1. One thing that helps
Can we use voice analysis to diagnose disease? Google thinks so.
Google’s HeAR (Health Acoustic Representations) is an AI model that uses sounds for early disease diagnosis. It was trained on a data set of around 300 million audio clips, including breathing and coughing. This allows HeAR to distinguish between sounds and detect early signs of respiratory illnesses such as tuberculosis. Another company, Klick Labs, has developed AI that uses voice recordings to detect high blood pressure. Their model analyses subtle vocals to predict hypertension with accuracy ratings of 84% for women and 77% for men.
Bioacoustics is more than a novelty in medical diagnostics. Using a smartphone’s microphone with the AI software installed, medical teams can have a diagnostics lab on the go that they can bring to even the remotest places. It’s cheap, handy, and reliable. This is another example of technology bringing healthcare to everyone.
2. One to be wary of
“Social media is great – I just wish it had more bots and fewer real people” – said nobody ever.
A new social media app called SocialAI promises “a private social network where you receive millions of AI-generated comments offering feedback, advice & reflections on each post you make”. Did anybody else break out in a cold sweat when they read that?
“SocialAI is designed to help people feel heard, and to give them a space for reflection, support, and feedback that acts like a close-knit community,” says Michael Sayman, SocialAI’s creator. To achieve that goal, SocialAI is designed to to be a place where bots constantly like your posts. You get to choose whether your followers are “supporters,” “nerds,” or “skeptics.” They then react to your posts accordingly.
Sigh…
3. One to amaze
Are you not keen on Elon inserting things into your brain? OK, I understand that. It would still be cool to be able to just “think” what you want and have it produced in text (and hopefully one-day movement and audio).
Researchers at the Ecole Polytechnique Federale de Lausanne in Switzerland have developed a miniaturized brain-machine interface, or MiBMI for short. This BMI measures just 8 mm2 compared to Neuralink’s 184 mm2. It can process real-time data using less power without implanting 64 electrodes into your brain (not ever Elon). At the moment, MiBMI is meant for people with severe motor impairment who find it challenging to communicate with others. The device’s primary use is to convert thought to speech, and it currently claims an impressive 91% accuracy rate.
The high accuracy rate is due to the use of distinctive neural codes. These are neural markers that fire when the user imagines a specific letter. Using these markers, which carry a size of just about a hundred bytes each, the speed of translation from thought to text is shortened. It currently can decode 31 characters, but it is hoped to be expanded to 100.
The next step is implementing MiBMI for processes other than writing text. The brain-to-machine interface is the next frontier and one that is incredibly exciting.