Automation, was selected a Financial Times Best Books of the Year in 2019, when it also won Princeton University’s prestigious Richard A. Lester Prize.

Institute for New Economic Thinking at the Oxford Martin School Oxford Martin Senior Fellow

In 2013, researchers at Oxford University published a startling number about the future of work: 47 percent of all United States jobs, they estimated, were “at risk” of automation “over some unspecified number of years, perhaps a decade or two.”

But a decade later, unemployment in the country is at record low levels. The tsunami of grim headlines back then — like “The Rich and Their Robots Are About to Make Half the World’s Jobs Disappear” — look wildly off the mark.

But the study’s authors say they didn’t actually mean to suggest doomsday was near. Instead, they were trying to describe what technology was capable of.

It was the first stab at what has become a long-running thought experiment, with think tanks, corporate research groups and economists publishing paper after paper to pinpoint how much work is “affected by” or “exposed to” technology.

In other words: If cost of the tools weren’t a factor, and the only goal was to automate as much human labor as possible, how much work could technology take over?

When the Oxford researchers, Carl Benedikt Frey and Michael A. Osborne, were conducting their study, IBM Watson, a question-answering system powered by artificial intelligence, had just shocked the world by winning “Jeopardy!” Test versions of autonomous vehicles were circling roads for the first time. Now, a new wave of studies follows the rise of tools that use generative A.I.

In March, Goldman Sachs estimated that the technology behind popular A.I. tools such as DALL-E and ChatGPT could automate the equivalent of 300 million full-time jobs. Researchers at Open AI, the maker of those tools, and the University of Pennsylvania found that 80 percent of the U.S. work force could see an effect on at least 10 percent of their tasks.

“There’s tremendous uncertainty,” said David Autor, a professor of economics at the Massachusetts Institute of Technology, who has been studying technological change and the labor market for more than 20 years. “And people want to provide those answers.”

But what exactly does it mean to say that, for instance, the equivalent of 300 million full-time jobs could be affected by A. I.?

It depends, Mr. Autor said. “Affected could mean made better, made worse, disappeared, doubled.”

One complicating factor is that technology tends to automate tasks, not entire occupations. In 2016, for instance, the artificial intelligence pioneer Geoffrey Hinton considered new “deep learning” technology capable of reading medical images. He concluded that “if you work as a radiologist, you are like the coyote that’s already over the edge of the cliff but hasn’t yet looked down.”

He gave it five years, maybe ten, before algorithms would “do better” than humans. What he probably overlooked was that reading the images is just one of many tasks (30 of them, according to the U.S. government) that radiologists do. They also do things like “confer with medical professionals” and “provide counseling.” Today, some in the field worry about an impending shortage of radiologists. And Mr. Hinton has since become a vocal public critic of the same technology he helped create.

Mr. Frey and Mr. Osborne calculated their 47 percent number in part by asking technology experts to rate how likely entire occupations like “telemarketer” or “accountant” were to be automated. But three years after their paper published, a group of researchers at the ZEW Center for European Economic Research, based in Mannheim, Germany, published a similar study that assessed tasks — like “explain products or services” — and found that just 9 percent of occupations across 21 countries could be automated.

“People like numbers,” said Melanie Arntz, the lead author of the ZEW paper. “People always think that the number must be somehow solid, you know, because it’s a number. But numbers can really be very misleading.”

In some scenarios, A.I. has essentially created a tool, not a full job replacement. You’re now a digger who can use an excavator instead of a shovel. Or a nurse practitioner with access to better information for diagnosing a patient. It’s possible that you should charge more per hour, because you’re going to get a lot more done.

In other scenarios, the technology is replacing your labor rather complementing it. Or turning your job from one that requires special skills to one that doesn’t. That is not likely to go well for you.

In either case, says Mr. Autor, technological developments throughout history have tended to mostly affect wages and wealth distribution — not how many jobs are available. “This kind of exercise risks missing the forest by focusing on one very prominent tree,” he said of studies that look at how much human work could be replaced by A.I.

What he considers to be another key focus — how artificial intelligence will change the value of skills — is difficult to predict, because the answer depends partly on how new tools are designed, regulated and used.

Take customer service. Many companies have handed the task of answering phones to an automated decision tree, bringing in the human operator only to troubleshoot. But one Fortune 500 enterprise software company has approached the problem differently. It created a generative A.I. tool to provide the agents with suggestions for what to say — keeping humans, and their ability to read social cues, in the loop. When researchers at Stanford and M.I.T. compared the performance of groups who were given the tool with those who weren’t, they found the tool significantly improved the performance of lower-skilled agents.

Even if a job becomes completely automated, how displaced workers fare will depend on how companies decide to use technology in new kinds of work, especially work we can’t yet imagine, said Daron Acemoglu, a professor at M.I.T. and an author of “Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity.” These choices will include whether to automate work entirely or use technology to augment human expertise.

He said that the seemingly scary numbers predicting how many jobs A.I. could eliminate, even if it’s not clear how, were a “wake up call.”

He believes that people could “steer in a better direction,” he said, but he is not optimistic. He does not think we are on a “pro-human” path.

All estimates for how much work A.I. could take over are very dependent on humans: the researchers making the assumptions about what technology can do. Mr. Frey and Mr. Osborne invited experts to a workshop to score the likelihood of occupations becoming automated. More recent studies rely on information such as a database tracking A.I. capabilities, created by the Electronic Frontier Foundation, a nonprofit digital rights group. Or they rely on workers using platforms like CrowdFlower, where people complete small tasks for money. The workers score tasks on factors that make them prone to automation. For instance, if it’s something with a high tolerance for error, it’s a better candidate for a technology like ChatGPT to automate.

The exact numbers are not the point, say many researchers involved in this type of analysis.

“I would describe our methodology as almost certainly precisely wrong, but hopefully directionally correct,” said Michael Chui, an A.I. expert at McKinsey who co-authored a 2017 white paper suggesting that about half of work, and 5 percent of occupations, could be automated.

What the data describes is, in some ways, more mundane than often assumed: Big changes are coming, and it’s worth paying attention.

Clear ethical standards and guidance must be in place to protect the relationship of trust between doctors and patients and to safeguard human rights, according to a Council of Europe report today, written by Dr Brent Mittelstadt, an Oxford expert specialising in Artificial Intelligence and medical ethics.

Dr Mittelstadt is Director of Research at the Oxford Internet Institute and a leading data ethicist. He says, ‘I hope the report will make people think about how AI might disrupt the core practices involved in healthcare.’

But he is concerned AI could be used as a way to reduce budgets or save costs rather than to improve patient care and says, ‘If you’re going to introduce new technology into the clinical space, you need to think about how that will be done. Too often it is seen solely as a cost-saving or efficiency exercise, and not one which can radically transform healthcare itself.’

Too often AI is seen solely as a cost-saving or efficiency exercise, and not one which can radically transform healthcare itself

Dr Brent Mittelstadt

Dr Mittelstadt was commissioned to write the report by the Council of Europe’s Steering Committee for Human Rights in the fields of Biomedicine and Health to investigate the potential impact of AI on the doctor-patient relationship. The report advises the use of AI remains ‘unproven’ and could undermine the ‘healing relationship’.

The ethicist points out, AI is already being used in clinical settings – not robot doctors, which ‘don’t exist’. He says, ‘Systems which use AI go a long way back...they might identify tumours in scans or recommend diagnosis.’ But, Dr Mittelstadt says, ‘They are not autonomously making decisions about patient care.’

His report states, ‘A radical reconfiguration of the doctor-patient relationship of the type imagined by some commentators, in which artificial systems diagnose and treat patients directly with minimal interference from human clinicians, continues to seem far in the distance.’

However, Dr Mittelstadt says, there are already robot systems being used in surgery and his report is predicated on the idea that automation is creeping forward in clinical settings – creating the need for clear standards.

The report warns, ‘The doctor-patient relationship is the foundation of ‘good’ medical practice, and yet it is seemingly being transformed into a doctor-patient-AI relationship. The challenge facing AI providers, regulators and policymakers is to set robust standards and relationships for this new clinical relationship ensure patients’ interests and the moral integrity of medicine as a profession are not fundamentally damaged by the introduction of AI.’

He explains AI could change healthcare as a third actor in the relationship – bringing new information and new dynamics. In turn, this could change how care is delivered.

According to the report, with the introduction of AI, a shift could take place in the relationship, but not in the patient, ‘The patient’s vulnerability [is] not changed by the introduction of AI...what changes is the means of care delivery, how it can be provided, and by whom. The shift of expertise and care responsibilities to AI systems can be disruptive in many ways.’

The report also warns AI can introduce bias into a situation but detecting biases is not straightforward. He writes, ‘AI systems are widely recognised as suffering from bias... Biased and unfair decision-making often...reflects underlying social biases and inequalities. For example, samples in clinical trials and health studies have historically been biased towards white male subjects meaning results are less likely to apply to women and people of colour.’

AI systems are widely recognised as suffering from bias... historically been biased towards white male subjects meaning results are less likely to apply to women and people of colour

And it underlines the challenges, and potential problems in the relationship between physicians and patients, including in terms of patients’ safety and protection, ‘If AI is used to heavily augment or replace human clinical expertise, its impact on the caring relationship is more difficult to predict.... The impact of AI on the doctor-patient relationship nonetheless remains highly uncertain’. And, it points out, ‘AI poses several unique challenges to the human right to privacy.’

The need for binding standards for how AI is deployed and governed is clear, the report adds, ‘Recommendations should focus on a higher positive standard of care with regards to the doctor-patient relationship to ensure it is not unduly disrupted by the introduction of AI in care settings.’