Can AI read our minds? And should we worry about that?

Melbourne:

Earlier this year, Neuralink implanted a chip inside the brain of American Noland Arbaugh, 29, who is paralyzed from the shoulders down. The chip has allowed Arbaugh to move the mouse pointer on a screen just by imagining it moving.

In May 2023, American researchers also announced a non-invasive way to “decode” the words someone thinks from brain scans in combination with generative AI. A similar project generated headlines about a “mind-reading AI hat.”

Can neural implants and generative AI really “read minds”? Will there come a day when computers can generate accurate, real-time transcripts of our thoughts for anyone to read?

This technology might have some benefits – particularly for advertisers looking for new sources of customer segmentation data – but it would demolish the last bastion of privacy: the isolation of our own minds. However, before we panic, we should stop to ask: what can neural implants and generative AI do? In fact “read minds”?

The brain and the mind

As far as we know, conscious experience arises from the activity of the brain. This means that any conscious mental state must have what philosophers and cognitive scientists call a “neural correlate”: a particular pattern of nerve cells (neurons) firing in the brain.

So, for every conscious mental state you may be in (whether thinking about the Roman Empire or imagining a moving cursor), there is a corresponding pattern of activity in your brain.

So clearly, if a device can track our brain states, it should be able to simply read our minds. Good?

Well, for real-time AI-powered mind reading to be possible, we must be able to identify precise, individual correspondences between particular conscious mental states and brain states. And this may not be possible.

tough matches

To read a mind from brain activity, it is necessary to know precisely which brain states correspond to particular mental states. This means, for example, that it is necessary to distinguish the brain states that correspond to seeing a red rose from those that correspond to smelling a red rose, or touching a red rose, or imagining a red rose, or thinking that red roses are yours. . mother’s favorite.

We must also distinguish all of these brain states from the brain states that correspond to seeing, smelling, touching, imagining, or thinking about something else, like a ripe lemon. And so on, for everything else you can perceive, imagine or have thoughts about.

To say this is difficult would be an understatement.

Take facial perception as an example. The conscious perception of a face involves all types of neural activity.

But much of this activity appears to be related to processes that occur before or after conscious face perception: things like working memory, selective attention, self-monitoring, task planning, and reporting.

Identifying those neural processes that are uniquely and specifically responsible for the conscious perception of a face is a herculean task that current neuroscience is not even close to solving.

Even if this task had been accomplished, neuroscientists would only have found the neural correlates of a certain type of conscious experience: namely, the general experience of a face. In this way they would not have found the neural correlates of experiences of particular faces.

So even if astonishing advances in neuroscience were to occur, the aspiring mind reader wouldn’t necessarily be able to tell from a brain scan whether he was seeing Barack Obama, his mother, or a face he didn’t recognize. .

That wouldn’t be much to write home about, as far as mind reading goes.

But what about AI?

But don’t recent headlines about neural implants and artificial intelligence show that some mental states can be read, such as imagining cursors moving and engaging in internal speech?

Not necessarily. Take the neural implants first.

Neural implants are typically designed to help the patient perform a particular task: moving a cursor on a screen, for example. To do this, it is not necessary that they be able to identify exactly the neural processes that correlate with the intention to move the cursor. They just need to get a rough idea of ​​the neural processes that tend to accompany those intentions, some of which might actually be underpinning other related mental acts, such as task planning, memory, etc.

Therefore, while the success of neural implants is certainly impressive (and future implants are likely to gather more detailed information about brain activity), it does not show that precise one-to-one mappings have been identified between particular mental states and particular brain states. . . And therefore it doesn’t make genuine mind reading any more likely.

Let us now take the “decoding” of inner speech using a system composed of a non-invasive brain scanner plus generative AI, as reported in this study. This system was designed to “decode” the content of continuous narratives from brain scans, when participants listened to podcasts, recited stories in their heads, or watched movies. The system is not very precise, but still, the fact that it worked better than chance in predicting these mental contents is really impressive.

So let’s imagine that the system could predict continuous narratives from brain scans with complete accuracy. Like the neural implant, the system would only be optimized for that task: it would not be effective at tracking any other mental activity.

How much mental activity could this system monitor? That depends: how much of our mental life consists of imagining, perceiving, or otherwise thinking in continuous, well-formed narratives that can be expressed in simple language?

Bit.

Our mental lives are flickering, lightning-fast, multi-stream affairs, involving real-time perceptions, memories, expectations, and imaginings all at once. It’s hard to see how a transcript produced by even the most refined brain scanner, coupled with the most intelligent AI, could capture all that faithfully.

The future of mind reading

In recent years, AI development has shown a tendency to overcome seemingly insurmountable obstacles. Therefore, it is not wise to completely rule out the possibility that artificial intelligence can read minds.

But given the complexity of our mental lives and how little we know about the brain (after all, neuroscience is still in its infancy), confident predictions about AI-powered mind reading should be taken with a grain of salt.

(Authors:Sam Baron, Associate Professor of Philosophy of Science, University of Melbourne and Jenny Judge, Professor of Philosophy of Mind and Cognitive Sciences, University of Melbourne)

(Disclosure Statement:The authors do not work for, consult with, own shares in, or receive funding from any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.)

This article is republished from The Conversation under a Creative Commons license. Read the original article.

(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated channel.)