Well worth a read.
Ignorance is insufficient data. So it doesn’t matter how smart you are; if you don’t have enough data to solve a problem, you’ll never solve it.
Intelligence is finding very simple solutions to complex problems. For example, if you say to someone, “You made that look effortless!” you’re saying to them, “You’re smart.” If you say to someone, “You made that look really hard,” you’re saying, “That’s stupid.” So stupidity is using a rule where adding more data doesn’t improve your chances of getting it right; in fact, it makes it more likely that you get it wrong.
If intelligence is making hard problems easy, genius is making problems go away! Prior to the so-called scientific revolution of the 17th century, we had extraordinary theories for how the planets pursue their orbits, the Ptolemaic system, Tychonic system, and so on. Every time someone made a new observation of a new planet, they had to add more and more complexity to their models—epicycles and deferents and so forth. Johannes Kepler came along and said, you know, I can replace all of that with the ellipse. So take that whole structure—he didn’t make that work better, he didn’t oil that; he just threw it away! And he replaced it with Kepler’s laws. It was an extraordinary simplification of the problem.
I don’t think studying the brain through neuroscience and psychology is a mistake, but I think it’s very limited. I think there’s a whole series of reasons why that’s true. One of them is the huge anthropomorphic bias that we have, which is, “What’s intelligence? Well, it’s just what we do.” In other words, it’s language. And so by definition, since as far as we are aware Homo sapiens is the only species on Earth with a sophisticated grammar, it’s the only species that’s intelligent. Of course, the ludicrous examples of that are: [Those] that use oil painting; or have a printing press... So that’s the history of thinking about intelligence in fact, excluding others—other human beings or other species. So that’s one huge problem and it’s manifested in the fact that most studies of intelligence are studies of intelligence in one species.
I agree that intelligence is a hard thing to define and has sometimes been defined in too limited a way. Modern animal studies seem to be overcoming that, however. Later, when he defines athletic skills to be a form of intelligence, I think he goes too far in the other direction. That's such a broad definition that it becomes meaningless. I also don't think having apps suggesting where to eat impinges on free will, but that's another story.
http://nautil.us/issue/23/dominoes/ingenious-david-krakauer
/surveil
ReplyDeleteBut it's simply not true! You never have enough data, but you have to manage very complicated problems. It is typical. That's why typical biological mechanisms works on two modes: fight or run away. And this is one of the best solutions of very basic problem: how to survive.
ReplyDeleteAnd there's a lot of other strategies that works in similar way: do not eat plants when they have milky juice, do not stay visible, be careful at unknown etc.
Even engendering problems have a lot of similar heuristics, which may give you better chance to solve it before you even started to perform deep study.
So very first sentence of this essay is just not in agreement with facts ( with data :) ). There is usually small subset of factors which is important when solving problems, and problems without such characteristics ( when you should "know everything" to solve it) usually are to complicated to be solved in any way at all...
Never enough data eh ? Well, you can have some of my years of backlogged HI data that needs analysing if you like... :P Never mind never enough data, it's easy to have far too much data ! And some of it is utterly irrelevant, like the entire content of the Daily Mail or Trump's "brain"...
ReplyDeleteKazimierz Kurz Part of intelligence is search despite inadequate or misleading data.
ReplyDeleteRhys Taylor a lot of data ( even "all possible data" ) is as much thread as no data at all. It is not data, which allows You to act. it is theory which data matters does...
ReplyDeleteKazimierz Kurz But you can't even construct a theory without data. Not one that has any chance of being applicable to the real world, anyway.
ReplyDeleteRhys Taylor A good theory is parsimonious both in its own complexity and in the data it requires.
ReplyDeleteEdward Morbius Indeed. But only insofar as simpler theories are easy to test and have fewer free parameters, and not in that they are, as is popularly expressed, more likely to be correct.
ReplyDeleteRhys Taylor parsimonious doesn't meaan minimising, but minimising wwithin an effectivenes constraint. There's such a thing as minimal essentiaal complexity to solve a given problem.
ReplyDeleteCarl Zimmer's "Meeet the Animats" is a great illustration of a complexity floor, though it doesn't add the additional element of a complexity constraint or cost.
https://www.nationalgeographic.com/science/phenomena/2013/08/02/meet-the-animats/?x
nationalgeographic.com - Meet the Animats
Edward Morbius Very nice article !
ReplyDeleteMy point stands though. The Universe isn't under any obligation to solve things efficiently or even be comprehensible. We should assume the former only because it makes things easier to test, and the latter because we don't have any choice.
Rhys Taylor Right, but the limit for theory and information is a 1:1 mapping to terrain (see Borges). Actually, long before that point, decisionmaking speed comes to the fore. If you can adapt to either the environment or an adversary more rapidly than it responds to you, and at least statistically not box yourself int a corner, then you have the advantage, so long as there is a selector which can assess your decision after the fact. Random decisionmaking (unbiased) is better than a model, in this case.
ReplyDeleteIf you're cooperating with a large cohort, either a predictable behaviour or comms (or both) is an assist. German WWII tanks beating Frech armour superior on all specifications but lacking radio. Germany could seize and capitalise on opportunity. France didn't know where to be or what to shoot.
Rhys Taylor No, You can construct theory without data. It is very easy! You cannot test it without data. It is completely different area.
ReplyDeleteOn the other side, if You have only data, and no theory at all - you know nothing and cannot act.
For example - could You tell me what about this data: 13, 16, 18, 37, 1,12, 34 m? What do You learn from them?
Kazimierz Kurz That raises the question of what exactly science is, and if there's a progression of development.
ReplyDeleteMythology, lore, craft, taxonomy & cataloguing, disconnecteed theory/ law, central theory/law, complexity / interconnection/ unintended consequences..
Biology and geology both come to mind. Early "theory" consisted of mythology, religion, and traditional practioners. Writing and education formalised this, and you saw a period of collectors, and formation of rules of thumbor generalisations, but still no central theory. Linneaus and early geologists such as Lyell served largely as organisers, cataloguing and classifying information, looking for (and exposing) patterns, but not coming up with central theory.
(Trivium: Linneus invented index cards, to be able to capture and restructure information.)
The breakthrough in biology came with Darwin and the theory of evolution, which not only organised existing knowledge, but predicted future developments: genetics of Mendel (modulo some fraud), and the double helix of Watson & Crick.
For geology, it was the discovery of radiation, the realisation of half-lives, and the clock that this created for telling the age of rocks, which ultimateley gave some awareness of just how long geology's "long time" was: not thousands or even millions of years, but billions. This in the early 20th century. It also set the stage for plate tectonics, giving several pieces of the puzzle which finally came together: time, the requisite thermal energy from radioactive decay, clocks for aging specific geological formations, and part of the toolkit for matching up formations, now wwidely separated, once proximate.
The upshot: beginning in the 1860s, and continuing through the 1950s and present, evolutionary genetics became the central organising and informing conceptof biology. And beginning in the 1910s, evidence for plate tectonics emerged until it became accepted theory, and the central organising principle of geology.
Which raises the question of what pre-evolutionary and genetic biology, or pre-tectonic geology were. Were they not science? Or a prior state?
And what does this say of various other ... problematic ... domains? What exactly are the central organising principles of psychology, sociology, anthropology, economics, political science, ecology, epidemiology? Questions I find myself asking frequently, and for whichI've failed to find satisfactory answers.
Kazimierz Kurz You have to have something to theorise about. Nothing will come of nothing.
ReplyDeleteOh, also this : astrorhysy.blogspot.com - Fifty Shades Of Science
ReplyDelete