This Beanie Is Made to Understand Your Thoughts

The ability to convert speech to text is now integrated into all contemporary computers. But imagine if you could type just by thinking, without having to speak to your device?

Sabi, a startup from Silicon Valley, is unveiling its ambitions with this technology. The company is working on a brain wearable that translates a person’s inner dialogue into text on a screen. CEO Rahul Chhabra mentions that their initial product, a brain-reading beanie, is set to launch by year’s end, along with a baseball cap variant.

This innovation falls under the category of brain-computer interfaces (BCI), devices that create a direct link between the brain and external technology. While firms like Elon Musk’s Neuralink focus on surgically implanted BCIs for those with severe disabilities, Sabi’s approach aims to make cyborg technology accessible to the general populace.

Although it diverges from Musk’s future vision—where humans merge with AI through implanted chips—venture capitalist Vinod Khosla, an early investor in OpenAI, believes that a noninvasive, wearable gadget is essential for widespread BCI adoption.

“The ultimate BCI application is having the ability to communicate with your computer just by thinking,” states Khosla, founder of Khosla Ventures and one of Sabi’s backers. “To successfully engage a billion users in daily computer interaction via BCI, the technology must be noninvasive.”

The technology behind Sabi’s headgear utilizes EEG, or electroencephalography, employing metal discs placed on the scalp to capture the brain’s electrical signals. While it’s currently possible to decode imagined speech from EEG, this capability is largely restricted to a limited set of words or commands, rather than fluid, natural language.

Sensors line the inside of the cap and read brain signals noninvasively.Photograph: Courtesy of Sabi

A challenge with wearable systems is that the sensors must receive brain signals through skin and bone, which can weaken the clarity of those signals. In contrast, implanted devices capture much stronger signals due to their proximity to the neurons. Sabi’s solution for improving accuracy involves exponentially increasing the number of sensors within the device. Most existing EEG devices feature a few dozen to several hundred sensors, while Sabi’s cap is expected to include between 70,000 and 100,000 miniaturized sensors.

“With such high-density sensing, we can precisely determine what and where brain activity is occurring. This data enables us to achieve more reliable interpretations of a person’s thoughts,” Chhabra states.

The company’s target for initial typing speed is around 30 words per minute. Although this is slower than the average typing speed, Chhabra believes that time spent using the cap will enhance this rate.