Tuesday, March 31, 2020

Another Increment in Mind Reading Technology


For the wonks in the crowd


In recent years, an area of research called brain-machine interface (BMI) technology has made incremental progress in reading minds and translating that into useful outcomes. A technical report in Nature Neuroscience describes another incremental improvement, namely an increase in the speed and accuracy of reading minds using electrodes that set on the surface of the brain.

The paper describes the state of the art like this:
“In the last decade, brain–machine interfaces (BMIs) have transitioned from animal models into human participants, demonstrating that some amount of motor function can be restored to tetraplegics—typically, continuous movements with two degrees of freedom. Although this type of control can be used in conjunction with a virtual keyboard to produce text, even under ideal cursor control (not currently achievable), the word rate would still be limited to that of typing with a single finger. The alternative is direct decoding of spoken (or attempted) speech, but heretofore such BMIs have been limited either to isolated phonemes or monosyllables or, in the case of continuous speech on moderately sized vocabularies (about 100 words), to decoding correctly less than 40% of words.”
A BBC article summarizes the results like this:
“Scientists have taken a step forward in their ability to decode what a person is saying just by looking at their brainwaves when they speak. They trained algorithms to transfer the brain patterns into sentences in real-time and with word error rates as low as 3%. Previously, these so-called "brain-machine interfaces" have had limited success in decoding neural activity. The earlier efforts in this area were only able decode fragments of spoken words or a small percentage of the words contained in particular phrases. Four volunteers read sentences aloud while electrodes recorded their brain activity. The brain activity was fed into a computing system, which created a representation of regularly occurring features in that data.”
As usual, there are cautions to consider. First, the electrodes need to be placed on the brain. That is highly invasive. Second, it took a lot of electrodes to attain high word reading accuracy. Third, the decoded speech was spoken, not read from text, and limited to 30-50 sentences with a 250 word vocabulary.

What was extremely interesting was the ability of the BMI and software to learn. The word decoder they used was not just classifying sentences based on structure because accuracy increased by adding new sentences that were not used in the original tests. That data was interpreted to mean that the machine interface can identify single words, and not just whole sentences. If that is true, then it could be possible to decode sentences never encountered in a training set. The authors commented: “Although we should like the decoder to learn and exploit the regularities of the language, it remains to show how many data would be required to expand from our tiny languages to a more general form of English.”

When the computer system was trained on brain activity and speech from one person before training on another, the decoding results improved. That was interpreted to mean that the technique may be transferable across people.

Obviously, this line of research will be pursued. It is too important to not pursue it.


Why this could be important
It is not clear how far mind reading technology can progress. Research over the next ~10 years should start to clarify what inherent limits, if any, there will be in how far it can progress. If one day it is possible use machines that can read minds without invasive procedures, the impact on society could be enormous. For example, people testifying in court probably would not be able to lie and deceive nearly as well as they can now. In theory, the mind could be contradicting oral lies and deceit. Politicians making statements to the public could also similarly be fact checked in real time for lies and deceit.

The thought experiment is this: What is going in in your mind when you lie to someone? How can you really know all that is going on up there in brainlandia? Specifically, you may not consciously think of the lie itself, but your unconscious mind by be thinking, ‘that's a big, fat whopper I just told’. If that turns out to be true, or even if your conscious mind cannot help but think, ‘that's a big, fat whopper I just told’, imagine how different the world would probably be.[1]

For better or worse, this line of research will continue. We will find whatever impassible limits there may be. If there are none, then mind reading machines just might transform the world in one of at least three ways. First would be a world of far less lying and deceit by rich and powerful people and by crooks and liars, more civility, and more widespread prosperity and well being. Second, would be about the opposite. Third would be something in between.


Footnote: 
1. Yes, I know. Many people, maybe most, will instantly and strongly object to such an outrageous intrusion on their privacy and maybe even their freedom. That is a legitimate concern. That is why one wants to live in a civilized democracy that operates under the rule of law and social comity. We all know what will happen if mind reading technology is in the hands of demagogues, tyrants, murderers and kleptocrats. They will use it against political opposition, not themselves. 


More evidence that no two people are alike --
just look at how those brains work differently


No comments:

Post a Comment