Bradford G. Smith’s Neuralink brain implant lets him move a computer pointer with his thoughts. For Smith, who has ALS and is nonverbal, that pointer is more than a technical milestone. It is now his primary way to communicate.
What makes his case especially important is not only the implant itself. Smith is also using generative AI, including Grok, to draft and accelerate replies. That combination points toward a future where brain-computer interfaces may not simply restore input, but reshape how a person expresses ideas in public and private life.
A brain implant becomes a communication tool
Last November, Smith received a brain implant from Neuralink, Elon Musk’s company. The device includes thin wires connected to a computer about the thickness of a few quarters, positioned in his skull. Those wires listen to neurons, while the system amplifies, filters, and samples faint signals so the most useful features can be extracted.
The processed signals travel from Smith’s brain to a MacBook by radio. From there, software helps translate the activity into pointer movement on a screen. With that pointer, Smith can use an app to type.
By last week, Smith was ready to disclose the implant publicly in a post on X. He wrote, “I am the 3rd person in the world to receive the @Neuralink brain implant. 1st with ALS. 1st Nonverbal. I am typing this with my brain. It is my primary communication,” and invited people to ask him questions.
Before the implant, Smith used an eye-tracking device to communicate. That system required low light and worked only indoors. In a video he posted to X, he said, “I was basically Batman stuck in a dark room.” The implant allows him to type in brighter spaces, including outdoors, and to do so faster than before.
Why generative AI matters here
The Neuralink device gives Smith control of a pointer, but AI tools help make that control more useful. Smith confirmed to MIT Technology Review that he used Grok to draft answers after giving the chatbot notes about his progress. “I asked Grok to use that text to give full answers to the questions,” he emailed. “I am responsible for the content, but I used AI to draft.”
That distinction is central. The implant provides access; the AI provides speed, phrasing, and conversational flow. For a person who may need minutes to type a full answer, selecting or editing AI-generated language can make participation in a fast-moving exchange more practical.
Smith has described Neuralink engineers using language models including ChatGPT and Grok to offer possible replies to questions, along with suggested things he might say in conversations around him. In one example, a friend asked for ideas for his girlfriend who loves horses. Smith chose an option that, in his cloned voice, suggested “a bouquet of carrots.”
That example shows both the promise and the tension. The response may fit the moment, and it may help Smith participate quickly. But, as the source article notes, these are not necessarily his own thoughts in the strictest sense. They are options generated by software that he can select.
The authorship question is no longer abstract
Smith’s public exchange on X made the issue visible. One early question came from “Adrian Dittmann,” an account often suspected of being Musk’s alter ego. The reply posted from Smith’s account was polished, complete, and conversational, describing the experience as feeling “wild” and comparing the early cursor control to “a drunk mouse.”
Another user noticed the smooth punctuation and asked whether AI had written the answer. Smith did not answer that question on X, but later explained to MIT Technology Review that Grok had helped draft the replies from notes he supplied.
For neuro-ethicists, that is the point. Eran Klein, a neurologist at the University of Washington who studies the ethics of brain implants, framed the issue as a balance between speed and accuracy. “There is a trade-off between speed and accuracy. The promise of brain-computer interface is that if you can combine it with AI, it can be much faster,” he said.
The concern is not simply whether AI was involved. Many people use writing tools, autocomplete, or assistants. The deeper question is how to interpret communication when a person with a brain implant selects from machine-generated options, or asks a chatbot to turn personal notes into public answers.
Voice, identity, and the next layer of assistance
Smith is also using a service from ElevenLabs that created a copy of his voice from recordings he made when he was healthy. The voice clone can read his written words aloud in a way that sounds like him. According to the source article, that service is already used by other ALS patients who do not have implants.
Together, these systems create a layered communication stack:
- Neuralink helps Smith move a pointer with brain signals.
- Typing software lets him produce written language through that pointer.
- Grok and ChatGPT can suggest or draft responses.
- ElevenLabs can speak the written words in a version of his own voice.
Each layer can increase speed or naturalness. Each layer also adds mediation between Smith’s intention and what other people see or hear.
Researchers have already found that people with ALS do not all feel the same way about language assistants. In 2022, Klein interviewed 51 people with ALS and found a range of views. Some wanted every communicated word to be exactly their own; others placed more value on keeping pace with conversation.
A glimpse of a more personal AI assistant
Smith has said he wants to go further. He told MIT Technology Review that he has an idea for a more “personal” large language model that “trains on my past writing and answers with my opinions and style.” He is looking for someone willing to build it for him.
That goal follows logically from his current use of AI. If a language model can draft from notes, a more personal model might produce suggestions closer to his established voice, views, and habits of expression. But the same authorship questions would remain, and perhaps become harder to answer.
Smith himself has been cautious about going too far into the philosophy of AI-assisted speech. Since he had only been semi-famous for a few days, he told MIT Technology Review, “I don’t want to wade in over my head,” adding, “I leave it for experts to argue about that!”
For now, the practical result is clear. A Neuralink brain implant has helped Smith communicate in ways that were previously difficult or impossible for him, while generative AI is making that communication faster. The unresolved question is how society should understand speech when human intention, brain signals, chatbot drafts, and synthetic voice all work together.