This could become a way for paralyzed people to communicate. It might become a way for the government to get information from people (and obviate attempts to get information by torture). At present the system requires not only our general knowledge of where things are typically thought in the brain, but knowledge of the brain operations of the specific individual, and this latter requires about 16 hours of investigation of the subject individual before successful mind reading.
If this system could overcome that arduous preliminary learning and if the system could be shrunken down to the size of a skull cap, perhaps hats would come back into fashion. A dating service might offer the hats to be worn for users of the dating service. It might be a sport to go on dates with these hats in which you get the low-down of what your date is really thinking about.
When x-rays were first discovered, the newspapers entertained the possible future in which people could walk down the street wearing glasses through which you could see the bodies underneath the clothes. But that was a very long time ago, and nothing like peeping glasses has eventuated so far as I know.
I’m very skeptical of these technologies. I don’t dispute that they are technically impressive. My objection is that news stories (and likely, scientific journal articles) seriously underplay the differences between technical laboratory findings and the requirements of real-world operationalization at scale. There always seems to be an assumption that once a novel scientific finding has been confirmed in a lab, the real-world operationalizations are trivial–just a matter of time, tweaking, details, etc. This is a little bit like the assumption that if you can find a metaphysical “grounding” for ethics, ethical decision-making becomes a trivial matter of “applying” it to this or that issue.
As you may or may not know, I work in health care revenue management. In effect, what we do is to take a bill from the hospital setting, and establish the workflow and data flow that takes the bill to final disposition–ideally, payment in full. There is nothing conceptually difficult about any part of this process. A procedure is scheduled, then done, then coded, then billed. The bill is sent out first to insurance, then to the guarantor. Insurance either pays or denies. The denials are either appealed or written off. The guarantor either pays or not. If not, the bill is either pursued or dropped. Each of these conditionals can be broken down into several dozen conditionals, and there are many processes involved. But none of this is rocket science, or neuroscience, or science fiction. “All” we have to do is get the data from the hospital to us, to the insurance company, and to the patient, maybe to a few other vendors or a collection agency or two, maybe to a few outsourced clinicians or two, maybe, occasionally, to a law firm. AI plays an important role along the way. AI mines for data, and AI does the coding.
How hard could this be? A high school kid could probably write every program required to get the whole thing done. The AI applications and algorithms involved are primitive. And yet, no one has found an efficient way to get the process from A to Z. That’s an understatement. I work in Operations Support. It’s hard to convey what my job is like. I wish I could, in a way that would be understandable and relevant to outsiders. And not just at my company. Everywhere.
Every existing hospital information system is a mess. The task of extracting and transforming that data into monetizable form is a trial by fire. Error plagues every process. Huge amounts of sensitive data are pointlessly exposed for years, even decades at a time, to thousands of users who have no reason to see it. It’s common knowledge that there is no reason for anyone to use a Social Security number for identification purposes in a health care setting, and yet they are ubiquitous. Doctor’s offices will routinely ask patients for Social Security numbers, and refuse service if the patient refuses to comply. About a week ago, I saw a bill for a balance of $185.50, for a procedure done in 1988, for a person who had been deceased for years. The account was still in collections, and the person’s Social Security number had been exposed for thirty years. We take this sort of thing for granted, but on reflection, one thinks: how can we reconcile news of such awe-inspiring technological advances with practices that are this primitive and absurd?
The answer is that “applying” science is not at all trivial. People forget how many variables are involved in applying any abstract finding to the real, friction-laden world of people and institutions. That’s my response to so many of these jaw-dropping findings. If applying them were trivial, we would have solved much simpler problems by now. But some of the simplest problems remain intractable. It will be a long time before AI solves any of them, or any of the others.
LikeLiked by 1 person
Irfan, there’s a simple solution to all these problems you mention. Just use the flowchart. Everyone should just use the flowchart.
LikeLiked by 2 people