Co-inventor of Apple’s FaceID and Vision Pro technologies, Giddy Litwin, has dedicated the last six years to developing groundbreaking artificial intelligence models. These revolutionary models aim to decode electrical activity in the brain and identify cognitive disorders.
Recently, Litwin’s startup, Hemispheric, secured $52 million in funding after analyzing brain data from over 100,000 individuals. This innovative approach enables the deep learning model to assess brain health without invasive procedures.
After leaving Apple in 2020 in pursuit of new opportunities, Litwin was approached by Hemispheric co-founder Hagai Lalazar via a LinkedIn message. Lalazar’s vision was to leverage artificial intelligence for brain study without surgical intervention, and he sought a commercially savvy co-founder to propel the company forward. After interviewing around 75 candidates, he found Litwin.
Having played a key role in developing FaceID, Litwin was also engaged in hand tracking for Vision Pro, an augmented reality device. He mentioned to WIRED that extensive data collection akin to “hundreds of thousands of subjects” was vital for training deep learning models behind such technologies.
“A massive data collection initiative was essential for our projects, which we successfully established at Hemispheric,” Litwin explained.
Traditionally, diagnosing conditions like depression, Alzheimer’s, and Parkinson’s relied heavily on subjective questionnaires and behavioral assessments. To address this, Litwin and Lalazar amassed their “most valuable asset”: 250,000 hours of brain data from 100,000 compensated volunteers in cities like Tel Aviv and Boston, as well as across Asia. Participants engaged in activities resembling games that activated various brain regions.
This extensive data set trained a pioneering model that deduces brain function from electrical activity in the skull, similar to how large-scale language models analyze text to extract meaning. They tested this generalized model on a select group, including individuals with PTSD, schizophrenia, and depression, successfully estimating their brain health. The research team is currently advancing clinical studies to ascertain whether the model can diagnose and predict Alzheimer’s disease.
The team aims to file their first product aimed at studying PTSD for FDA approval early next year, with hopes of making it publicly accessible by late 2027.
To assist in diagnosing cognitive impairment, patients will wear a lightweight EEG headset for about 15 minutes while interacting with a tablet app. Hemispheric’s AI models are designed to help clinicians interpret signals, aiding in diagnostics, predicting treatment, and monitoring patient progress.
“We envision a future where this process resembles a blood test,” said Lalazar. “This device will be affordable and could be distributed in mental health clinics, hospitals, and even psychologists’ offices.”
AI-powered diagnostic tools for conditions like lung cancer are already being utilized in clinical settings across Europe, expediting treatment access. In addition, major AI players like OpenAI and Anthropic are venturing into healthcare, escalating competition among emerging startups in the field.
Hemispheric has attracted early-stage investments from venture capital firms and private investors in the U.S. and Israel, including early Uber supporter Howard Morgan. The funding will facilitate partnerships with governments, medical institutions, and pharmaceutical companies, create jobs in the U.S., and progress toward regulatory approval. They also aim to gather extensive brain data from millions to enhance their models.
The co-founders are developing their own brain scanner believed to yield more valuable data than conventional EEG methods. “These devices were never designed for machine learning; they were never built for deep learning,” Litwin stated.
Source: www.wired.com