Brain-Computer Interfaces

Devices that read brain signals to control computers, robotic arms and synthetic voices: BrainGate, speech neuroprostheses, Neuralink and Synchron, non-invasive BCIs, and the ethical questions they raise.

The idea

The term brain-computer interface was coined in 1973 by Jacques Vidal at UCLA. A BCI records brain activity, decodes it with algorithms, and translates it into commands — moving a cursor, typing letters, controlling a robotic arm or synthesising speech. Some also send information back into the brain (for example, touch feedback via stimulation).

Implanted BCIs

In 2004 the BrainGate trial implanted a small Utah array of 96 electrodes in the motor cortex of Matthew Nagle, who was paralysed from the neck down; he could move a computer cursor and open email by imagining movements (published in 2006). In 2012 a woman with paralysis, Cathy Hutchinson, used a BrainGate implant to control a robotic arm and drink coffee from a bottle by herself — the first time in nearly 15 years.

Progress has accelerated. In 2021, Frank Willett, Krishna Shenoy and colleagues decoded imagined handwriting, letting a man type about 90 characters per minute. In August 2023, two teams published speech neuroprostheses in Nature: a Stanford team decoding attempted speech at 62 words per minute, and a UCSF team led by Edward Chang at 78 words per minute, including an animated avatar with a voice. In 2024 a UC Davis team reported a man with ALS communicating with about 97% accuracy within weeks of implantation.

Companies and approaches

Neuralink implanted its first human participant, Noland Arbaugh, in January 2024; its device uses flexible threads inserted by a robot and wireless transmission, and he used it to control a computer, including playing chess. Synchron takes a less invasive route: its Stentrode is delivered through blood vessels to lie against the motor cortex without opening the skull. Other companies, such as Precision Neuroscience and Paradromics, are developing thin-film and high-channel-count systems. Most remain in clinical trials.

Non-invasive BCIs

EEG-based BCIs need no surgery and have been used for spelling (via the P300 response) and controlling wheelchairs or games, but they're slower and less precise because the skull blurs the signals.

Ethics

BCIs raise questions about neural data privacy, long-term safety and support when companies or trials end, identity and agency, equitable access, and potential future use for enhancement rather than therapy.