Challenger Artificial Intelligence - Everything You Need to Know
Artificial Intelligence is a fundamental shift in computing technology.
It does for brain power what engines did for muscle power. It’s a lever, a multiplier.
Artificial Intelligence is a complementary technology, it does things that a human can’t, or won’t. The same way that engines freed us from distance and the limits of our physical bodies, AI offers us freedom from the limitations of how much we can think about and do.
Artificial Intelligence, As We Know It Today, Isn’t Intelligent
The AI we see today is not intelligent. It’s a very advanced and sophisticated search and summarization algorithm, trained on millions of pieces of data. This is not the AI of the movies - it isn’t Skynet, or HAL 9000. Artificial General Intelligence may be created one day, or our machines and algorithms may just grow so sophisticated that we can’t tell the difference, but today we have a search engine on steroids.
With it, we can perform some amazing technology feats, and it’s just getting started. AI offers breakthroughs in how we talk to computers. Translation, text conversation, voice conversation, image and video generation, immersive environments, image recognition, and in talking to databases.
Your next programming language for computers is likely to be the language you speak every day.
AI Didn’t ‘Just Appear’
The sudden advent of AI algorithms has created a lot of concerns and overwrought claims. Some speak of AI extinguishing humanity, some are trying to make a silicon god that will solve all the unsolvable problems. There’s millions of words written already both in praise and in condemnation of intelligent computing, when in fact nothing has really happened yet.
The truth is a lot more mundane and boring.
The transformer networks that power most of what we see today originated in 2017. They themselves are just a type of neural network technology called transducer networks. You can trace the history of neural networks on computers all the way back to the 1950’s and 1960’s. The concept itself dates back much further than that, when weighted networks were just a mathematical abstraction for solving complex statistical problems.
Programmers and scientists have been working towards better types of artificial brains for decades, and this is just another phase of them. All that’s happened is that the capabilities have risen above the visibility curve - they’re useful now for more general problem solving.
Education Will Go Through Fundamental Shifts
AI is going to create changes everywhere, because it is a fundamental economic shift. Things that required too much expensive human brain power to do can now be done. Rote or mundane tasks can be handed off to some relatively dumb computers. Things that weren’t possible now are economically feasible.
In this environment, human learning is still essential, and is in fact more important than ever.
Already, an AI can tutor and teach in a lot of subjects better than a human. It’s never impatient, it will go through material with an individual as many times as required, it can work through complex topics with people at their own pace, own language, and even their own learning style. The cost is a tiny fraction of a human, far less than 1%.
An AI can assess and manage learning in ways that a human never could, because of the economic cost. These economics are going to change education, whether it wants to be changed or not. They’re going to change medicine and medical education as well. And the world of education is going to be divided into those that embrace the changes, and those that try to cling to the traditional.
Competency-based education is an approach dealing with teaching and proving specific skills and mastery of knowledge, rather than classrooms and seat time as a measure of mastery. The benefits this sort of individualized learning provides are tremendous, and there was a lot of hope in the 1960’s and 1970’s that computers would be able to implement individualized learning for all. The promises of CBE couldn’t be realized because of the human labor cost - the costs were more than we were willing to spend on students, the economics favored structured classroom-based education.
With AI, CBE becomes possible.
AI is on an Exponential Path
Moore’s Law implies that computing power doubles every couple of years, and that the cost of computing gets cut in half. AI algorithmic capability is expanding at an even more rapid pace, as software catches up to computer hardware. But even once they catch up, we can still expect the capabilities of AI to double in usefulness and usability every couple of years.
Humans tend to think in linear patterns, and not exponential curves. When we look at AI and its impact, we need to consider this exponential factor.
In education, another technology is approaching in the form of Extended Reality (XR) - which encompasses virtual reality (VR), augmented reality (AR), and mixed reality (MR) environments. These immersive technologies will also bring a new set of capabilities into education. You can see their earliest forms in today’s AR glasses and VR devices.
With limitations based in material science, these immersive technologies are on a slower growth path - batteries, computing power, visual components and weight all play a factor in adoption and cost. Towards the end of the decade, ubiquitous immersive environments will bring another set of changes to education.
Where Challenger is Using AI
AI doesn’t replace humans in education, or medicine, or anywhere else. Instead it lowers human labor requirements. It’s a complementary technology rather than a replacement. For Challenger, which makes and administers Learning Management Systems and Content Management Systems (course creation), we are deploying and using AI to do things that would have had costs that made them unfeasible just a couple of years ago.
Our view is that in the type of online learning we produce, AI can be applied in three areas in the learning environment. These are the types of things that we are deploying right now, fully expecting technological shifts throughout the decade that will enhance or render them obsolete.
- Data Enhancement - automatic translation, didactic learning material, supplementary learning material, fact checking, grammar checking, spelling, rewriting, monitoring external resources for changes in standards of care or current guidelines, monitoring for articles or controversies, question restructuring, developing core material around learning objectives.
- Application Enhancement - student-paced learning, constant assessment, implementation of CBE (CBME) principles, interactive tutoring, case studies, history simulation, interactive differential diagnosis teaching.
- External Information - material enhancement through pulling in external information on comorbidities, confounding diagnostics, differential testing, pathophysiology outlines and timelines, diagnostic criteria checklists, comparison tables, global health perspectives, treatment pathways and protocols, among others.
The AI is good at this, and getting better, but if it were a student, it would score around 70%-80% on a USMLE Step 3 test. That’s not really a comparison we’d use - people are different from AI. If I sat a high school student down with Google and an unlimited amount of time, they’d easily pass a USMLE Step 3 as well - that’s all we’re doing when we make these comparisons to AI.
What we’re really looking at is the AI training. The AIs we have now are trained in general responses. We expect by the end of 2024 to see more specialized LLMs better trained for medicine.
If You Want to Keep Up
We’re working on a free education course, serialized and updated frequently, for clinicians interested in understanding the technology and its implications in practice and healthcare environments. We also publish newsletters for residency programs, nursing programs, and individual clinicians that keep you abreast of the technology and Challenger’s product capabilities as we adopt it.
