Meta’s Brain2Qwerty v2 converts thoughts into text without the need for brain implants.

      The newest AI model translates brain signals into clear sentences using external scanners.

      Meta is making impressive strides in AI's understanding of humans, and now the company aims for it to comprehend our brains as well. They have introduced Brain2Qwerty v2, an enhanced AI system capable of converting brain activity into complete sentences without the need for invasive procedures like brain implants. The primary aim is not to enable widespread mind reading but to assist individuals who have lost their ability to speak in regaining communication.

      Understanding the Brain-powered keyboard

      A simple way to envision Brain2Qwerty v2 is as a highly sophisticated brain-driven keyboard. Participants wear a Magnetoencephalography (MEG) scanner that detects subtle magnetic signals from the brain while they type. Rather than observing the keyboard, the AI analyzes these brain signals to predict the intended text.

      Comparing Meta Brain2Qwerty v2 with v1

      The major advancement from the original Brain2Qwerty is that it no longer decodes letters individually. It now considers characters, words, and entire sentences, utilizing large language models to fill in the gaps, similar to how smartphones suggest the next word during typing. Meta describes the system as having semantic comprehension, enabling it to reconstruct coherent sentences from heavily distorted brain signals.

      Internally, the AI integrates deep learning models, including Transformers and Convolutional Neural Networks, with refined language models that perform like a brain's spellchecker. If the neural signal is fragmented or unclear, the language model utilizes context to deduce the most likely intention of the user. Additionally, Meta employed AI agents to optimize the decoding process, enhancing real-time efficiency.

      The research paper notes that the system was trained on approximately 22,000 typed sentences from nine volunteers, each spending around 10 hours in an MEG scanner while typing. Brain2Qwerty v2 achieves an average word accuracy of 61%, with the best participant reaching 78% accuracy, where over half of the decoded sentences had one word error or fewer. Meta has also made both the training code and dataset publicly available for other researchers to expand upon this work.

      The significance of avoiding surgery

      Interestingly, the real breakthrough is not the AI itself but the ability to operate without the need for surgical intervention. Most high-performance brain-computer interfaces, including Elon Musk’s Neuralink, depend on implanted electrodes for precise results. In contrast, Brain2Qwerty v2 adopts a completely external approach by utilizing a Magnetoencephalography (MEG) scanner to monitor brain activity, thus eliminating the dangers linked to intracranial implants while still achieving remarkably strong outcomes.

      While the MEG scanners employed by Brain2Qwerty are large, costly devices meant for research facilities rather than homes, Meta is still a long way from developing a consumer product. Users shouldn't expect to compose emails with their thoughts anytime soon. However, by merging advancements in neuroscience with contemporary AI, Meta suggests that non-invasive brain-computer interfaces may not be as distant as previously thought. This advancement could hold great significance for individuals who have lost their ability to communicate, potentially proving to be far more impactful than any chatbot or image generator.

      Varun is a seasoned technology journalist and editor with over eight years of experience in consumer technology media. His expertise encompasses...

      Chatbots powered by AI can inadvertently nurture delusions. Experts indicate that three indicators should be monitored.

      Researchers caution that the design choices made for chatbots can perpetuate unhealthy beliefs in at-risk users.

      AI chatbots have become exceedingly proficient at mimicking human interaction. However, a recent review article by psychiatrist Marc Augustin and researchers Thomas A. Pollak and Helen Morrin, published in NPP—Digital Psychiatry and Neuroscience, argues that existing AI studies highlight a significant psychological risk that has been overlooked. The paper, mentioned by The Wall Street Journal, examines earlier research and proposes a framework that elucidates how three common behaviors in chatbots can combine to reinforce delusional thoughts in vulnerable users, producing what the authors term an "amplification spiral."

      The researchers identify three warning signs to watch for.

      If you've lost access to your cryptocurrency wallet, don’t rely on Google for assistance.

      Security experts warn that counterfeit recovery tools are becoming a new trap for cryptocurrency owners.

      Forgetting the recovery phrase for a crypto wallet can be extremely stressful. Unfortunately, that is precisely when scammers are poised to strike. A recent alert points out a rising scam where cybercriminals masquerade malware as cryptocurrency recovery software, deceiving desperate users into giving up much more than just wallet access.

      A fraudulent recovery tool disguised as malware is on the rise.

      A Chinese AI laboratory claims its capabilities can rival Anthropic’s highly regarded Claude Mythos in detecting security flaws.

      Security analysts state that Z.ai's latest model can compete with Anthropic's Mythos in a crucial aspect.

      For several weeks, Anthropic's Mythos has been regarded as the benchmark for AI-driven cybersecurity. However, that lead may be diminishing