Understanding super-exponential growth: the challenges and potentials of medical AI development

What challenges does medical AI development face today? Where does the untapped potential lie? In his presentation at this year's MedTech Summit Budapest conference, futurologist Dr. Rab Árpád discussed his experience in medical AI developments. Having witnessed some setbacks while working with US companies, he hopes Hungarians can avoid making the same mistakes.

Healthcare is an outstanding innovation market. As a result, some IT companies try to profit off less responsible solutions, while the problems that arise often get delegated to professional healthcare providers.

Success rates are way higher if a doctor or nurse guides the development process, Rab points out. Thus, the healthcare sector must create knowledge that tech companies do not yet possess.

Why is super-exponential growth a game-changer?

Technology is democratizing extremely fast. Even a year ago, several external players were involved in AI development, and many stakeholders were wary of uploading data to external platforms. Today, the development process can even take place on your home computer.

According to Rab, many people underestimate the role of super-exponential growth. This concept has nothing to do with the speed of data processing. It is about removing the biological factor from the path of knowledge production.

If you have taught a human being something, there is a constant risk that they will forget, grow old, or die. By transferring knowledge to AI, we can overcome this biological barrier, Rab explains.

"For example, the AI-assisted knee surgery robot is a device that makes the necessary adjustments while a doctor makes the right incision.

Today, there are only 2-3 knee specialists in Hungary, and it's hard to get the knowledge out of them. Let's say they have performed 5,000 operations in their lifetime, while these machines can carry out as many as 10 million operations in a year without forgetting the know-how.

Thus, after training them, we can tick off this job and move on to the next one - and that's what super-exponential growth is all about."

Time is on our side

"Artificial intelligence consists of three main parts. The first and most important is data. Its purity, validity, and measurement determine the quality of AI's operation.

The middle element is computation, which is a given. Technology is way ahead of the services we use it for today. The third element is the output, in other words, what we decide to do with the development."

The problem, according to Rab, is that we often deliver complicated answers to simple questions. Companies often try to solve all the potential issues instead of focusing on the task ahead.

For instance, the use of a transcription tool for administrative purposes can save a lot of time for GPs. But what if the AI misinterprets something and the patient gets mistreated? Who's going to take the responsibility? To solve this, Dr. Rab and his team have created a simple interface where the user can select keywords with a single tap, no typing required.

"Most companies want to solve everything at once, but it would be crucial to develop AI solutions in a modular fashion. With organic development aided by super-exponential growth, we win out every single time. Overall, time is our most powerful development tool", Rab says.

The use of hard empathy is another challenge. This term refers to putting ourselves in a situation we have not yet experienced, for example, understanding the needs of an African doctor. We often use easy empathy instead and ask, "What is MY problem?" But this is a grave mistake in the context of development.

"Today, AI is at the level of a 6-8-year-old child. The teaching process is pretty much the same: to provide them with knowledge, test them, and give them frequent feedback. Super-exponentiality leads to rapid success. Such a student will never forget what they have learned, and once they have graduated, you can create any number of students with a tap of Control C and Control V.

The last aspect is honesty, even if it sounds cliché. Participants tend to be dishonest about their interests. Often, an e-learning curriculum does not work because the teacher is concerned that transferring the same knowledge to AI might render them expendable.

Hungary is not lagging in AI development. The knowledge and the money are a given. It's the know-how, the credentials, the vision, and the honesty that counts to have successful developments."

Share this Post: