We all use artificial intelligence (AI) every day. I regularly ask Siri to call my mom; Siri is my ever-present assistant who remembers all my important phone numbers, whose birthday it is, my grocery list, and when it’s time for bed. While some still think of AI as a futuristic concept, the truth is it’s already part of our everyday lives.
The idea of using AI for help is catching on. According to the 2020 Zendesk Customer Experience Trends Report, people are becoming more comfortable with the idea of interacting with AI if it means fast, efficient resolution to their problems.
Chatbots, for example, can be a great customer service tool—helping to answer questions quickly and route service tickets to the right teams. Algorithms can help companies predict customer needs and serve up recommendations based on preferences. Amazon provides an easy example of a company that uses algorithms that make connections based on your searches and purchase preferences to predict what you want—and also to shield you from what you don’t want.
Artificial intelligence has immense potential to help us do great things, some of which were relegated to wishful thinking a few years ago. Read on for just a few of the ways the power AI is being harnessed for humanity.
1. Predictive speech
Speaking doesn’t come easily to everyone. Frustratingly slow tools can be a particular challenge for people with motor disabilities who often use assistive technology to help them communicate. Research shows that people usually type between five and 20 words per minute, but speak between 100 to 140 words per minute. So, people who need keyboards to communicate can struggle to have a conversation.
Researchers from the universities of Cambridge and Dundee are using AI to help solve this problem. They developed a predictive text tool that analyzes a person’s words and context as they type. With this data, the tool suggests sentences for the person to use, eliminating between 50 to 96 percent of keystrokes.
[Related read: Be an #A11Y—why inclusive design is good design]
2. AI for companionship
For many, the spread of COVID-19 has deeply impacted our ability to connect with other humans, and the resulting loneliness can affect both mind and body. Loneliness has been connected to everything from heart disease to dementia, as large a health threat as smoking 15 cigarettes a day. And here’s an unexpected area where AI can help. At the height of the pandemic, more than half a million people downloaded the Replika app to find companionship in a chatbot. Designed and built by Luka, a California start-up, Replika uses AI to send messages to people who just want to talk to someone (even if they’re not a human). It’s no replacement for a hug, but some users say it helps them get by.
3. Advanced image recognition
Predictive AI can be used to spot things that might be hard for humans to readily see. At the University of León in Spain, scientists worked with the Spanish National Cybersecurity Institute to create a tool to identify objects in crime scene photographs. The team uses the images to train AI to spot crucial clues. The image-recognition tool catalogues information about items in the scene, and can recognize known faces and estimate age and gender. All of this makes it possible for officers to quickly find details without having to manually look through hundreds of photos.
4. Bringing medicines to market, faster
AI can also speed up processes that otherwise take people a long time (and a lot of effort). A biotechnology startup based in Hong Kong used machine learning to create a potential new drug for treating scar tissue. What’s significant here is the timing: only 46 days from molecular design to testing. In comparison, it usually takes more than 10 years and $2.6 billion to bring a new drug to market. Though this is still in the testing stage, there is enormous potential for AI in pharmaceuticals.
5. Identifying at-risk veterans
The U.S. Department of Veterans Affairs piloted an initiative called REACH VET to help save the lives of veterans. Using a predictive model, the department can analyze patient records from the VA to identify veterans at highest risk for suicide. AI uses 61 variables, including missed appointments, brain injuries, and mental health diagnoses, to create a predictive model. When they find individuals at high risk, staff proactively reaches out and suggests treatment options like visiting their social worker. So far, the results have been promising: veterans in the program have fewer missed appointments, fewer inpatient mental health admissions, and lower all-cause mortality.
To solve human problems, AI needs a human touch
AI has great potential, but it can’t replace human intelligence completely. The power of machine learning should be used to empower human beings, not hinder human rights. To that end, the European Union has developed guidelines on the ethical use of AI. These guidelines map out the key requirements AI needs to be trustworthy, including the need to avoid unfair bias that could result in discrimination.
“Most AI systems learn how to function from data. So it's incredibly important to understand the data you're using to train the system — what it contains, its lineage, and if it introduces any liabilities. This is easier said than done,” said Brandon Purcell, principal analyst at Forrester, in the AI Aspirants: Caveat Emptor Report.
AI is only as good as the data you give it. We all have innate biases, even if we don’t always recognize them. Since AI learns from humans, it is apt to learn our biases, too. But keep in mind that biases aren’t all bad.
“Bias can be helpful, identifying important differences between customers that CI pros can use to drive customer acquisition, retention, and loyalty,” said Purcell in another report. “Bias can also be quite harmful, blocking customers from accessing your products and services. Algorithms learn different types of bias in insidiously imperceptible ways.”
Preferential biases are used in business to help serve us the products we might like best, and help to create better customer experiences. For example, when you log into Netflix, you’ll see different recommendations than your neighbor. You want the algorithm to recognize what you like and what you don’t.
The important thing is to ensure prejudicial bias isn’t used against people to do harm. That takes a lot of thoughtful planning, consulting with a diverse range of people for their insights, and a willingness to correct mistakes when they happen.
We’re just getting started when it comes to seeing all the applications that are possible with artificial intelligence. What was once science fiction is now science. And we are only limited by what we can imagine.
Photo credit: Christina Morillo