Artificial intelligence (AI) is one of the hottest fads in technology right now. Just Google the phrase and you’ll see post after post talking about AI in relation to everything from autonomous cars to energy production. AI is even a hot topic in healthcare. It’s often mentioned side-by-side with advanced signal processing techniques that, when combined, seem to have quite a bit of potential.
As exciting as AI and signal processing are, we all need to temper our expectations. The one thing history has taught us about technology is that hype rarely matches reality. Just because combining AI and signal processing shows great potential does not mean that this potential is achievable now, or ever will be. Potential is just that. It never becomes reality until someone figures out how to do it.
U.S. News & World Report published a great article on September 26 (2018) discussing artificial intelligence and its potential for treating cancer. What the article reveals perfectly illustrates the need to temper expectations. All the hype surrounding AI’s ability to improve cancer treatment is just that: hype.
The Artificial Intelligence Principal
AI is not a new concept. Scientists have been working on it since the mid-1940s. But only in the last two decades has the science of AI advanced enough to produce viable results we can actually use. Perhaps that’s why there’s so much hype surrounding AI. Be that as it may, let’s talk about what AI actually does.
True artificial intelligence – the ability of a machine to think autonomously – does not actually exist. What we call AI really is a series of advanced computer algorithms capable of comparing multiple levels of data in order to reach data-driven conclusions.
The AI being utilized to improve cancer treatment involves what is known as deep learning. What is deep learning? It is the process just described above. It is a process that involves going through multiple layers of data and comparing key points in each layer. Those points are weighted for importance, giving the computer algorithm some direction as to what to do with the information.
All artificial intelligence systems require a human teacher to instruct machines about what to do with the data at hand. As such, the machines are not really thinking at all. They are merely responding to inputs resulting from human thought.
AI and Signal Processing
Rock West Solutions, a California company that specializes in signal processing technologies for the healthcare sector, explains that better signal processing improves AI by giving AI systems more accurate data to work with. Combining signal processing with AI can theoretically improve cancer diagnoses and treatment. Researchers are especially excited about the ability of the combined sciences to improve early detection capabilities.
All of this sounds very good, indeed. So what’s the problem? Both signal processing and AI are limited to the finite knowledge of existing data sets. Even though the sheer volume of data collected globally in the last 10 years is almost unfathomable, there still are limits to the information thereof. Those limits are intrinsically determined by the capacity of the human brain.
While signal processing and AI can certainly improve diagnostics and treatment methodologies, they are incapable of predicting outcomes. Furthermore, they rely heavily on human understanding of the data being analyzed.
In the end, a lot of what we are hearing about AI as a tool for improving healthcare is just hype. In fact, that is probably true in every area where AI is being explored. It would be best for us to slow down a bit and temper our expectations.