Back in the earlier days of IT (information technology), a distinction was made – and emphasized – between data, information, and knowledge:
-
Data was that which had no real content or context or, perhaps stated more accurately, simply existed as fact, e.g., 6 Aug 1945. It is no different than any other date, it is a mere way of marking time in the modern age. Similarly 32 deg. F is simply a place on a temperature scale, no more, no less. $25 is simply an amount specified in a given currency.
-
Information was defined as data that could be used for something, especially for making decisions. Using the same examples, it turns out that 6 Aug 1945 was the day the Americans dropped an atomic bomb on Hiroshima. 32 deg F is the point on the (Fahrenheit) temperature scale at which water freezes at sea level. $25 is the price of a book which I recently purchased. The information -value of these examples does depend on context: perhaps there is a discussion going on about when the Nuclear Age actually began, or you’re taking a science quiz, or I realize I was actually overcharged for my purchase.
-
Knowledge, at that time, or at least in that particular context, was considered information that could be applied in particular ways to achieve particular results: say, the fact that the bombing of Hiroshima preceded the bombing of Nagasaki by three days, or to document the case that it was a superfluous act as it did nothing to actually shorten the war. And, knowing the freezing point of water is helpful when designing, say, consumer products for use in the home for when water freezes it expands. Or that dollars area currency used in many places but without further specification the one meant is the US dollar.
These examples deal with facts, but a similar case obtains with processes. I always found the above to be a useful distinction, but the Google came along and redefined everything that can be digitized as “information”, riding on the coattails of early information theorists who reduce fundamental binary distinctions (bits) as carriers of information. The two meanings of the words were conflated and now we have no real distinctions anymore. It is very difficult to discuss these matters in a reasonable way because many discussion participants are unaware of or unwilling to recognize what terms may mean outside the technical domain.
Many process which seem exceedingly complicated or complex from the outside when analyzed can be broken down into meaningful chunks. In the early 80s, I was working on defense-department documentation that was so designed that any soldier/sailor/airman (or their female counterparts) could repair the given equipment or system in question even without technical knowledge or training in maintenance and repair. It takes a lot of work to identify all the minute little steps, and it produces a ton of documentation, so the hardcopy production process was stopped, but then digital technologies came along allowing you to store what might be tons of documentation on a small, portable, digital medium, and the program was back in business.
This is precisely what is behind the self-diagnostic systems in all modern cars. There was a time you had to go to your mechanic of trust to find out what was wrong with your car, but today the car tells you to go to the garage. All that knowledge, if you will, has been digitally encoded into the software diagnostics of the car, reducing the need for highly skilled technicians. The software says swap out module X and module X gets swapped out. The skills necessary to do the swapping are minimal. On the other side of the coin, the manufacturing of the cars can be highly automated because standard procedures can be encoded and programmed into robots to do the jobs that were once done by humans.
Back in the 60s our family went on one of our few vacations to Detroit of all places, and we toured one of the car factories there to watch how 20,000 workers in three shifts pushed tons of metal out the doors to make American mobile. In 2012, I had the opportunity to tour the BMW factory in Spartanburg, SC which was producing four times as many cars as that factory in Detroit, but with a mere quarter of the personnel. All of those skills (also a kind of knowledge) and techical knowledge had been encoded into robots that were doing the bulk of the heavy work in particular.
Now, we should recognize that all of this is possible because humans can do such things. We can take what we know and we can convert a lot of it into digital systems that reduce the number of humans needed to do that kind of work. What they’ve refined in manufacturing, they have expanded into many other areas: low-level legal processes, certain areas of medical diagnosis (!), accounting, automated teller machines and checkouts, just to name a few of the most obvious. And all of this has significant consequences for humans. Kurt Vonnegut’s Player Piano was one look at it.
I don’t want to even address the AI aspect of it all. Just as “knowledge” has been more or less redefined as “access to information”, so too has “intelligence” been redefined into something akin to “manipulation of information”. Though I’m a curmudgeon by nature, I’m not a luddite, yet I can say with a certain amount of confidence that the “I” in AI is more likely “Ignorance” … it has little, if anything, to do with intelligence.
Just because a whole team of experts and programmers can get together and program a computer to play chess by calculating probabilities faster than perhaps humans can certainly gives one the impression that the system is “smart”, but it isn’t. In this connection, I highly recommend TJ William’s “At Play: A Personal Odyssey in Chess” and its related discussion, as many of the themes that were raised by Durwin’s post were focused on there.