Brain imaging study finds AI chatbots do not process meaning like the human brain
A study published in eLife, signed by researchers from the University of Melbourne and Hitotsubashi University, compared eight transformer-based language models with 7-tesla functional MRI scans of 30 people reading 108 sentences. The models outperformed a system that ignores word order, but their fit to brain activity was low and lower than that of a model that explicitly represents structural relations between words. The difference concentrated on sentences with inverted semantic roles.
In 40 seconds
- The study used 7-tesla functional MRI to record the brain activity of 30 people while they read 108 sentences.
- The eight language models analyzed outperformed a model that ignores word order, but their overall performance in predicting brain activity was low.
- The models rated sentences with inverted roles as more similar to each other than human participants did.
IASummary generated by artificial intelligence from a story by Infobae. It may contain errors: the original source prevails.




Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.