The artificial intelligence industry continues to push the boundaries of what is possible, and the latest move in this race is the transition of a former key OpenAI researcher to the ambitious Conduit project. Naomi Bashkansky, known for her work in AI safety, now serves as one of the co-founders of the startup's research division, which sets itself an almost futuristic task: to train models in "non-invasive reading of human thoughts."

Vision for 2035: Telepathy as an Interface

In her programmatic manifesto, Bashkansky outlines a radical paradigm shift. According to her forecasts, by 2027 we will be able to transmit thoughts directly to an AI agent via Bluetooth, which will interpret them into specific commands and actions. This is not just an improvement of voice assistants—it is a fundamental transformation of human-machine interaction.

"These will not be super-smart automated interns that you desperately try to keep up with. This will be a natural, joyful extension of yourself," the researcher emphasizes, formulating the project's philosophy. She insists: the goal is not to replace humans, but to augment them.

Technology and Criticism

Technically, Conduit plans to train models based on brain activity, predicting the semantic outcome of a person's actions—for example, the text they are typing. "Input is brain activity, output is what the person was doing at that moment," Bashkansky explains.

However, the San Francisco-based company has already faced skepticism. Critics rightly point to the complexity of implementing a "non-invasive" approach without direct surgical access to neurons. This is a key challenge: current electroencephalography (EEG) methods provide too coarse a signal for precise semantic decoding. For comparison, a recent Meta system that reconstructs text from brain activity achieved only 61% accuracy at the word level—versus 8% for previous methods, but this is still far from full-fledged "telepathy."

My Analysis

Bashkansky's statement about "converging brain and AI" echoes the forecasts of Elon Musk and Sam Altman, but it is important to maintain a sober perspective here. Even if Conduit manages to create a working prototype by 2027, mass adoption will require not only a breakthrough in neuroscience but also addressing ethical questions of thought privacy. The market clearly underestimates the complexity of this task, but it is precisely such "crazy" projects that often drive the industry forward. I will follow Conduit's development with great interest—this is potentially the most ambitious AI project of the decade.