I was fascinated by the news story about the study regarding the ability of Artificial Intelligence (AI) to “read minds.” Different stories told slightly different versions, meaning they either did or did not include the authors’ caveats about the limitations of AI. Recently there has been a spate of news items warning about the dangers of AI taking over mankind.
Not to diminish the strengths of AI, the full article published in Nature Neuroscience reveal critically important facts about the study:
- Subject cooperation is essential for AI to train and apply the decoder which “reads” your mind
- You have to climb into a big MRI to enable the AI to even get started
- The subject can resist the AI by silently repeating simple tasks such as counting by sevens, naming and imagining animals, and imagined speech
The authors of the study caution that even if the subject doesn’t cooperate and the AI is inaccurate, humans could still deliberately lie about the results for “malicious purposes.” Nothing new under the sun there.
The current technology here would not be usable in the emergency room to assist psychiatrists ascertain suicide risk. It probably wouldn’t help psychiatrists and other physicians diagnose Factitious Disorder in patients whose main feature is “lying” about their medical and psychiatric disorders in order to get attention from health care professionals.
This reminds me of news stories about the propensity of AI to tell lies. One story called them pathological liars. I interviewed Google Bard and found out that it makes stuff up (see my posts about Bard). Does that mean that it’s lying? Humans lie, but I thought machines were incapable of deception.
Another interesting sidelight on lying is whether or not you could use AI like a lie detector. For example, the case of people who report being abducted by extraterrestrials. Travis Walton and co-workers reported he was abducted in 1975 and they all took lie detector tests. They all “passed.” There are many articles on the internet which essentially teach how to beat the polygraph test.
And if you can beat the AI by repeating the names of animals, it will not detect lying any better than a polygraph test.
I think it’s too soon to say that AI can read your mind. But it’s clear that humans lie. And it wouldn’t hurt those who are enthusiastically promoting AI to brush up on ethics.
Tang, J., LeBel, A., Jain, S. et al. Semantic reconstruction of continuous language from non-invasive brain recordings. Nat Neurosci (2023). https://doi.org/10.1038/s41593-023-01304-9
“A brain–computer interface that decodes continuous language from non-invasive recordings would have many scientific and practical applications. Currently, however, non-invasive language decoders can only identify stimuli from among a small set of words or phrases. Here we introduce a non-invasive decoder that reconstructs continuous language from cortical semantic representations recorded using functional magnetic resonance imaging (fMRI). Given novel brain recordings, this decoder generates intelligible word sequences that recover the meaning of perceived speech, imagined speech and even silent videos, demonstrating that a single decoder can be applied to a range of tasks. We tested the decoder across cortex and found that continuous language can be separately decoded from multiple regions. As brain–computer interfaces should respect mental privacy, we tested whether successful decoding requires subject cooperation and found that subject cooperation is required both to train and to apply the decoder. Our findings demonstrate the viability of non-invasive language brain–computer interfaces.”