Artificial Intelligence (AI) is a popular term you’ve probably been hearing a lot lately. But how does AI work as a hearing aid technology?
The term “artificial intelligence” was coined in the 1950s by scientists attempting to create a machine that could perform the functions of “thinking” that humans are capable of. Today, this technology is not just something computer scientists and tech aficionados dabble with; it has crossed into the mainstream. Most of us use AI in our daily lives without really thinking or knowing about it. 

That’s because AI can make life a lot easier. It can do tasks better and faster than humans can, like mapping directions, filtering spam out of your email, or even guessing what word you’re going to type next in a text message.

In more technical fields like healthcare, AI can even save lives. It has been shown to diagnose skin cancer with higher accuracy than experienced doctors

In short, Artificial Intelligence can perform tasks that previously only humans could.

AI enables a computer or machine to carry out a task that a person normally would have to do, like looking at a map and finding the fastest route home. Today, you’d probably use your GPS to help you find your way. That GPS program uses AI to get you home, no matter where you’re starting. It hasn’t memorized the way home; it figures out the best possible path at a give moment. 

This form of AI is very common. All it takes is a graphical representation of a map with dots to indicate the intersections and roads. When the computer in your GPS has this information, the AI can easily find the fastest route to your house. 

You’re also exposed to AI when you shop online. For example, remember a time when you almost bought a pair of pants on a website but decided not to. Then, moments later, you start seeing ads for those very same pants across different websites. That’s because AI knows you’re interested in the pants and is trying to get you to buy them. Or, at least, the company that sells them is--and uses AI to promote them.  

This simple kind of AI is mostly based on mathematics like combinatorics, logics and algorithmics. It uses explicit models, which means that to perform the task that you’re asking of it, you have to represent every possible outcome or condition in the computer, and it will choose among them.

So, for instance, to create an AI-based chess robot, you would need to include every possible move and all the rules for the AI robot to play intelligently and win. The first time a computer used AI to exceed human abilities in chess was in 1997, when the IBM computer Deep Blue beat the world chess champion Garry Kasparov.     

Machine learning takes artificial intelligence to the next level.

A more complex form of AI is called machine learning. That’s when a machine learns structures and can predict future outcomes.

For instance, if you want to teach a computer what a dog looks like, you train it by repeatedly showing the computer pictures of dogs. That way it learns how to categorize a dog as a dog, because it gradually understands the traits that define a dog. It learns not by memorizing but by making up rules to answer the question it is asked.

This form of learning is implicit rather than explicit, and it makes the computer more capable of defining things on its own and predicting outcomes based on experience. Of course, you can never be 100% sure that the computer will always be right in its predictions or categorizations, but machine learning algorithms can become extremely smart extremely quickly. For example, AlphaGo Zero went from complete novice to the arguably the best Go player ever in just 40 days, with no inputs other than the rules of the game.

Machine learning can truly help humans.

AI in its simplest form performs a task that a human could do. But Machine Learning, the deeper form of AI, can predict future outcomes beyond what a human could do.

A good example comes from a healthcare study using a computer to spot more patterns than humans could. The study used Computer Assisted Diagnosis to review early mammography scans of women who later developed breast cancer. The computer predicted 52% of the cancers a year ahead of the official diagnoses. With this kind of information, humans doctors can act faster and more easily prevent the cancer from progressing.  

Artificial intelligence and machine learning as hearing aid technologies.

When it comes to hearing aids, AI is used for different purposes. For many years, it has classified sound environments, so the hearing aid can automatically adjust to the appropriate settings for each environment. Today, this feature is expected of any hearing aid.

Building on that foundation, today hearing aids use AI to personalize hearing in the moment (through technologies like SoundSense Learn). SoundSense Learn uses a machine learning algorithm to calculate the best possible hearing aid parameters for a given situation in just a dozen comparisons. 

To reach the same result, a user would have to compare almost 2,500,000 combinations of hearing aid settings. So, Machine Learning is completing a task which a human would not be able to do in practice, because it has been trained to do so. 

Every time this AI is used, it stores the information it’s given in the cloud, so that it can learn and improve hearing for other users of this feature. And it also remembers your preferred volume settings for every sound class and automatically adjusts the settings the next time you’re in that environment. 

Artificial Intelligence still needs human intervention.

AI will get smarter over time as humans continue to build on the technology, but humans will dictate how AI develops and when it’s successful or not. AI will continue to need humans to give input and keep it up to date.  

As for hearing aids, we can only speculate about what’s to come. One day, hearing aids may be able to perform all sorts of tasks themselves, with no assistance from apps. But one thing is for sure: AI will always need humans to succeed. 

Need help explaining AI to your patients? Check out our animation below: