New Cap Utilizes Artificial Intelligence to Interpret Thoughts and Transform them into Written Text.

Researchers at the University of Technology Sydney (UTS) have created a cap with artificial intelligence (AI) that “reads” the user’s thoughts using electroencephalography, or EEG. Even while the system is far from flawless, it is impressive—if not a little unsettling—that it can roughly translate internal mental processes into text and speech without the need for extra inputs like eye tracking.

The system, named DeWave, is made up of two main parts: a large language model (LLM) and an EEG-reading cap. The non-invasive EEG electrodes included in the cap start to detect the wearer’s mental activity as soon as they put it on since electrical impulses from the brain run through them. These brain waves are fed into the DeWave LLM, which the UTS group developed based on the ZuCo dataset, which associates natural reading activities with eye-tracking and EEG patterns. The user reads the statement aloud while matching each pattern with a word or phrase.

DeWave obtained a highest accuracy score of 42.8% on the BLEU-1 machine translation metric in testing with 29 individuals. The researchers observe that DeWave matches verbs more accurately than nouns, with the latter frequently translating into synonymous pairings, in a preprint study uploaded to the arXiv. (For instance, although DeWave uses the word “version,” the user could think of “edition.”)

They state, “Our analysis points to two possible causes for this.” First, semantically related words may cause comparable brain wave patterns during brain processing.Secondly, compared to traditional language translation, a far smaller volume of EEG-to-text pairs is accessible for training. Therefore, it is reasonable to expect some degree of mistake when translating unknown nouns or sentences.

DeWave should ideally score 90% or above for accuracy. However, in terms of usefulness, EEG-only “mind-reading” technologies are already thought to be superior to comparable techniques like eye tracking or Elon Musk’s Neuralink. A system like DeWave doesn’t actually require training when it comes to teaching stroke survivors or patients with amyotrophic lateral sclerosis (ALS) how to utilize an eye-tracking interface; all you have to do is read or think, and DeWave takes care of the rest. It’s also far simpler to swallow donning a swimming cap with EEG electrodes than having an implant physically inserted into your brain, particularly if that implant was created by a wild millionaire with a questionable sense of ethics.

Exit mobile version