Scientists At The University Of Texas Develop An AI System That Can Translate Human Thoughts Into Words

Researchers from the University of Texas at Austin have developed an artificial intelligence (AI) system capable of interpreting and reconstructing the human mind. (photo:doc. pexels)

JAKARTA - Researchers from the University of Texas at Austin have developed an artificial intelligence (AI) system capable of interpreting and reconstructing the human mind. They recently published a paper in the journal Nature Neuroscience exploring the use of AI to translate the human mind non-invasively into words in real time.

According to the researchers, the current method of encoding the mind into words is whether or not it is invasive or limited because it "can only identify the stimulus of a small number of words or phrases". The team in Austin managed to overcome this limitation by training neural networks to decode the signal of functional magnetic resonance (fMRI) from several areas of the human brain simultaneously.

In carrying out these experiments, the researchers tested several subjects by listening to podcasts for hours while the fMRI machines recorded their brain activity non-invasively. The resulting data is then used to train the system on a specific user's mindset.

After training, test subjects have their brain activity monitored again while listening to podcasts, watching short films, and imagining telling stories secretly. During this part of the experiment, the AI system was given fMRI subject data and decoded the signal into a language that is easy to understand in real-time.

According to a press release from the University of Texas at Austin, AI is capable of getting things right about 50% of the time. However, the results are not eccentric - researchers designed AI to convey the general ideas they were thinking about, not the right words they were thinking about.

The good news for anyone worried about their thoughts being infiltrated by AI without their permission, scientists is very clear that this is not the current possibility. The system only works if it is trained on certain user brain waves. This makes it useless to scan individuals who haven't spent time providing fMRI data. And even if the data is generated without user permission, the team ultimately concludes that both decoding data and machine capabilities to monitor the mind in real-time require active participation from scanned people.

However, the researchers confirmed that this is unlikely to happen at this time. This system can only function if trained with a certain user's mindset. Therefore, this system is not useful for scanning individuals who do not provide fMRI data for hours. Even if the data is obtained without user permission, the team finally concluded that both data decryption and machine capabilities to monitor the mind in real time require active participation from people being scanned.

However, the researchers also note that this may not always happen: "[A] our privacy analysis shows that current subjects' cooperation is needed both for training and using decoders. However, future development may allow decoders to bypass these requirements. Moreover, even if the decoder's prediction is inaccurate without the subject's cooperation, they can be misinterpreted intentionally for malicious purposes."

In related news, a team of researchers in Saudi Arabia recently developed a method to improve accuracy in diagnosing brain tumors by processing MRI scanning through blockchain-based neural networks.

In their paper, Saudi researchers showed how to process cancer research on secure and decentralized blockchains could increase accuracy and reduce human error.

Although both experiments were cited as working on their research papers, it is essential to note that the technologies used in each of those experiments have become widely available.

AI, which is the basis for experiments conducted by teams at the University of Texas at Austin, is a generating pre-trained transformer (GPT), the same technology used by ChatGPT, Bard, and similar large language models.

Meanwhile, cancer research conducted by the Saudi Arabian team was carried out using AI trained at Nvidia GTX 1080, a GPU that has been available since 2016.

Realistically, nothing stops the smart developer (with access to the fMRI machine) from combining the two ideas to develop an AI system that can read someone's mind and record it on a blockchain.

This can lead to a "proof-of-throught" paradigm, where people may be able to print nonfungible tokens (NFTs) from their minds or record huge books that cannot be changed from their feelings and ideas for the future, legal goals, or just exhibitions.

The impact of printing NFT thoughts into blockchain could have implications for copyright and patent applications where blockchain serves as proof of exactly when a mind or idea is recorded. It can also allow famous thinkers such as contemporary Nobel winners or philosophers to encode their ideas in irreversible records - which can be co-modified and made into collectible digital assets.