I am a parent, futurist, and technologist. My career has spanned the birth of personal computers to the rise of cloud computing.
Can Machine Intelligence (AI) adapt when the variable is its creator?
One of the concepts in AI, or what I prefer to call machine intelligence, is the concept of adaptation. Most of us learn about adaptation when we are introduced to the theory of evolution by Charles Darwin. I will paraphrase Charles Darwin; species adapt to their environment to survive. That concept of survival increases based on the species' ability to adapt. Most of us have already seen adaptation in the world of machine learning. Suppose you use Google docs or Microsoft Word and are typing a document. As you type many times, additional recommended text appears on the computer in gray ahead of what you are typing. It isn't perfect, by the way. Writing this article and starting with Charles Darwin has finished my adaptation sentences with the biological use of adaptation. Not the Machine Intelligence or AI adaptation I am typing. But that's because most people who use the word adaptation tend to be biologists. So today, we were to talk a little bit about the concept of adaptation.
I've already given you one form of Machine Intelligence and adaptation, the use of type ahead in either an official document or other documents where you are typing. AI-driven type ahead is not completely new, and most email programs support the concept of type ahead by utilizing the local cache on your machine to reuse email addresses you've used in the past. If it recognizes the start of one of the names you use, it finishes the address. That is an early form of Machine Learning and a very early form of adaptation. The computer is adapting to the way you use email. Now, adaptation has several applications far beyond simply adding words ahead as you type. One of the extensive uses of adaptation is the automation systems required for the driverless car. Suppose you consider the number of options and variables within the driving process. In that case, it's very easy to see that the computer system must adapt to multiple changes. Add ice to a road, and you drive differently than if you add rain. Although, in some parts of the world, when it rains, people speed up. At least where I live, that's what happens.
But the autonomous car will have to adapt to the road conditions and the people driving near it. Humans, at that point, become a variable; that is, we don't always do the predictable thing. My favorite example of that is someone that is scared of spiders and someone that is terrified. Someone scared of spiders will keep a distance from a spider they see. A person terrified of spiders will freeze, and their brain will immediately begin processing the fight or flight options. Likely if they are terrified, they will flee. But in building autonomous cars, flight or fight is not an adaptation we can utilize. When building an autonomous car, it's more important than the car understands the reaction of the humans around t but also the car it is operating. That's a lot of adaptation. Now computers can perform calculations at a much faster rate than human beings. Computers gather information at a much higher rate than humans. The computer takes input from the Lidar system, the camera system, and the car's other road sensor. Lidar is the system used for "lane protection" systems. It is a form of radar that constantly measures the distance between you and the car next to you. Cross into a lane with a car in it, and the system beeps at you. Don't pull into the lane next to you. There's a car there already. So an autonomous vehicle gathers more information more quickly than a human driver.
When the variable is human, the reality is scary.
Of course, the struggle for an autonomous car will be the other human drivers on the road. If you take a hundred human beings and put them on an icy road, they will turn into a slide if you're lucky. However, the reality is a hundred drivers on an icy road; the likelihood is more of them will turn against the skid and therefore cause a crash. The autonomous car must prepare for people acting correctly and incorrectly. That ability or adaptation is the true value of artificial intelligence for machine intelligence in the future. When the systems can adapt to the variable of the human driver next to them, you will begin to see more and more autonomous cars on the road. All that end with this simple view of tomorrow. The more machine systems can adapt to human variability and the variability of the system they are operating, the more value there will be in the autonomous system. The vast majority of road accidents in the United States are rear-ended collisions. If we reduce rear-end crashes by just 50%, the resulting safety addition to drivers on the road will be massive. Imagine a world where you as a person don't have to adapt to not having your car for a week as the body shop repairs the damage done in the rear-end collision. Let alone the medical problems caused by being in a car that stopped or nearly stopped and hit by another vehicle going fast, causing you to suffer whiplash or worse. Like birds that developed longer peaks to handle and consume nuts with different shells, autonomous systems will eventually adapt to the variable of humanity over time. I can't wait for that day to come!
This article is accurate and true to the best of the author’s knowledge. Content is for informational or entertainment purposes only and does not substitute for personal counsel or professional advice in business, financial, legal, or technical matters.
© 2022 DocAndersen