When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
It’s now possible to use a “brain scanner” to hear the words someone is thinking. Whether that’s exciting or terrifying depends on how you imagine that technology is being used.
The technology is still in its very early stages and not likely to see any application—nefarious or benign—in the immediate future. However, a pair of researchers from the University of Maryland have made important progress on the fundamentals of how it would work.
The “brain scanner” in question is a magnetoencephalogram, the standard non-invasive technology for monitoring the brain’s activity through its magnetic field. Such devices have been around for more than half a century, but interpreting the output has always been a challenge. This is one area where machine learning is proving its strengths.
Previous efforts with machine learning had shown that an algorithm trained on a particular subject’s neurological responses can make educated guesses at what they’re listening to. However, hearing their internal monologue was out of reach, because brain activity when imagining speech doesn’t look quite the same as activity when listening to it.
Interpreting Imagined Speech is Harder
Adequately training a model on imagined speech is difficult. That’s because there’s no way to verify exactly what words the subject is thinking, and when.
Maryam Maghsoudi and Shihab Shamma solved this problem in two ways. Firstly, they recruited musicians as their experimental subjects and had them memorize music and poetry with a specific rhythm. The musicians’ ability to conjure the imagined audio with the same tempo as the real thing allowed direct comparison of the heard and imagined audio.
The researchers then used two separate models to decipher the brain activity. The first takes a scan of imagined speech and converts it into a prediction of what the brain activity would have looked like if the subject was actually listening to a recording. The second component then does the more familiar task of deciphering that auditory response and outputting guesses at what words were involved.
The results were significant, but far from a ready-for-market product. The machine’s guesses were “better than chance,” but not perfect. Moreover, the experiment involved a vocabulary of only 76 words and, of course, required the willing cooperation of trained musicians.
Still, it’s a fascinating proof-of-concept that it is possible for a machine to “hear” the words you’re thinking.
Thought-Listeners for Good or Evil
Dystopian science fiction has taught us to fear the idea of technological mind-reading, and for good reason. Nonetheless, the researchers envision their project as life-changing for disabled patients—and not the CIA.
In the introduction of their paper, they explain:
“The ability to decode imagined speech could benefit patients with motor disabilities such as amyotrophic lateral sclerosis (ALS) or locked-in syndrome (LIS) who cannot produce overt speech, thus enabling them with other individuals.”
Additionally, there is evidence that for autistic people who are deemed “nonverbal,” the issue often occurs only at the point of generating speech. That is, they can understand speech and generate internal monologues, but just can’t “get the words out.”
For these and other categories of people who have thoughts that they can’t vocalize, a computer thought-to-speech interface could be life-changing.
Still, it’s hard not to think about such a technology without imagining its potential for abuse by parties ranging from scammers to foreign powers to overreaching law enforcement. The good news is that we’re still several steps away from those possibilities.
Getting the technology to work reliably with an individually-trained model and a willing subject is one thing. Something that would pry thoughts from anyone against their will is another matter altogether.
Also in Science News
Could Evaporating Black Holes Have Driven the Big Bang?
A “Swiss Cheese” model of the very early Universe could resolve one of the big puzzles surrounding its formation. If spacetime (the “cheese”) was pockmarked with a large number of tiny black holes, that would produce the outward acceleration that puts the “bang” in the Big Bang. So says a trio of Greek astrophysicists, anyway.
Black holes are gravitationally powerful, so we typically associate them with attraction, not repulsion. However, there’s a phenomenon called Hawking radiation that makes small ones “evaporate.” When they do that, the emitted energy pushes other nearby things away. In essence, the theory of these three Greek astrophysicsts is that the early Universe was full of tiny, evaporating black holes that flung each other apart and conveniently consumed themselves in the process. That tidily answers both why “inflation” happened, and also why it stopped.
Hawking radiation occurs due to an interaction between quantum mechanics and general relativity. Quantum mechanics forbids space from ever being truly empty, as that would be too much certainty. So, particle-antiparticle pairs are constantly popping up in a vacuum and quickly annihilating each other before their energy is swallowed back up into that baseline uncertainty. However, if they happen to pop up at the edge of a black hole, one can get swallowed up before colliding with its mate. If the other one escapes the pull, it’s emitted as radiation… but the black hole now owes an energy debt to the laws of physics, which it has to pay back with some of its own mass.
The smaller the black hole, more abrupt its event horizon, which means this happens more often. Below a certain threshold, the black hole will quickly Hawking-radiate itself into oblivion while emitting energy in all directions.
Biodiversity of Flowering Plants at Particularly High Risk
Climate change threatens to drastically reduce the world’s biodiversity, and among plants, the flowering kind appear to be at the greatest risk. A group of researchers claim that up to 16% of the world’s plant species could be at high extinction risk by the year 2100, but that flowering plants are already close to losing 21% of their evolutionary history.
This idea of “evolutionary history” is based on the idea that species on more sparsely populated branches of the evolutionary tree are worth more to biodiversity than others. For instance, if you had to choose between losing one of 14,000 species of ants or the red panda (with no living relatives), you’d choose to save the panda.
Weighted that way, turtle and crocodile species are the animal groups of greatest concern under climate change. Less attention has been paid to plants, but the researchers found that flowering plants (angiosperms) are in a more precarious position than other clades.
LunaPure: Portable Solar Device Could Be Key to Water on the Moon
An Ontario company won the Canadian Space Agency’s $400,000 contest to solve the problem of producing drinkable water for a future manned Moon mission. LunaPure, by Canadian Strategic Missions Corporation (CSMC), was the winning entry: a unit the size of “a box of books” that uses solar power to melt ice and then purify it.
Although the Moon is largely devoid of the necessities for life, water ice exists in the permanent shade at the bottom of certain craters. That, and the technology to turn it into usable water, will be key for any hypothetical moon base.
Once electrolyzed, water also provides the necessary components for rocket fuel — hydrogen and oxygen. The rationale for such a base, then, would be as a resupplying station for other missions, as the Moon’s weaker gravity would make each relaunch more affordable.
CSMC’s other projects include miniaturized nuclear reactors for use in the Arctic and other extreme environments, potentially including space.
