In an initial study, Canary analyzed speech recordings with AI and identified early stage Alzheimer’s with 96 percent accuracy.
Dave Arnold retired in his 60s and began spending time volunteering in local schools. But then he started misplacing items, forgetting appointments and losing his sense of direction. Eventually he was diagnosed with early stage Alzheimer’s.
“Hearing the diagnosis made me very emotional and tearful,” he said. “I immediately thought of all my mom had experienced.” His mother suffered with the condition for years before passing away. Over the last year, Arnold has worked for the Alzheimer’s Association as one of its early stage advisors, sharing his insights to help others in the initial stages of the disease.
Arnold was diagnosed sooner than many others. It's important to find out early, when interventions can make the most difference. One promising avenue is looking at how people talk. Research has shown that Alzheimer’s affects a part of the brain that controls speech, resulting in small changes before people show other signs of the disease.
Now, Canary Speech, a company based in Utah, is using AI to examine elements like the pitch of a person’s voice and their pauses. In an initial study, Canary analyzed speech recordings with AI and identified early stage Alzheimer’s with 96 percent accuracy.
Developing the AI model
Canary Speech’s CEO, Henry O’Connell, met cofounder Jeff Adams about 40 years before they started the company. Back when they first crossed paths, they were both living in Bethesda, Maryland; O’Connell was a research fellow at the National Institutes of Health studying rare neurological diseases, while Adams was working to decode spy messages. Later on, Adams would specialize in building mathematical models to analyze speech and sound as a team leader in developing Amazon's Alexa.
It wasn't until 2015 that they decided to make use of the fit between their backgrounds. ““We established Canary Speech in 2017 to build a product that could be used in multiple languages in clinical environments,” O'Connell says.
The need is growing. About 55 million people worldwide currently live with Alzheimer’s, a number that is expected to double by 2050. Some scientists think the disease results from a buildup of plaque in the brain. It causes mild memory loss at first and, over time, this issue get worse while other symptoms, such as disorientation and hallucinations, can develop. Treatment to manage the disease is more effective in the earlier stages, but detection is difficult since mild symptoms are often attributed to the normal aging process.
O’Connell and Adams specialize in the complex ways that Alzheimer’s effects how people speak. Using AI, their mathematical model analyzes 15 million data points every minute, focusing on certain features of speech such as pitch, pauses and elongation of words. It also pays attention to how the vibrations of vocal cords change in different stages of the disease.
To create their model, the team used a type of machine learning called deep neural nets, which looks at multiple layers of data - in this case, the multiple features of a person’s speech patterns.
“Deep neural nets allow us to look at much, much larger data sets built out of millions of elements,” O’Connell explained. “Through machine learning and AI, we’ve identified features that are very sensitive to an Alzheimer’s patient versus [people without the disease] and also very sensitive to mild cognitive impairment, early stage and moderate Alzheimer's.” Based on their learnings, Canary is able to classify the disease stage very quickly, O’Connell said.
“When we’re listening to sublanguage elements, we’re really analyzing the direct result of changes in the brain in the physical body,” O’Connell said. “The brain controls your vocal cords: how fast they vibrate, the expansion of them, the contraction.” These factors, along with where people put their tongues when talking, function subconsciously and result in subtle changes in the sounds of speech.
Further testing is needed
In an initial trial, Canary analyzed speech recordings from phone calls to a large U.S. health insurer. They looked at the audio recordings of 651 policyholders who had early stage Alzheimer’s and 1018 who did not have the condition, aiming for a representative sample of age, gender and race. They used this data to create their first diagnostic model and found that it was 96 percent accurate in identifying Alzheimer’s.
Christian Herff, an assistant professor of neuroscience at Maastricht University in the Netherlands, praised this approach while adding that further testing is needed to assess its effectiveness.
“I think the general idea of identifying increased risk for cognitive impairment based on speech characteristics is very feasible, particularly when change in a user’s voice is monitored, for example, by recording speech every year,” Herff said. He noted that this can only be a first indication, not a full diagnosis. The accuracy still needs to be validated in studies that follows individuals over a period of time, he said.
Toby Walsh, a professor of artificial intelligence at the University of New South Wales, also thinks Canary’s tool has potential but highlights that Canary could diagnose some people who don’t really have the disease. “This is an interesting and promising application of AI,” he said, “but these tools need to be used carefully. Imagine the anxiety of being misdiagnosed with Alzheimer’s.”
As with many other AI tools, privacy and bias are additional issues to monitor closely, Walsh said.
Other languages
A related issue is that not everyone is fluent in English. Mahnaz Arvaneh, a senior lecturer in automatic control and systems engineering at the University of Sheffield, said this could be a blind spot.
“The system may not be very accurate for those who have English as their second language as their speaking patterns would be different, and any issue might be because of language deficiency rather than cognitive issues,” Arvaneh said.
The team is expanding to multiple languages starting with Japanese and Spanish. The elements of the model that make up the algorithm are very similar, but they need to be validated and retrained in a different language, which will require access to more data.
Recently, Canary analyzed the phone calls of 233 Japanese patients who had mild cognitive impairment and 704 healthy people. Using an English model they were able to identify the Japanese patients who had mild cognitive impairment with 78 percent accuracy. They also developed a model in Japanese that was 45 percent accurate, and they’re continuing to train it with more data.
The future
Canary is using their model to look at other diseases like Huntington’s and Parkinson’s. They’re also collaborating with pharmaceuticals to validate potential therapies for Alzheimer’s. By looking at speech patterns over time, Canary can get an indication of how well these drugs are working.
Dave Arnold and his wife dance at his nephew’s wedding in Rochester, New York, ten years ago, before his Alzheimer's diagnosis.
Dave Arnold
Ultimately, they want to integrate their tool into everyday life. “We want it to be used in a smartphone, or a teleconference call so that individuals could be examined in their home,” O’Connell said. “We could follow them over time and work with clinical teams and hospitals to improve the evaluation of patients and contribute towards an accurate diagnosis.”
Arnold, the patient with early stage Alzheimer’s, sees great promise. “The process of getting a diagnosis is already filled with so much anxiety,” he said. “Anything that can be done to make it easier and less stressful would be a good thing, as long as it’s proven accurate.”
In the realm of cancer, Dr. Song is similarly setting his sights on another group of patients for whom treatment options are few and far between: people with solid tumors. Whereas some gradual progress has been made in treating blood cancers such as certain leukemias in past few decades, solid tumors have been even more of a challenge. But Dr. Song’s approach of using natural killer cells to treat solid tumors is promising. You may have heard of CAR-T, which uses genetic engineering to introduce cells into the body that have a particular function to help treat a disease. NKGen focuses on other means to enhance the 40 plus receptors of natural killer cells, making them more receptive and sensitive to picking out cancer cells.
Paul Y. Song, MD is currently CEO and Vice Chairman of NKGen Biotech. Dr. Song’s last clinical role was Asst. Professor at the Samuel Oschin Cancer Center at Cedars Sinai Medical Center.
Dr. Song served as the very first visiting fellow on healthcare policy in the California Department of Insurance in 2013. He is currently on the advisory board of the Pritzker School of Molecular Engineering at the University of Chicago and a board member of Mercy Corps, The Center for Health and Democracy, and Gideon’s Promise.
Dr. Song graduated with honors from the University of Chicago and received his MD from George Washington University. He completed his residency in radiation oncology at the University of Chicago where he served as Chief Resident and did a brachytherapy fellowship at the Institute Gustave Roussy in Villejuif, France. He was also awarded an ASTRO research fellowship in 1995 for his research in radiation inducible gene therapy.
With Dr. Song’s leadership, NKGen Biotech’s work on natural killer cells represents cutting-edge science leading to key findings and important pieces of the puzzle for treating two of humanity’s most intractable diseases.
Show links
- Paul Song LinkedIn
- NKGen Biotech on Twitter - @NKGenBiotech
- NKGen Website: https://nkgenbiotech.com/
- NKGen appoints Paul Song
- Patient Story: https://pix11.com/news/local-news/long-island/promising-new-treatment-for-advanced-alzheimers-patients/
- FDA Clearance: https://nkgenbiotech.com/nkgen-biotech-receives-ind-clearance-from-fda-for-snk02-allogeneic-natural-killer-cell-therapy-for-solid-tumors/Q3 earnings data: https://www.nasdaq.com/press-release/nkgen-biotech-inc.-reports-third-quarter-2023-financial-results-and-business
Some researchers argue that active, playful engagement with a "robot nanny" for a few hours a day is better than several hours in front of a TV or with an iPad.
A 2017 study led by Google executive James Manyika suggested that skills like creativity, emotional intelligence, and communication will always be needed in the classroom and that robots aren’t likely to provide them at the same level that humans naturally do. But robot teachers do bring advantages, such as a depth of subject knowledge that teachers can’t match, and they’re great for student engagement.
The teacher and robot can complement each other in new ways, with the teacher facilitating interactions between robots and students. So far, this is the case with teaching “assistants” being adopted now in China, Japan, the U.S., and Europe. In this scenario, the robot (usually the SoftBank child-size robot NAO) is a tool for teaching mainly science, technology, engineering, and math (the STEM subjects), but the teacher is very involved in planning, overseeing, and evaluating progress. The students get an entertaining and enriched learning experience, and some of the teaching load is taken off the teacher. At least, that’s what researchers have been able to observe so far.
To be sure, there are some powerful arguments for having robots in the classroom. A not-to-be-underestimated one is that robots “speak the language” of today’s children, who have been steeped in technology since birth. These children are adept at navigating a media-rich environment that is highly visual and interactive. They are plugged into the Internet 24-7. They consume music, games, and huge numbers of videos on a weekly basis. They expect to be dazzled because they are used to being dazzled by more and more spectacular displays of digital artistry. Education has to compete with social media and the entertainment vehicles of students’ everyday lives.
Another compelling argument for teaching robots is that they help prepare students for the technological realities they will encounter in the real world when robots will be ubiquitous. From childhood on, they will be interacting and collaborating with robots in every sphere of their lives from the jobs they do to dealing with retail robots and helper robots in the home. Including robots in the classroom is one way of making sure that children of all socioeconomic backgrounds will be better prepared for a highly automated age, when successfully using robots will be as essential as reading and writing. We’ve already crossed this threshold with computers and smartphones.
Students need multimedia entertainment with their teaching. This is something robots can provide through their ability to connect to the Internet and act as a centralized host to videos, music, and games. Children also need interaction, something robots can deliver up to a point, but which humans can surpass. The education of a child is not just intended to make them technologically functional in a wired world, it’s to help them grow in intellectual, creative, social, and emotional ways. When considered through this perspective, it opens the door to questions concerning just how far robots should go. Robots don’t just teach and engage children; they’re designed to tug at their heartstrings.
It’s no coincidence that many toy makers and manufacturers are designing cute robots that look and behave like real children or animals, says Turkle. “When they make eye contact and gesture toward us, they predispose us to view them as thinking and caring,” she has written in The Washington Post. “They are designed to be cute, to provide a nurturing response” from the child. As mentioned previously, this nurturing experience is a powerful vehicle for drawing children in and promoting strong attachment. But should children really love their robots?
ROBOTS AND THE PEOPLE WHO LOVE THEM: Holding on to Our Humanity in an Age of Social Robots by Eve Herold (January 9, 2024).
St. Martin’s Publishing Group
The problem, once again, is that a child can be lulled into thinking that she’s in an actual relationship, when a robot can’t possibly love her back. If adults have these vulnerabilities, what might such asymmetrical relationships do to the emotional development of a small child? Turkle notes that while we tend to ascribe a mind and emotions to a socially interactive robot, “simulated thinking may be thinking, but simulated feeling is never feeling, and simulated love is never love.”
Always a consideration is the fact that in the first few years of life, a child’s brain is undergoing rapid growth and development that will form the foundation of their lifelong emotional health. These formative experiences are literally shaping the child’s brain, their expectations, and their view of the world and their place in it. In Alone Together, Turkle asks: What are we saying to children about their importance to us when we’re willing to outsource their care to a robot? A child might be superficially entertained by the robot while his self-esteem is systematically undermined.
Research has emerged showing that there are clear downsides to child-robot relationships.
Still, in the case of robot nannies in the home, is active, playful engagement with a robot for a few hours a day any more harmful than several hours in front of a TV or with an iPad? Some, like Xiong, regard interacting with a robot as better than mere passive entertainment. iPal’s manufacturers say that their robot can’t replace parents or teachers and is best used by three- to eight-year-olds after school, while they wait for their parents to get off work. But as robots become ever-more sophisticated, they’re expected to perform more of the tasks of day-to-day care and to be much more emotionally advanced. There is no question children will form deep attachments to some of them. And research has emerged showing that there are clear downsides to child-robot relationships.
Some studies, performed by Turkle and fellow MIT colleague Cynthia Breazeal, have revealed a darker side to the child-robot bond. Turkle has reported extensively on these studies in The Washington Post and in her book Alone Together. Most children love robots, but some act out their inner bully on the hapless machines, hitting and kicking them and otherwise trying to hurt them. The trouble is that the robot can’t fight back, teaching children that they can bully and abuse without consequences. As in any other robot relationship, such harmful behavior could carry over into the child’s human relationships.
And, ironically, it turns out that communicative machines don’t actually teach kids good communication skills. It’s well known that parent-child communication in the first three years of life sets the stage for a very young child’s intellectual and academic success. Verbal back-and-forth with parents and care-givers is like fuel for a child’s growing brain. One article that examined several types of play and their effect on children’s communication skills, published in JAMA Pediatrics in 2015, showed that babies who played with electronic toys—like the popular robot dog Aibo—show a decrease in both the quantity and quality of their language skills.
Anna V. Sosa of the Child Speech and Language Lab at Northern Arizona University studied twenty-six ten- to sixteen- month-old infants to compare the growth of their language skills after they played with three types of toys: electronic toys like a baby laptop and talking farm; traditional toys like wooden puzzles and building blocks; and books read aloud by their parents. The play that produced the most growth in verbal ability was having books read to them by a caregiver, followed by play with traditional toys. Language gains after playing with electronic toys came dead last. This form of play involved the least use of adult words, the least conversational turntaking, and the least verbalizations from the children. While the study sample was small, it’s not hard to extrapolate that no electronic toy or even more abled robot could supply the intimate responsiveness of a parent reading stories to a child, explaining new words, answering the child’s questions, and modeling the kind of back- and-forth interaction that promotes empathy and reciprocity in relationships.
***
Most experts acknowledge that robots can be valuable educational tools. But they can’t make a child feel truly loved, validated, and valued. That’s the job of parents, and when parents abdicate this responsibility, it’s not only the child who misses out on one of life’s most profound experiences.
We really don’t know how the tech-savvy children of today will ultimately process their attachments to robots and whether they will be excessively predisposed to choosing robot companionship over that of humans. It’s possible their techno literacy will draw for them a bold line between real life and a quasi-imaginary history with a robot. But it will be decades before we see long-term studies culminating in sufficient data to help scientists, and the rest of us, to parse out the effects of a lifetime spent with robots.
This is an excerpt from ROBOTS AND THE PEOPLE WHO LOVE THEM: Holding on to Our Humanity in an Age of Social Robots by Eve Herold. The book will be published on January 9, 2024.