Scientists Develop Non-Invasive AI-BCI | ForkLog
Researchers at the University of California, Los Angeles (UCLA) have unveiled a wearable, non-invasive brain-computer interface that pairs with artificial intelligence as an “autopilot” to speed up and sharpen control of devices like robotic arms and computer cursors, according to Neuroscience News.
The prototype reads brain activity using electroencephalography (EEG) and converts those signals into movement commands through custom decoding algorithms. A complementary AI-driven vision system—described by the team as AI cameras—interprets the user’s intent in real time, enabling collaborative control that is both faster and more precise than EEG alone.
“By using artificial intelligence in addition to brain-computer interface systems, we aim to find much less risky and non-invasive approaches,” said Jonathan Kao, the study’s lead researcher and an assistant professor of electrical and computer engineering at UCLA’s Samueli School of Engineering.
How the system was tested
The team evaluated the AI-assisted BCI with four participants, including one person with paralysis. Across tasks, the combined EEG-and-AI approach consistently outperformed control without AI, and in some cases enabled results that were otherwise not possible.
- Cursor navigation: Participants moved a mouse cursor among eight on-screen targets, dwelling for at least half a second on each target before moving on.
- Robotic manipulation: Participants used a robotic arm to pick up and move blocks on a tabletop—an embodiment task that demands coordinated, multi-step control.
Notably, the participant with paralysis successfully completed the robotic arm challenge in roughly six and a half minutes with AI assistance. Without the AI co-pilot, the task could not be completed at all.
Why this matters
Today’s most capable BCI systems often rely on surgically implanted electrodes that can translate neural activity into highly accurate commands. While powerful, those approaches carry significant surgical risks and costs. Non-invasive and wearable options, by contrast, avoid surgery but historically have struggled with signal reliability and speed.
UCLA’s AI-augmented design aims to bridge that gap. By layering intent inference from AI vision onto EEG-driven control, the system can predict likely goals and smooth out noisy brain signals, allowing users to complete tasks more quickly and with greater confidence. The results point to safer, more accessible assistive technologies for people with paralysis or other motor impairments.
“Ultimately, we want to create AI-BCI systems with combined control that will allow people with motor impairments, such as paralysis or ALS, to regain some independence in daily activities,” Kao emphasized.
The bigger picture
The study adds momentum to a broader wave of AI-enabled neurotechnology. In April, a woman who had been unable to speak since a stroke 18 years earlier regained her voice using an experimental BCI paired with artificial intelligence—another sign that AI is accelerating progress across both invasive and non-invasive interfaces.
While more research is needed to validate performance in larger, real-world trials, UCLA’s work underscores a promising path forward: combining non-invasive brain sensing with AI intent understanding to deliver practical, everyday assistance without surgery. If the approach scales, it could reshape how people with motor disabilities interact with computers, mobility devices, and the physical world.
Source: Neuroscience News; UCLA Samueli School of Engineering.
