‘Mind reading’ was the science phrase of today, with the publication of a mechanism to reconstruct speech from brain waves. It is unfortunate that so many headlines sensationalised the findings, since this science fiction façade conceals some genuinely remarkable neuroscience research.
Researchers at the University of California, Berkeley, measured brain activity in a region called the superior temporal gyrus – one area that processes auditory information – while subjects listened to recorded speech. Subjects were undergoing neurosurgery for epilepsy or tumours, allowing electrodes to record directly from the surface of the brain.
The measured brain activity was then used to formulate a mathematical model that would allow decoding of signals in this region to reconstruct speech. The models they used took into account information about the rhythm, pitch and patterns of speech. Once refined, these models were used to identify words that the subjects were thinking from just subjects’ brain activity that was stimulated by hearing those words.
At first glance, this does certainly appear to be mind-reading, but there are two caveats that were largely excluded from news reports. Firstly, imagining that this mechanism could be used for clear-cut, every-day mind-reading requires an assumption that the same area of the brain is activated in the same way when formulating speech rather than listening to it.
This difficulty was noted in the press release by the first author of the paper, Brian N Pasley: “This research is based on sounds a person actually hears, but to use it for reconstructing imagined conversations, these principles would have to apply to someone’s internal verbalisations.”
Pasley continues to point out that there is “some evidence that hearing the sound and imagining the sound activate similar areas of the brain.” This suggests that such an assumption is valid, but nonetheless introduces the need for further clarification before any ‘mind-reading machines’ are in production.
The second stumbling block is a greater hurdle for general purpose mind-reading. When words were identified from brain waves alone, they were words from a defined set of 47. This is radically different from identifying words generated at will from a choice of hundreds of thousands.
In the paper itself, the authors note that word “identification is more difficult when the word set contains many acoustically similar sounds”. Perhaps the best the current system could do with natural language would therefore be to produce a list of words that may have been the imagined one.
These shortcomings are not to take anything away from the achievements of this study. At a fundamental level, such detailed analysis of neural activity tells us an awful lot about how the brain works and processes sensory information. Similar research on the visual system has allowed some manipulation of visual perception through brain stimulation, and comparisons of these studies may give us more insight into how we process the world around us.
On a practical level, a refined version of this system could be used to help people who are unable to speak but can think of what they want to say, such as people with locked-in syndrome. Using a system finely tuned to individuals and with a limited vocabulary may allow some communication, and further research could undoubtedly improve this.
These outcomes and prospects are incredible, but at the very least, they have been placed for many in a context that damages their credibility. Mind reading as the general public thinks of it is a long way off, if it is at all possible or desirable. Magnificent neuroscience that unravels the puzzle of the brain and has extraordinary medical potential is happening right now.
Comment with Facebook