💻 This AI transcribes your thoughts into images with disconcerting reliability (examples)

2024-08-21 04:00:07

A major breakthrough in understanding visual thoughts has recently been made by Dutch researchers. Thanks to sophisticated artificial intelligence (AI), it is now possible to reconstruct mental images based solely on theactivity brain of individuals.

Image d’illustration Pixabay

Scientists at Radboud University in the Netherlands have developed an AI that can recreate images seen by people by analyzing their brain activity. The results of their study, pre-published in bioRxivreveal remarkable precision in the reconstruction of mental images. This result is based on data obtained from two separate studies.

In the first study, volunteers underwent magnetic resonance imaging functional (fMRI) while they looked at photographs. The second study involved a macaque whose brain activity was recorded using implanted electrodes while he observed AI-generated images. The data from these experiments was used to train the AI ​​to reconstruct the original images.

The AI ​​used integrates predictive attention mechanisms (PAM) to target the most relevant brain signals. By processing these signals with a generative adversarial network, the AI ​​manages to create images that are visually close to the originals. The best results were obtained with the images seen by the macaque, whose brain activity, recorded directly by the electrodes, is more precise than fMRI data, which is often disrupted by noise.

Umut Güçlü, one of the lead researchers, points out that these reconstructions are the most accurate ever made to date. He explains that the AI, by interpreting brain signals, learns to focus its attention on the essential aspects to reconstruct the images seen. The ability to faithfully recreate what an individual perceives could pave the way for more advanced brain implants, allowing the restoration of the vue in a more natural way.

Results of the reconstruction.
The top block shows ten examples from the B2G dataset (GAN-generated stimuli), and the bottom block shows ten examples from the GOD dataset (natural stimuli). The first rows show the original stimuli, the middle rows the reconstructions by the PAM model (P), and the last rows the reconstructions by the linear decoder (L).

The potential applications of this technology are vast. In addition to restoring vision, this advancement could transform communication for people with visual impairments and improve neural interfaces. Rapid advances in generative modeling could further improve the accuracy of visual reconstructions in the near future, the researchers say.

1724218292
#transcribes #thoughts #images #disconcerting #reliability #examples