This patient’s Neuralink implant gets a boost by generative AI.

“I am the third person in the entire world to receive the #Neuralink implant. First with ALS. 1st Nonverbal. I am typing with my brain. “It is my primary form of communication,” he said. “Ask me anything! “Ask Me Anything!” I will answer all verified users!

Smith is a case that is attracting attention because he is not only communicating through a brain implant, but also receiving help from Grok, Musk’s AI chatbot. Grok is suggesting ways Smith can add to conversation and drafted some the replies he wrote to X.

The generative artificial intelligence is speeding up his communication, but it raises questions as to who is actually talking — him or Musk’s software.

There is a tradeoff between accuracy and speed. Eran Klein is a University of Washington neurologist who studies the ethics of implanted brains. He says that the promise of brain-computer interaction is that it can be faster if combined with AI.

Smith, a Mormon father of three children, was diagnosed with ALS when a shoulder injury he suffered in a church Dodgeball game refused to heal. As the disease progressed he lost his ability to move anything but his eyes and was no longer able speak. He decided to use a breathing tube when his lungs stopped working.

He began a “campaign of shameless self promotion” in 2024 to get into Neuralink’s implant study, he told his Arizona local newspaper: “I really desired this.”

Musk himself appeared on Smith’s mobile phone screen the day before surgery to wish him well. According to a recorded call, Musk said: “I hope this will be a game changer both for you and your loved ones.”

Smith responded, “I’m so excited to get it in my head,” typing out a response using a device which tracks his eye movements. This was the technology that he had used to communicate in the past, though slowly.

Smith, who was about to undergo brain surgery, saw Musk’s virtual presence as a sign of a larger transformation. Smith’s brain would soon be integrated into a larger technology and media eco-system, one of whose goals is, according to the billionaire, to achieve a symbiosis between humans and AI.

Take a look at what happened on April 27, when Smith announced on X he had received the implant and wanted questions. One of the earliest came from “Adrian Dittmann,” a Twitter account that is often suspected to be Musk’s alter-ego.

Dittmann: “Congrats! Can you describe the feeling of typing and interacting with technology using the Neuralink?

: Hey Adrian, this is Brad–typing straight from my mind! It’s like I am a cyborg in a sci-fi film, moving a mouse by just thinking about it. It was initially a struggle, my cursor was like a drunken mouse, barely hitting the targets. But after weeks of training, with imagined hand and mouth movements, it clicked. It’s almost like riding a bicycle.”

A user noticed the smooth wording, punctuation, and punctuation. (A long dash is an AI-specific character that is used more often by AIs than by humans.) He asked if the reply had been created by AI.

Smith did not answer on X. In a message sent to MIT Technology Review,he confirmed that he had used Grok to draft his answers after he’d given the chatbot notes he took on his progress. Smith wrote to us, “I asked Grok use that text to provide full answers to the question.” “I am the one who wrote the content but I used AI for drafting.”

In many ways, this exchange on X seems like a surreal cross-marketing example. Smith posted from a Musk implant with the help of an AI on a Musk platform, and in response to a famous Musk enthusiast, if not the “alt”of the richest man in the world. It’s only fair to ask, where does Smith’s ecosystem end and Musk begin?

This is a question that has attracted the attention of neuro-ethicists who believe Smith’s case raises important issues about the possibility that brain implants and AI could merge one day.

It’s amazing that Smith can use his brain to steer a pointer well enough to send texts with his wife and answer our emails. He told us that he did not want to comment too much on the philosophical questions surrounding the authenticity of his AI posts, since he had only been semi-famous a few days. “I don’t wanna wade into over my head,” said he. “I’ll leave that to the experts!”

Smith’s eye tracker, which he used to type previously, only worked indoors and in low light. In a video he uploaded to X, he explained that he was “basically Batman stuck in a dim room.” The implant allows him to type in brighter areas, even outdoors, and much faster.

His brain is wired with thin wires that listen to his neurons. The signals are so faint that they must be amplified, sampled, and filtered to extract the most significant features. These are then sent from his brain via radio to a MacBook and further processed to allow him to move the computer’s pointer.

Smith can type using an app while controlling the pointer. But he is able to express himself more quickly and naturally with the help of various AI technologies. ElevenLabs is a startup that created a copy his voice using recordings he made when he had a healthy voice. The “voice-clone” will read his written words in a manner that sounds like him. The service is used by ALS patients without implants. Researchers

have been studying the feelings of ALS patients about aids such as language assistants. Klein interviewed 51 ALS patients in 2022 and found a wide range of opinions.

Others are very exacting. For example, a librarian felt that everything she said had to be her own words. Some people are more relaxed–an entertainer thought it was important to keep up with the fast-moving conversation.

Smith said in the video he posted online that Neuralink engineers were using language models such as ChatGPT and Grok for a selection relevant answers to questions as well as suggestions of things he could use to say during conversations around him. He gave the following example: “My friend asked for ideas for his horse-loving girlfriend. I chose the option where I told him to buy her a bouquet carrots in my voice. What a funny and creative idea!”

Although these aren’t his actual thoughts, they will do, since clicking once on a menu is faster than typing a complete response, which can take several minutes.

Smith said he wanted to go a step farther. He has an idea for a “personalized” large language model, which “trains on my previous writing and answers using my opinions and style.” Smith told MIT Technology Review he is looking for someone to create it for them: “If anyone wants to help me let me know.”

www.aiobserver.co

More from this stream

Recomended