What ever happened to “fuzzy logic.” It was all the rage in the early 1990’s. Japanese companies were building washing machines and bullet trains controlled by fuzzy logic. How much of this was hype? Are people still using fuzzy logic (aside from a few aging acolytes of Lotfi Zadeh)?
Wikipedia on Fuzzy Logic
It’s still very damn useful as a metaphoric concept, at least!
Many of our common-sense sets do not have sharp edges. I mentioned the set of “chairs” in another thread. The central core of the set is obvious: office chairs and kitchen chairs and big overstuffed living room chairs. But farther out toward the boundaries of the set, the definitions are fuzzy. Is a chaise-longue a “chair?” Is a “rocking chair” a chair? How wide, exactly, does a “chair” get before it becomes a sofa?
And then there’s the real fringes: a wooden box with a cushion on top. Or (garish image!) a person on hands-and-knees, upon whom one might sit. Or a bar stool, with the seat removed: just a pylon. You could sit on top of it. (Most uncomfortably.)
The idea of sets having “fuzzy boundaries” makes for a topology that more accurately reflects the way our language treats objects. The traditional Venn topology – an item either is or is not inside the set – doesn’t work as well in a universe full of gradations and spectra.
Sure, but how does that allow us to build better electronic controllers? What attributes of a system make it a candidate for fuzzy control?
The thing is, this ws well known to anyone who had spent more than thirty seconds thinking about things long, long before Mr Zadeh dreamt up a new, cuddly name for it. Fuzzy logic was always very old wine in flimsy new bottles (or maybe in boxes).
It’s still used somewhat in Machine Learning, but in my experience people tend to strongly prefer probability. There’s a very real bias that Fuzzy Logic is just probability in drag. The thing is, in reality they’re not mutually exclusive, and can be used in tandem – though to be honest, I’ve never tried similar problems with both of them to see if the math works out significantly differently (i.e. certainty that thing IS a ghost vs likeness to a ghost), but I find it useful for my own reasoning. The bias isn’t bad enough that you’ll get laughed out of the room for using it or proposing it as a solution to a problem, but most times I’ve suggested it as an elegant solution to a problem on a conceptual level I get a lot of “eh… too much trouble for little to no clear cut benefit” reaction.
ETA: Though I have gotten a group to agree to use it once and there were some blood wars over the definition of what constitutes “similarity” in some cases, on the order of ridiculousness as “well these things have at least SOME similarity, after all, they’re both matter!” 
Just bought a couple of LG washing machineswith fuzzy logic control over in India. (LG is Korean, rather than Japanese)
Fuzzy logic is 63% dead, 52% alive and 76% undead.
I think you’ve just hit on the problem with fuzzy logic. There’s no way to objectively determine that an object is X% similar. There is a way to objectively determine that an object has an X% probability of being the same–you run multiple tests.
Three chairs for this late reply.
Stealing…
7 years late, but stealing it anyhow.
Is fuzzy logic basically a ‘continuous’ multimodal logic, i.e. a logic with an infinite number of operators, or is it the ‘usual’ logic but with the truth values allowed to vary continuously between 0 and 1? How does it differ from continuous probability?
Thanks,
It has some strong resemblances to probability, but it isn’t quite the same. I’m not expert enough to give a better answer than that.
This.
The wiki cited in post #1 is decent. Here’s my favorite soundbite summary from that:
Both degrees of truth and probabilities range between 0 and 1 and hence may seem similar at first, but fuzzy logic uses degrees of truth as a mathematical model of vagueness, while probability is a mathematical model of ignorance.
As argued elsewhere, there is some legitimate epistemological distinction here, but whether it amounts to a practical difference is increasingly evident to be “No”. Or at least not enough to matter.
I would dispute “well known”. Heck, you can still find plenty of instances of people claiming sorites paradox (which grain of sand marks the transition to a heap or whatever) is real and unsolved.
I don’t understand this. The paradox is certainly real and is encountered in many places.
As to whether it’s solved, none of the resolutions given in Sorites paradox rise to the level of solving the problem. They just try to get around it, unsatisfactorily. Unless you mean that Supervaluationism does, but I don’t understand that either.
Simply accepting that “it is not true for all heaps of sand that removing one grain from it still makes a heap” gets around it fairly well, IMO. It reminds that the paradox arose because of unwarranted assumptions in much the same way that Russell’s paradox and similar conundrums involving self-reference can be avoided by clarifying your axioms and/or restricting what can be said about what (i.e. classes). Although I admit it is not a terribly satisfying way out.
I will concede that since the wikipedia page doesn’t say it is solved, then there is no consensus that it has been solved. Thus I take back what I said previously.
However, I would maintain that IMO fuzzy logic should basically put the issue to rest. It is not that any particular grain transitions a heap to non-heap or vice versa. It’s that each grain affects the “heap-ness” of the set. Like how making alterations to a seat can affect the “chair-ness” of it, in a non-binary way.
And, when you ask a human whether a collection is a heap, or something is a chair, or a man is bald, there is something like a fuzzy collapse to decide which binary classification we’re going to approximate to, this time.
Ref @Mijin & @KarlGauss. Exactly. The fault line is in the idea any collective class has a sharp boundary. Or more precisely, it’s in the idea that every collective class has a sharp boundary and so therefore e.g. heaps must have such a boundary. When they self-evidently don’t.
Lots and lots of Ancient Greek logic foundered on the falsely excluded middle that they simply refused to acknowledge existed.
To the degree “fuzzy logic” by whatever name helps us avoid falsely excluding middle cases and searching vainly for bright lines in uniformly varying shades of gray, it performs a useful service. All the more so if it brings some logical or computational rigor to the discussion.
Whether Zadeh’s formulations and others’ subsequent work up that alley amount to that rigor is a different question. But at least he was talking about the problem.
I can accept “you’re looking at it wrong” as a resolution. That does have a long history of reworking problems.
n.b. I just realized that resolution is solution with a re- prefix. You’d think therefore that they are related words, but etymonline.com shows that they have separate histories.