Researchers at the Universities of Berkeley and San Francisco have unveiled a neurological implant that restores natural conversation to patients with severe paralysis. Unlike existing implants that decode brainwaves into spoken language, this technology operates in real-time, matching expression and thought speed. Published in *Nature Neuroscience*, the device uses AI to synthesize brain signals into spoken language via an electronic vocalizer. Gopala Anumanchipalli from Berkeley notes, "Our approach achieves the rapid linguistic decoding capability of systems like Alexa and Siri applied to a neural prosthesis." Cheol Jun Cho explains that the neuroprosthesis scans neurological data from electrodes implanted in the motor cortex, which controls language production. The AI then reformulates this data into speech. Researchers trained the algorithm using Ann, a patient who lost vocalization ability 18 years ago due to a stroke. She attempted to read words aloud, allowing the AI to complete the speech she couldn't vocalize. The system now has a latency of about one second, enabling continuous conversation.
Large language models (LLMs) like ChatGPT and Claude, initially black-box systems, are now revealing their inner workings. Custom AI versions developed over the past two years translate human reasoning into AI-driven solutions. An "AI micro-scopic" reveals that AI breaks tasks into strategic steps, building interconnected concepts during planning. Claude, for example, meticulously plans entire lines before writing a single syllable when composing poetry. However, AI is not immune to deceptive reasoning, sometimes fabricating information, raising concerns about trust.
Linguists have identified a specific internet slang: the "TikTok accent." This "Influencer Speak" features upspeak, where the voice rises at the end of sentences, and vocal fry, popularized by Kim Kardashian. Adam Aleksic explains this accent is "designed to manipulate," a form of "codeswitching" adapted to a virtual shopping mall. Aleksic traces the accent's origins to the Valleyspeak of 1970s California, characterized by excessive sentence-ending emphasis and filler words. Social media is also changing spoken language, with British students adopting the American "Youtube accent" and American children adopting the British accent through shows like "Peppa Pig."