ChatGPT: I asked the AI bot to do my job as a data journalist - here’s why my livelihood is safe for now
It might not have produced a Pulitzer Prize piece of journalism but the impact AI could have on the journalism industry shouldn't be ignored.
and live on Freeview channel 276
It’s writing first class academic essays, reportedly allowing workers to take on multiple full-time jobs and has even been hooked up to a Furby to ‘take over the world’ – but could ChatGPT replace me as a journalist?
Artificial intelligence (AI) is anticipated to replace more than 300 million jobs worldwide, according to Goldman Sachs. Perhaps I’m digging my own grave but I wanted to see if the AI chatbot everyone is talking about is capable of doing my job (work smarter, not harder, they say).
I work in the data unit at NationalWorld. Along with my two editors we crunch the numbers to tell the stories hidden in the depths of spreadsheets. These vary from crime to the climate crisis but we particularly keep a close eye on the state of the UK’s health system, so what better task to give my new journalism colleague than to get it to write a piece on the state of England’s A&E departments.
Last week (14 April) I asked the AI tool to “write me a journalism article about NHS A&E waiting times in England”. My article was spat out within seconds. On first glance, it looked legitimate enough but whether it would be a plausible submission is another question. Here’s what happened when I tasked ChatGPT with being a journalist.
Robot journalist versus human journalist
ChatGPT: “The NHS (National Health Service) has long been a point of pride for the people of England, providing free and accessible healthcare to all who need it. However, recent data has revealed a growing problem within one of the most vital aspects of the NHS - Accident and Emergency (A&E) services.
“According to the latest statistics, waiting times for A&E services in England have hit their worst levels since records began in 2010, with only 84.9% of patients being seen within the target time of four hours in January 2023. This means that over 300,000 patients had to wait longer than four hours to receive the care they needed.”
ChatGPT has fallen dramatically at the first hurdle in becoming a journalist. On Thursday (13 April) NHS England published its latest statistics on A&E waiting times which cover from 2010 up to March 2023 – two months more recent than ChatGPT’s latest statistics claim. The actual latest figures show 71.5% of patients were seen within the four hour waiting time target at A&E – not 84.9% that ChatGPT claims. Even when looking at January’s figures the information it provides is still wrong. In January 72.4% were seen within the target timeframe. I’m not entirely sure where it got the 84.9% figure from because since records began not a single month recorded this rate.
My AI assistant also claimed “over 300,000 patients had to wait longer than four hours to receive the care they needed”. While this claim is technically correct (483,000 people were waiting over four hours in January and 551,000 in March), I assume it hasn’t just gone with some obscure rounding technique and is again just publishing wrong information.
Making up figures is generally seen as a sacking offence at NationalWorld but let’s see if it improves.
ChatGPT: “One contributing factor is the increasing demand for A&E services, with a growing number of patients presenting with complex and urgent conditions. The Covid-19 pandemic has also had a significant impact, with increased pressure on hospitals and staff due to the high number of patients requiring treatment for the virus.”
Without doubt the pandemic’s impact on the country’s healthcare system has been profound but ChatGPT implies that hospitals are still being overwhelmed with Covid patients, which is no longer the case. The latest figures published by the UK Government show there were 6,428 people in hospital with Covid as of 12 April, a 19.5% decrease on the same day the previous month. The number peaked in January 2021 when more than 34,000 people were hospitalised with the virus.
ChatGPT: “However, some healthcare experts argue that the problem is not just one of demand, but also of supply. There is a shortage of doctors and nurses in the UK, and the high turnover rate of healthcare staff means that many A&E departments are struggling to fill key positions. This, in turn, leads to longer waiting times and increased pressure on the staff who are available.”
ChatGPT is right but doesn’t back up its statement on staff shortages with any evidence so I doubt we would run with this. The NHS does have a chronic shortage of staff and the latest figures published by NHS England show tens of thousands of vacancies across nursing and medicine. In December of last year, there were almost 44,000 nursing vacancies and 9,000 medical vacancies (doctors).
ChatGPT: “Another issue is the lack of available beds within hospitals, which can result in patients being kept waiting in A&E departments for longer than necessary. This is often due to a backlog of patients who require admission to hospital, but for whom there are no available beds on the wards.”
Credit where it’s due – bed blocking remains a major issue across the health service but ChatGPT still doesn’t back up its claims with any evidence. The latest NHS figures show almost 13,000 beds were occupied by patients who no longer needed to be there at the end of April.
ChatGPT: “To address these problems, the UK government has pledged to increase funding for the NHS and to hire more healthcare staff. However, healthcare experts argue that these measures alone will not be enough to solve the problem of A&E waiting times.
“Instead, they suggest a more holistic approach, which includes investing in community healthcare services to reduce the number of patients who need to be admitted to hospital, improving patient flow within hospitals, and investing in technology to streamline processes and reduce waiting times. Ultimately, the issue of A&E waiting times in England is a complex one, requiring a multi-faceted approach to address the root causes. However, with the right investments and support, it is possible to ensure that patients receive the care they need in a timely and efficient manner, and that the NHS continues to provide world-class healthcare to all who need it.”
This sort of anecdotal writing wouldn’t normally land in a news article, unless it was credited to an actual source. Who are these healthcare experts? We don’t know because ChatGPT doesn’t tell us.
When I first pitched this story idea I’d be lying if I said I wasn’t a bit apprehensive about talking myself out of a job but needless to say I’ll still be working as a journalist for a while yet. AI is certainly not to be undervalued by the journalism industry though. We would be utterly foolish to not see what is staring us straight in the face and the benefits/risks that come with it. I’ve pulled apart this poor piece of journalism but I don’t doubt we’ll see dramatic improvements in a very short space of time. Whether you’re terrified, fascinated or just utterly confused, journalists need to get with AI and use it to our advantage.
Whether AI will replace me is up for debate and prior to asking ChatGPT to write me an article on A&E waiting times, I asked if it would. It said what we all already know: “journalism requires a level of human intuition, creativity, and critical thinking that cannot be replicated by machines”. Thanks for the reassurance, ChatGPT, but I'm not so confident AI won't radically affect the industry.