Arnav Kapur, the CEO of AlterEgo, demonstrates the wearable device.Credit: AlterEgo
A wearable device could make saying ‘Alexa, what time is it?’ aloud a thing of the past. The artificial intelligence (AI) neural interface called AlterEgo promises to allow users to silently communicate by just thinking words. Sitting over the ear, the device facilitates daily life through live communication with the Internet. “It gives you the power of telepathy but only for the thoughts you want to share,” says AlterEgo’s chief executive Arnav Kapur, based in Cambridge, Massachusetts. Kapur unveiled the device on 8 September.
The device does not read brain activity, but predicts what a wearer wants to say from signals in muscles used to speak, then sends audio information back into their ear. The researchers say that their non-invasive technology could help people with amyotrophic lateral sclerosis (ALS) and multiple sclerosis (MS) who have trouble speaking, but also want to make the devices commercially available for general use.
In a promotional video on the AlterEgo website, Kapur says that “it’s a revolutionary breakthrough with the potential to change the way we interact with our technology, with one another and with the world around us”.
“The big question about this is ‘how likely is that potential to be realized?,” says Howard Chizeck, an electrical and computer engineer at the University of Washington in Seattle. Chizeck says that the technology seems workable and is less of a privacy risk than listening devices such as Amazon’s Alexa are, but isn’t convinced that the device will catch on for commercial use.
Silent speech
Brain-computer interfaces, like those from Elon Musk’s firm Neuralink,typically require electrodes to be surgically implanted in the brain to record and send signals between neurons. AlterEgo works differently. “It’s using neuromuscular signals that would be used in speech, so it’s like you’re silently talking,” says Chizeck. “The clear advantage is that it’s not invasive. You don’t have to put chips in the brain with risks of infection and surgery.”
The device works by measuring speech patterns in the face. When you speak aloud, silently mouth words or internally articulate words, the information is sent from the brain to the cranial nerves and articulatory muscles, even if the face doesn’t move. AlterEgo detects electrical signals in the articulatory muscles, then sends the data to AI models that predict what the wearer wants to say. The device then communicates audio back to the wearer through bone-conduction headphones.
The AlterEgo project was born in the MIT Media Lab in 2018. The original headset was large and ungainly, says Kapur, and its detection of speech was held back by the capabilities of artificial intelligence at the time. It had limited functionality but could search the web and “order pizzas”. Since then, the engineers have incorporated newer AI capabilities of speech recognition, and the project was spun off as a for-profit company earlier this year.
Clinical applications
The technology is aimed at people with conditions that make speech difficult, such as dysarthria or dysphonia. AlterEgo is currently being trialled in people with ALS or MS.
“A totally locked-in patient wouldn’t be able to use the system, but a lot of the time there is a sparse signal that exists in the speech system and that sparse signal is enough,” says Kapur. “We started with early-stage ALS patients, but we also were able to use the system with late-stage ALS patients.” The trials are ongoing, and results are yet to be published.
“A totally locked in patient wouldn’t be able to use the system, but a lot of the time there is a sparse signal that exists in the speech system and that sparse signal is enough,” says Kapur. “We started with early-stage ALS patients, but we also were able to use the system with late-stage ALS patients.” The trials are still ongoing, and results are yet to be published.
Chizeck thinks that it might be usable for people with early-stage ALS, but because the disease is progressive, gradually reducing people’s ability to move the muscles required for speech, it might be less useful in later stages of the condition. “The device is only as good as the person’s control of speech,” he says.
