We write tons of detailed articles about niche topics, for our seasoned audience. However, every now and then it is important to zoom out and just look at the big picture as well. Here we decided to provide a high-level outlook of what’s coming next for AI in healthcare, and hope it will be useful.
So let’s take a birds-eye view of what we can expect from AI in healthcare in the coming period. In this post, we will not discuss abstract principles but will instead focus on providing specific examples. By examining recent advancements and exploring potential future developments, we aim to provide a comprehensive picture of the role AI may play in transforming healthcare.
- Regulation for adaptive algorithms will arrive
Adaptive algorithms are notoriously tricky to regulate due to their constantly changing nature.
So far, only locked algorithms have been approved. Regulating adaptive algorithms presents a significant challenge, as they may change from day to day, becoming different from when they were initially approved. However, the U.S. Food and Drug Administration (FDA) has promised to start addressing this issue this year and is expected to release at least some regulatory descriptions or recommendations providing guidance in this field.
We will also be launching a database where users can search for AI healthcare patents. This resource will enable the identification of trends and predictions about which areas or specialties are likely to stand out in AI developments, potentially providing valuable insights into the future landscape of healthcare.
2. Specialty-based AI recommendations and guidelines will come
In the American College of Radiology, an important development has been the introduction of instructions on how to submit studies, papers, or scientific publications related to AI technology. Guidance is also provided on how radiologists should approach reading a study about AI. It would be beneficial to see such recommendations and guidelines implemented across all medical specialties, helping professionals better understand and utilise AI technologies in their respective fields.
3. Large Language Models (like ChatGPT and MedPaLM) will receive some form of regulation
Large Language Models (like ChatGPT and Google’s MedPaLM) will likely receive some form of regulation this year, as it is impossible that millions of people, including doctors and patients use them daily without any regulatory guidance.
LLM regulation will probably be an entirely new category, distinct from the way other AI-based medical technologies have been regulated so far.
Regulators need to develop oversight mechanisms that consider the unique challenges posed by LLMs, such as interpretability, fairness, and unintended consequences. Tokenization, a crucial aspect of LLMs in natural language processing, is currently not regulated in healthcare and requires attention. On top of that, given LLMs’ broad applicability across various domains, a one-size-fits-all regulatory framework is not suitable. Regulators must address diverse industry-specific concerns, and healthcare – a matter of life and death – will be the most complicated of all.
4. Doomsday Knights will ban large language models
It is very likely that some countries or regions will decide to address potential issues by banning the use of large language models. However, this may not be an effective solution for several reasons. First, even moderately motivated users can easily find ways to circumvent such regulations. Moreover, a better long-term approach would involve the smart utilization of LLMs, the development of appropriate regulations, and education for users on how to work with AI.
Nevertheless, there surely will be individuals who prefer a more extreme approach, akin to “salting the earth” to prevent the impending doom.
5. We will see yet undiscovered medical areas for AI solutions
While the number of applied AI solutions will increase in established areas (e.g., radiology, oncology) we will finally see new ones (e.g., mental health) as well.
This is exciting because even in fields rich in interactions and creative tasks that are less prone to automation, AI usage will begin to emerge. New, innovative models – like paid subscribers can access an AI chatbot but are also routed to a human therapist if their queries escalate – will be developed for these purposes.
6. AI will arrive in the everyday professional lives of healthcare workers
Millions of doctors, nurses, and healthcare workers will experiment with various AI-based tools, such as voice-to-text applications, where they only need to review the output. This will lead to increased efficiency and adoption of AI in healthcare.
There is a wide selection of AI-driven tools, many of them not strictly related to medicine or healthcare, that can help anyone in a wide range of tasks, from building a website to creating videos, from designing presentations to generating FAQ texts or emails.
7. Drug discovery is the Trojan Horse for AI in pharma
AI may break into the pharmaceutical industry through drug discovery. This area has significant potential for cost savings and profit generation for pharmaceutical companies, making it an attractive target for AI applications.
Easy to see why: while traditional drug discovery takes about $1 billion and 10 years to bring a new drug to market, AI can accelerate this process significantly in many ways: creating new molecules virtually (in silico), and this progress is mind-blowing, some algorithms find potential target molecules 1,200 times (!) faster than the previous supercomputer model.
Thus the pipeline of AI-first drug candidates is growing, with 18 drug candidates in clinical trials in 2022, compared to zero in 2020. AI is reportedly taking months, if not years, off of early drug discovery.
8. ChatGPT integration in core healthcare business
This year, dozens or even hundreds of companies will incorporate ChatGPT into their core business. Currently, the most typical use cases are health management and coaching, but other applications are emerging, such as using LLMs as medical scribes.
These tools use large language models to transcribe and summarize in-person and remote medical encounters, including patient-doctor interactions, diagnoses, and treatment plans.
The benefits of implementing LLMs in medical scribe applications are easy to grasp: they can save tons of time, and address a major cause of physician burnout: excessive amounts of administration.
Welcome to the world of The Jetsons!
While we’ve touched upon the incredible potential of AI in healthcare, it’s crucial to realize that these applications are still only a small part of the much broader impact AI can have on our lives. As AI technology continues to grow and seep into various sectors, we might find ourselves living in a world that feels like a page out of the classic cartoon “The Jetsons.”
Wouldn’t it be amazing to live in a Jetsons-esque future, where flying cars solve our traffic woes, robot assistants tend to our needs, and smart home systems manage our households effortlessly? In this high-tech world, AI has the power to revolutionize not just healthcare, but also transportation, education, entertainment, and countless other aspects of daily life. The possibilities seem endless, bounded only by our imagination and commitment to responsible innovation.
The post What’s Next For AI In Healthcare In 2023 appeared first on The Medical Futurist.