The company is now working with the Champalimaud Foundation in Lisbon to explore potential applications for patients with conditions like ALS and Cerebral Palsy who struggle to communicate verbally. The device is expected to be deployed to the first ALS users by the end of the year. Unbabel's Halo is currently capable of around 20 words per minute of equivalent communication, with a target of 80 words per minute for consumer use.
Key takeaways:
- Unbabel, a language translation services startup, is developing a brain-to-computer interface that uses an EMG system and Generative AI to translate thoughts into text messages.
- The system, dubbed 'Halo', measures how the wearer of an EMG device reacts when thinking of a word, building up a set of signals which correlate to real words and feeding those signals into an LLM to create a 'personalized LLM'.
- Unbabel is working with the Champalimaud Foundation in Lisbon to potentially use the system for patients with ALS and other conditions that make communication difficult.
- The current version of Halo can achieve around 20 words per minute of equivalent communication, with a target of 80 words per minute for consumer use.