ChatGPT – why for the first time the future of (medical) authors is uncertain

(This text was originally published in German, I translated it with deepl.com, another cool AI page).

Recently I dealt with the book “Wolff/Weihrauch – Internistische Therapie” (Wolff/Weihrauch – Internal Therapy), published by my father over almost 50 years and written with ambition and great care by about 30 authors. This work is a standard work in its field, is published anew every 2 years and is aimed at physicians practicing internal medicine who would like to look up the current therapy recommendations. For more than 10 years I wrote the chapter for oncology – my own specialty, which I left 3 years ago in favor of the IT industry. Every 2 years, I spent a few days editing my chapter, rereading the current literature, researching, and summarizing my findings in updates to the chapter.

In the 1970s, internal medicine was still a manageable specialty because the options for treating disease were still quite narrow. For medical students in the 60s, there was even still as a textbook the “one-man book,” a book that contained internal medicine and neurology (!) and was written by a single person. In the last 25 years, however, knowledge in internal medicine has virtually “exploded”, which can be seen impressively in oncology through the countless new therapeutic paths. Today, it is hardly possible to be a good oncologist for all diseases, because one cannot stay up-to-date for all tumors.

So how can you keep up with the pace of medical developments, especially if you want to cover large areas in terms of content? “Amboss” is a company, which has created and continues to expand a massive online body of work covering all of medicine. The company accomplishes this with a huge team of physicians who are full-time authors. In addition, they have several medical illustrators who make anatomy books (“Grey’s anatomy”) or other illustration works (“Netter”) obsolete. Started as a company focused on medical students and their exams, they conquered the student market virtually overnight (95% of Germany’s medical students are customers) and have been extending articles to physicians for some time.

The “Wolff/Weihrauch” also requires about 30 renowned authors who condense their knowledge into lines in an excellent way.

And then, ChatGPT came along

I must admit, DALL-E, a product of the company OpenAI has already excited me, even if the results are not perfect yet. DALL-E is an artificial intelligence that “dreams up” images on command. These images can be output as photos as well as oil paintings, impressionism or any other art form. The image to the blog article was created by DALL-E on my command “Robot with speech bubble with text ChatGPT, digital art”, not quite perfect (it says “Chat” instead of “ChatGPT” in the speech bubble).

It’s free, already extremely impressive, and makes you realize that the future of “stock photo” sites is ticked.

But since the launch of ChatGPT, the world is no longer the same – the invention resembles other groundbreaking inventions such as the telephone, television or the Internet. ChatGPT is an artificial intelligence that understands language, grasps inquiries in context, and produces texts. It starts innocuously with questions you can ask ChatGPT (“who was Albert Einstein?”) and goes on to automatically create simple computer software on command.

Students send themselves ChatGPT a week after start as a secret tip, because ChatGPT can already take over a part of the homework (“write an analysis about the meaning of an inflation!”). People who are not proficient in English can have ChatGPT formulate polished emails to their landlord. The possibilities are endless.

But what does this mean for medical writers? In one sentence: to have summaries formulated (royalty/rights free!) that make much of the work redundant.

I challenged ChatGPT to write something about chronic obstructive pulmonary disease with this sentence:

“Write a medical article about the therapy of COPD with indication of medications!”

And that’s what ChatGPT returned to me within seconds (initially in German, I translated it with deepl.com):

Chronic obstructive pulmonary disease (COPD) is a progressive lung disease characterized by difficulty breathing, coughing and sputum. Treatment for COPD aims to relieve symptoms and improve lung function. In advanced stages of the disease, therapy can also help extend life expectancy.

An important treatment option for COPD is the use of medications. These include bronchodilator medications, which widen the airways to relieve breathing difficulties. The most commonly used bronchodilator medications in COPD are long-acting beta-2 sympathomimetics (e.g., salmeterol and formoterol) and anticholinergic medications (e.g., tiotropium).

Anti-inflammatory medications may also be used to reduce inflammation in the airways. These include glucocorticoids, for example, which are usually given in combination with a bronchodilator medication. However, glucocorticoids can also have side effects and should therefore be taken under medical supervision.

In advanced stages of COPD, the use of oxygen therapy may also be useful. This involves delivering oxygen to the patient through a nasal mask or oxygen concentrator to improve oxygen delivery to the body.

To relieve COPD symptoms and extend life expectancy, it is important that sufferers take their treatment regularly and follow their doctor’s instructions. Early diagnosis and treatment of COPD can help relieve symptoms and improve lung function.

Then I wrote ChatGPT:

“Write something about diagnostics in COPD!”.

The answer:

The diagnosis of COPD is usually made on the basis of clinical symptoms and the results of lung function tests. An important test in the diagnosis of COPD is spirometric measurement, which measures lung function. This determines maximum respiratory flow and lung capacity to determine if COPD is present and how severe it is.

In addition, X-rays of the lungs and blood tests may be performed to provide further information about the condition of the lungs and to rule out possible other causes of breathing difficulties.

In advanced stages of COPD, a bronchoscopy may also be performed to examine the airways and lungs more closely and to determine if any changes have occurred in the lungs.

Overall, it is important that COPD is diagnosed early in order to initiate effective treatment and relieve symptoms.

To which I asked for the following addition:

“What is the dosage of anticholinergic medications?”

The answer:

The dosage of anticholinergic medication in the treatment of COPD depends on various factors, such as the patient’s age, the severity of the disease and the individual’s response to the medication. It is therefore important that the dosage is determined by a physician and reviewed regularly to ensure that it is optimal.

Usually, anticholinergic medications are administered in the form of inhalers. The recommended starting dose is usually one inhaled puff of the medication once or twice daily. The dosage can be adjusted depending on the patient’s needs and tolerance.

It is important that anticholinergic medications be used regularly and correctly for optimal effect. Patients should therefore follow their doctor’s instructions exactly and have regular check-ups with their doctor.

It is easy to see that the answers are already almost “perfect”. ChatGPT is also, like any good software, a work-in-progress and will be continuously improved. The learning curve is rapid and the U$1 billion (!) that Microsoft has injected in 2019 into the company that Elon Musk founded with others will have secured its future.

This technique shows that ChatGPT can soon (1-5 years?) help to massively reduce the workload of authors and speed up the update cycle. One must then also ask whether pre-authored therapy texts are even necessary anymore – after all, in the future a physician can directly ask ChatGPT (or a similar AI) and possibly receive a research-current answer in seconds.

In concrete terms, this means that a medical author could have ChatGPT create the article in the first step. Then he/she would only have to edit it. The expert is given his/her own “writing assistant” by the AI. The better ChatGPT will be trained with medical data, the more superfluous the expert will possibly become.

Should it become possible to let ChatGPT read and “understand” the entire Pubmed (continuously), it may be possible to disruptively shake entire content companies with one line.

“ChatGPT, write me my own ‘Amboss’!”