In a significant breakthrough in the field of artificial intelligence (AI), researchers have developed a tool that can interpret thoughts and translate them into text. This technology, which is still in its prototype stage, boasts a conversion reliability of over 60%.
The Technology Behind Thought Translation
The thought translation tool, aka BrainGPT is based on two distinct AI models:
- First Model: This model analyzes the electrical signals generated by the brain during various activities such as thinking, moving, or sleeping.
- Second Model: This model then translates these signals into text.
The Role of Neuralink
Neurotech startups like Neuralink are already utilizing these cutting-edge technologies to aid paralyzed individuals in communicating or controlling electronic devices with their thoughts. However, the Neuralink device requires surgical implantation of electrodes into the brain tissue to read the signals accurately, posing certain risks.
A Non-Invasive Alternative
Researchers at the University of Technology Sydney’s department of human-centered artificial intelligence, GrapheneX-UTS, have developed a non-invasive system that eliminates the need for surgery. This device, which translates thoughts into text, was presented at the NeurIPS conference on advances in artificial intelligence in Louisiana.
How It Works
To develop this device, an experiment was conducted with 29 participants who were asked to read passages of text while wearing special equipment that recorded their brain’s electrical activity. This activity was measured using an electroencephalogram (EEG), a device that records electrical signals produced by the brain via sensors placed on the scalp.
The signals recorded by the EEG were then fed into an AI model called DeWave, which was trained to associate each signal with specific words or phrases. After DeWave learned to understand these brain signals, it was connected to the open large tongue model (LLM). LLM not only generates individual words based on brain signals but also constructs complete sentences by considering context and linguistic structure.
Accuracy and Future Applications
Initially, the system’s accuracy was about 40%. However, ongoing research has improved this figure to over 60%. While further reliability is needed for practical application, researchers hope that their system will eventually help individuals who have lost their speech, including stroke victims, to communicate again. The technology could also find applications in robotics, offering a more natural and intuitive interface that bridges the gap between humans and machines.
In summary, this technology holds great promise for the future, potentially revolutionizing the way we communicate and interact with machines. As research progresses, we can expect to see even more exciting developments in this field.