I know what you're thinking...

AI text tools like ChatGPT are based on the logical order of words or strings of words. The sentence “I saw a dog” completes the AI program with the word “with” or “that”, depending on which words follow: “I saw a dog with three legs” or “I saw a dog that ran away”.
Recent research shows that words activate different areas of the human brain. In a so-called fMRI scanner (an extremely powerful MRI scanner), researchers at the University of Texas can see which area in a brain lights up when you hear, read or even think about a certain word. The AI program can predict with 80% accuracy whether you think of the word “with” or “that” after the sentence “I saw a dog”.
The program was able to retell with reasonable accuracy what story someone had listened to, or what video clip he/she had watched. If you have only seen the computer-generated version of the story, you can correctly answer nine of the sixteen questions about the story heard by the subject.

Snags

Of course there are still many snags to this. If a subject is uncooperative, "mind reading" becomes much more difficult, and an fMRI scanner is an incredibly powerful and large device; if you stand next to it with a bunch of keys in your pocket, it will fly out through your clothes.
But still, developments are moving fast. After all, in addition to language, it is already possible to reconstruct images that a test subject has seen. George Orwell's comment in 1984 that you only have privacy in your own skull seems quickly to be undone. Hence the growing concern of scientists and the call for politicians to intervene.
This is also evident from the resignation of AI pioneer Geoffrey Hinton at Google. This 75-year-old British-Canadian computer scientist is also called the 'godfather of artificial intelligence'. In the 1980s, he pioneered the technology that is now used by AI systems such as ChatGPT, the so-called ‘deep learning’.
Although Hinton emphatically does not explain his departure from Google as criticism, he does advocate AI legislation, such as the AI Act that the European Commission is working on. This should categorize and restrict AI systems.