An island is not necessarily love
Reports from laboratories trying to decipher the contents of the human brain are certainly worrying to many. Looking closely at these techniques, you will calm down a little.
In 2013, Japanese scientists from the University of Kyoto succeeded with an accuracy of 60% "read dreams »by decoding some signals at the beginning of the sleep cycle. The scientists used magnetic resonance imaging to monitor the subjects. They built the database by grouping objects into broad visual categories. In the latest round of experiments, the researchers were able to identify the images the volunteers saw in their dreams.
Activation of brain regions during MRI scanning
In 2014, a group of researchers from Yale University, led by Alan S. Cowen, exactly recreated images of human faces, based on brain recordings that were generated from respondents in response to the images shown. The researchers then mapped the participants' brain activity and then created a statistical library of the test subjects' responses to individuals.
In the same year, Millennium Magnetic Technologies (MMT) became the first company to offer the service "recording thoughts ». Using our own, patented, so-called. , MMT identifies cognitive patterns that match the patient's brain activity and thought patterns. This technology uses functional magnetic resonance imaging (fMRI) and biometric video analysis to recognize faces, objects, and even identify truth and lies.
In 2016, neuroscientist Alexander Huth of the University of California at Berkeley and his team created a "semantic atlas" for deciphering human thoughts. The system helped, among other things, identify areas in the brain that correspond to words with similar meanings. The researchers conducted the study using fMRI, and the participants listened to broadcasts telling different stories during the scan. Functional MRI revealed subtle changes in blood flow in the brain by measuring neurological activity. The experiment showed that at least a third of the cerebral cortex was involved in language processes.
A year later, in 2017, scientists at Carnegie Mellon University (CMU), led by Marcel Just, developed a way to identify difficult thoughtsfor example, "the witness screamed during the trial." The scientists used machine learning algorithms and brain imaging technology to show how different areas of the brain are involved in building similar thoughts.
In 2017, Purdue University researchers used mind reading Artificial Intelligence. They put a group of subjects on an fMRI machine, who scanned their brains and watched videos of animals, people, and natural scenes. This type of program had access to the data on an ongoing basis. This helped his learning, and as a result, he learned to recognize thoughts, patterns of brain behavior for specific images. The researchers collected a total of 11,5 hours of fMRI data.
In January of this year, Scientific Reports published the results of a study by Nima Mesgarani of Columbia University in New York, which recreated brain patterns - this time not dreams, words and pictures, but heard sounds. The collected data was cleaned and systematized by artificial intelligence algorithms that mimic the neural structure of the brain.
Relevance is only approximate and statistical
The above series of reports of successive advances in mind-reading methods sounds like a streak of success. However, development neuroformation technique struggles with enormous difficulties and limitations that make us quickly stop thinking that they are close to mastering them.
First, the brain mapping joke long and costly process. The aforementioned Japanese "dream readers" required as many as two hundred trial rounds per study participant. Secondly, according to many experts, reports of success in "mind reading" are exaggerated and misleading the public, because the case is much more complicated and does not look like it is portrayed in the media.
Russell Poldrack, a Stanford neuroscientist and author of The New Mind Readers, is now one of the loudest critics of the wave of media enthusiasm for neuroimaging. He clearly writes that activity in a given area of the brain does not tell us what a person is actually experiencing.
As Poldrack points out, the best way to watch the human brain in action, or fMRI, is just indirect way by measuring the activity of neurons, as it measures blood flow, not the neurons themselves. The resulting data is very complex and requires a lot of work to translate it into results that can mean something to an outside observer. also no generic templates – each human brain is slightly different and a separate frame of reference must be developed for each of them. Statistical analysis of data remains very complex, and there has been much debate in the fMRI professional world about how data is used, interpreted, and subject to error. That's why so many tests are needed.
The study is to infer what the activity of specific areas means. For example, there is an area of the brain called the "ventral striatum". It is active when a person receives a reward such as money, food, candy, or drugs. If the reward were the only thing that activated this area, we could be pretty sure which stimulus worked and with what effect. However, in reality, as Poldrack reminds us, there is no part of the brain that can be uniquely associated with a particular mental state. Thus, based on activity in a given area, it is impossible to conclude that someone is actually experiencing. One cannot even say that since “we see an increase in activity in the brain island (island), then the observed person should experience love.”
According to the researcher, the correct interpretation of all the studies under consideration should be the statement: "we did X, and this is one of the reasons causing the activity of the islet." Of course, we have repetition, statistical tools and machine learning at our disposal to quantify the relationship of one thing to another, but they can at most say, for example, that he is experiencing state X.
“With fairly high accuracy, I can identify the image of a cat or a house in someone’s mind, but any more complex and interesting thoughts cannot be deciphered,” Russell Poldrack leaves no illusions. “However, remember that for companies, even a 1% improvement in ad response can mean big profits. Thus, a technique does not have to be perfect to be useful from a certain point of view, although we do not even know how great the benefit may be.
Of course, the above considerations do not apply. ethical and legal aspects neuroimaging methods. The world of human thought is perhaps the deepest realm of private life that we can imagine. In this situation, it is fair to say that mind-reading tools are still far from perfect.
Scanning brain activity at Purdue University:
