Brain computing is the science of directly connecting the brain’s electrical signals to a computer. Technologists and investors have high hopes that advances in artificial intelligence, particularly neural networks, will revolutionize the field of brain computing, which is currently experiencing a surge in funding and popularity thanks in part to Elon Musk’s Neuralink startup. As my Eye on A.I. colleague Jeremy Kahn and I reported for a recent Fortune feature, Neuralink has been plagued by a pressure-cooker work culture, impossible deadlines, and a missing leader. Despite the problems, experts are keeping a close eye on Neuralink, which hopes to conduct human trials this year pending approval by the Food and Drug Administration. 

Musk has publicly said that brain implants could save humanity from an A.I. apocalypse, a statement that many brain computing experts regard as science fiction but one that seems in character for the Tesla-founder, given his taste for hyperbolic, attention-getting comments. A more realistic near-term future regarding brain implants and A.I. is that neural networks could help researchers better understand the human brain, which could in turn lead to more advanced medical breakthroughs.

For several years, researchers have been able to surgically implant tiny electrodes into the brains of paraplegics and use the implants to help patients regain limited body movements and to command computers to execute simple tasks such as moving robotic arms. Many experts believe the next generation of A.I.-powered brain implants could enable users to performmore advanced tasks, such as helping patients power and steer wheelchairs by thought alone.

The underlying premise is that the brain is essentially a “data organ” that processes information and in turn takes actions, such as sending signals to muscles to command them to move, explained Matt Angle, the CEO of brain computing interface startup Paradromics.

Last year, neuroscientists at the University of California, San Francisco published a high-profile paper in The New England Journal of Medicine detailing how artificial neural networks coupled with a brain implant helped a paralyzed and voiceless man regain the ability to talk via a computer. With the help of a surgically implanted brain interface, the researchers were able to record the patient’s brain activity while he attempted to say 50 words during multiple sessions totaling 22 hours. They then trained a neural network on the patient’s brain data so that the software eventually learned to associate certain brain activity with each of the 50 words he would repeat during his sessions. As a result, the A.I. could help decode the patient’s brain activity when he attempted to talk, enabling it to display short sentences like “I am very good” on a computer screen when the researchers asked certain questions.

The UCSF paper helped draw attention to A.I.’s potential usefulness for the brain computing industry. If A.I. could help a paralyzed, mute man communicate with 50 words merely by thinking, perhaps the technology will improve so that he could one day regain his entire vocabulary. However, there’s still much more research needed for that breakthrough to occur.  

A.I. specialist Robert Edgington, a co-founder of the brain computing startup Braingrade, said that the brain computer industry still requires much more data to make A.I. really useful in the field. Tech giants like Google-parent Alphabet and Facebook-parent Meta were able to create powerful computer vision systems because they had access to a massive amount of photos of cats from the Internet that they could use for training their neural networks, Edgington explained. At this point in time, the brain computer space lacks its own enormous trove of valuable data, he said.

Braingrade aims to eventually build a BCI to be implanted deep into the brain near the hippocampus, which plays a major role in memory function—and it’s currently building the data infrastructure its leaders believe will be necessary for its devices. Edgington said he spends much of his time building the proper data tools and so-called pipelines needed to parse and prep brain data so that it can eventually be analyzed by A.I. These kinds of data infrastructure tools may be commonplace for Internet web companies, but they don’t currently exist in the brain computing industry. These data tools, Edgington believes, will lay the foundation for the heavy duty A.I. analysis that the company hopes to accomplish. 

In the meantime, expect more breakthroughs in deep learning and brain implants to occur at academic labs like UCSF. The research stemming from these academic labs gives hints of what’s yet to come, inspiring the startups to develop future brain implant technology that executives like Paradromics’ Angle hope will do tasks as grand as restoring vision to the blind. It will likely be artificial intelligence that powers all those future devices for the human brain.

Jonathan Vanian 
@JonathanVanian
jonathan.vanian@fortune.com

A.I. IN THE NEWS

Top Chef star bets on A.I.  Tom Colicchio, celebrity chef and star of the competitive cooking show “Top Chef,” told Fortune that he invested in the restaurant technology startup Tattle. Restaurants use Tattle to gather customer feedback that they can use to improve the quality of their food or dining experience. The startup plans to use its latest funding of $5.5 million to build an A.I. and machine learning team to develop sales forecasting models for restaurants based on customer feedback.

Let’s keep this a secret for now. The California Superior Court ruled that Waymo could keep certain details secret related to its self-driving car development, TechCrunch reported. Waymo originally filed a lawsuit against the California Department of Motor Vehicles arguing that disclosing certain information would expose trade secrets. From the article: Waymo filed the lawsuit because it wants to protect details about how its AVs identify and navigate through certain conditions, how they determine the circumstances under which the AV will revert control to a human driver, when to provide support to an AV fleet, and how the company addresses disengagement incidents and collision incidents, according to the lawsuit.

Meta’s big A.I. plans. Facebook-parent Meta held an online event on Wednesday in which executives detailed recent A.I. developments, such as an A.I. project that CEO Mark Zuckerberg described as a universal language translator, the tech news service Protocol reported.“The ability to communicate with anyone in any language — that’s a superpower people have dreamed of forever, and AI is going to deliver that in our lifetimes,” Zuckerberg said.