In 1908, a Columbia University psychiatry professor named Frederick Peterson claimed to have built a “psychometer”—a contraption which he said could read people’s thoughts using, among other components, a mirror and a lamp. “It is almost like sorcery to observe the behavior of the instrument with a series of words, some purposely chosen to be emotionally indifferent and others to strike some emotional complex deep in the soul of the individual experimented upon,” Peterson told the Seattle Star about his device.
Peterson’s psychometer may have been a Ouija board-like illusion, but there are plenty of scientists today who believe that we will soon have access to machines that can record and interpret human thoughts using the pattern-recognizing prowess of AI. In June, Meta unveiled the latest version of its Brain2Qwerty system, designed to translate brain scans into text; the company is promoting it as an early step towards helping people with ALS and other neurodegenerative illnesses regain their ability to communicate.
The San Francisco-based startup Conduit is also working to crack the AI mind-reading problem. Founded in 2024, Conduit is endeavoring to build what it describes on its website as “telepathy at scale”—AI systems that can read minds entirely through noninvasive means. (In other words, without the need for a surgically implanted brain-computer interface, like the one Elon Musk is building at Neuralink.) The company’s long-term goal is to build and sell a wearable device that conveys commands to AI assistants directly via thoughts, as opposed to voice or text prompts—which, as anyone who’s interacted with ChatGPT or Siri knows very well, can often be a very buggy process.
Modern AI systems require mountains of training data to operate, and the technology that Conduit is building is no different. To harvest the data it needs, the startup has been recruiting participants to visit its office and spend two hours conversing with AI while wearing a four-pound, electrode-studded helmet that looks like something House Harkonnen might wear.
The company is offering $50 to $55 to anyone willing to participate in these “freeform” conversations. The only requirement is that you must be able to type without ever needing to look at your fingers. There’s a 32-session limit per person. “Sessions vary,” the company wrote about its data-collection process in a December blog post. “Some are listening and speaking with an LLM, and some are reading and typing. The goal is to maximize the amount that subjects type or say during the two-hour period, without constraining the topics they discuss.”
Conduit was founded by Rio Popper, an Oxford graduate who went blind at age five, and an AI researcher named Clem von Stengel. Naomi Bashkansky, a former alignment researcher at OpenAI, announced in a blog post last week that she was leaving the company to join Conduit.
“I like that this is literally a human-in-the-loop vision of the future, where AI directly empowers humans rather than replacing us,” Bashkansky wrote of her decision to join the startup, before calling the work “frighteningly ambitious.” She expects Conduit, if it achieves its goals, to eventually be valued at more than $1 trillion.
Bashkansy also envisions that by 2030, Conduit’s “neural headband” will have evolved to the point where users can seamlessly communicate commands to AI tools like OpenAI’s Codex using only their thoughts. “Without it, I feel like I’m talking with a superpowered alien who’s trying its best to be helpful but isn’t sure what I want and is scared I’ll get mad if it does something I don’t want,” she writes in the blog post, imagining her future self. “Ugh.” As soon as next year, she believes the Conduit headband could become a vital part of her daily life at work, allowing her to dash off commands to her AI assistant while she casually plans out the day ahead over a cup of coffee.
“It feels like magic,” she writes—a close echo of Dr. Peterson all those years ago.
Read the full article here
