Artificial Intelligence often gets talked about like it’s some kind of magic. There is a perception that if you just take a powerful enough program, put it in a powerful enough machine and just point it at any problem, it will spit out the answer in a matter of minutes if not seconds. Unfortunately, that isn’t the way things work.
For instance, recently, there were some rather breathless headlines about how AI had just figured out how to recognize the most at risk COVID-19 patients. Finally, we would be able to quickly recognize those who need treatment the soonest, allowing hospitals to prioritize their resources accordingly. Yet, this isn’t really true.
So, what is wrong with the statement? Plenty of things. First, the AI didn’t actually solve anything, the person who wrote the algorithm did. See, true artificial intelligence doesn’t really exist yet. All any program can do is process the information it is given based on the algorithms that are programmed into it. At best, they can learn how to process information better and reach related conclusions faster. It also didn’t necessarily help anyone identify the most at risk patients. At the most all it did was take a bunch of symptoms we already know and sort through patients with and these other symptoms faster.
Just because someone has a certain list of symptoms, doesn’t mean they are most at risk. It could be a thirty year-old who is otherwise healthy and will recover quickly. He looks bad now, but in two or three days he may well be walking out of the hospital. On the other hand, a sixty year-old with mild symptoms but an advanced case of diabetes may be dead in two days, even with treatment. That is a significant consideration given that the vast majority of deaths from COVID have been the elderly with pre-existing conditions. Diabetes in fact is one of the deadliest conditions to have for those who contract COVID.
It isn’t that the algorithm isn’t useful or that it can’t be improved, it’s just that, as it is, it’s reactive only. It only tells us and helps us identify what has already happened. That isn’t artificial intelligence, it’s data analysis, analysis that only helps see what is in front of us and so isn’t actually a great planning tool.
TARTLE by contrast is trying to make real progress in machine learning, helping to write algorithms that don’t just help us process the information we already have and make sense of it, but to use that information to make predictions about the future.
Think of it as cause and effect. If we can understand thoroughly the causes, we can make reasonable predictions based on that. I’m not sitting here typing this just because I decided to sit down and do it. A whole host of causes fed into this moment, choices made over years, choices that could be traced back to buying my first laptop two decades ago. If we or someone else could take your data now and understand it, predictions about what your future may look like would be possible.
Analyzing data like this will be extremely useful in the medical field. If it were possible to gather certain basic data (heart rate, blood pressure, weight, and eating habits) we could make solid predictions on the likelihood that you’ll have heart problems, be overweight, or develop cancer. True even with something as objective and quantifiable as medical data the predictions won’t be perfect. There are simply too many variables in the real world, especially when dealing with people, to expect perfect accuracy. However, it only needs to provide probabilities in order to make a difference in people’s lives. Think about it, if after we or a medical researcher looked at your health data and came back saying there is a 90% chance you’ll die in ten years unless you change things will you make some changes or play the odds, counting on that 10% of uncertainty? I’m willing to bet that most people will make those changes.
That’s the world that TARTLE is trying to build, a world where when you do decide to sell your data, it is going not just to a faceless company but a group that is taking that data and using it to help you and others as well.
What’s your data worth?