Is the IQ meme finally over?

Some guys at Northwestern developed a network model that outperforms the majority of humans at Raven's Progressive Matrices.

mccormick.northwestern.edu/news/articles/2017/01/making-ai-systems-see-the-world-as-humans-do.html

qrg.northwestern.edu/papers/Files/QRG_Dist_Files/QRG_2010/Lovett_CogSci10_Ravens_Final.pdf


What does this imply about the true relationship between "intelligence" and IQ scores?

Other urls found in this thread:

urbandictionary.com/define.php?term=boop
twitter.com/AnonBabble

Not to defend IQ tests, but why shouldn't a machine be able to show "intelligence"?

>evaluating a function is a show of intelligence

I never said that. But now you're hinting at a better argument than "this test is bullshit because a machine can do it".

the thing is g is a part of a larger system.

expert systems are hyperautistic networks that will always outperform humans, but does that mean that human mind as a whole does not exist?

As I said, g in humans makes sense because is a part of a whole. A fucking calculator predicting patterns is not a proof that IQ is meaningless.

IQ tests are, when applied to a large human population, positively correlated with other parameters that could typically denote intelligence. That doesn't say anything about the possibility to have a dumb person score high in IQ tests if they've been trained their whole life to do that and just that. They could be the most moronic person, good only at specific types of pattern recognition but unable to solve basic algebraic problems, to cook an egg, to show good reading comprehension, or to solve basic problems in other settings, but yet perform very well on IQ tests. It's not usually a problem, because, again, IQ tests are used to measure some kind of indicator of the average intelligence of a large population, so special outliers like that wouldn't matter.

Similarly, if you can train a machine to perform super well on IQ tests, it still tells you absolutely nothing about IQ tests being bad, about the machine being as clever as humans, or anything. If you attempt to draw any philosophical conclusion from that experiment, you're only showing that you don't understand what IQ tests are. The conclusion you *can* draw from this experiment is that some people managed to create a pretty damn good pattern recognition machine specifically tuned for IQ tests and that's a nice part-theoretical part-technical achievement. That's it.

what is the solution?

>IQ tests are used to measure some kind of indicator of the average intelligence of a large population
So IQ tests are not used to measure an individual's intelligence?

Only braindeds need to do IQ test because they are insecure about it

i would give a slight change to your answer:
sometimes brainlets (school teachers) make others (gifted kids) do an IQ test because they need to feel in control due to their insecurity in presence of an intelligence they cannot comprehend, and putting a number to an intelligence that vastly eclipses yours is a form of "controlling" the unknown.

They're not originally designed for that. People use them like that anyway because it boosts their ego, and people who actually make/run tests let you believe that it works that way because they benefit from you taking this test. But no, the correlation between one's intelligence and one's results at a IQ test is much less relevant than if you're working on larger populations, because the law of large number doesn't help cancel out the various biases that you may modify a single person's results.

>now you're hinting at a better argument
they're essentially the same argument actually

i get it! so if we want to know a black's intelligence we just find out the average black intelligence and that'll be the most accurate measure.

I'd say about 60% of the IQ meme is due to insecure whiteboys wanting a reason for intellectual dickwaving.

IQ isn't a great measure of success/intelligence; gratification deferral is.

They hold fast to their belief that IQ as measured by 3x3 matrices with little pictures in them is purely genetic and have nothing to do with nurture, wealth, and training, when we can literally design a mathematical function to solve them.

Do you believe that your thoughts are anything more than a function of the input your brain receives?

i think because they follow an algorithm. they don't have a thought of their own

D

What bullshit logic. Why does "following an algorithm" or "evaluting a function" not count as intelligence? That's the most general thing in the world for a computer, and the difference between that and """""actually really thinking"""""" is purely semantics. You'd say "Well, a computer can't do anything BUT follow its own logic! It has no choice!" but that's stupid because its like saying you have no choice but to act like yourself.

The moment when a programmer can't tell you exactly why his program output what it did, and particularly when that output turned out to be correct (such as deepblue or alphago beating world-class masters; I can guarantee you their programmers probably sucked at those games) is when programs began to show intelligence. No one told it what moves to play, they told it HOW to play, and it generated its own way to win that none of its programmers could have done.

It's still very niche and limited, but refusing to call it some form of intelligence, just shows that you think intelligence is "uniquely human" for some arbitrary reason.

correct!

D

Good post.

something something chinese room thought experiment

>What does this imply about the true relationship between "intelligence" and IQ scores?

That intelligence is, more or less, pattern recognition and processing. Which proponents of IQ have always claimed it to be.

Some better IQ puzzles for fun

If you don't get them in 60 seconds each you fail the puzzle

...

...

...

...

...

...

IQ tests ain't meant to detect intelligence, they're a diagnosis aid for doctors to evaluate the level of retardedness of ill/genetically deficient individuals. If you had to sit for an IQ test while you were young, no matter your score, I have bad news for you.

Raven's Progressive Matrices are a meme.

>five letters

that's funny i've been literally thinking about making a program that would do just that
the problem is it can be learnt, meaing you can improve your score, meaning the charts you see where majors having higher iq's the more math they do could mean it's the major that caused the higher iq and not the other way around
even iq questions that are letter/number-free aren't math free and a lot of the operations are something you can train when you are taking a math education

been wondering what the iq of people who make iq tests is

8
square with a circle inside
1st one
i don't know the alphabet
1+6 + 1+1 + 0(coutner) = 9
2+5 + 1+4 + 1(counter) = 13
1+6 + 5 + 2(counter) = 14
14
letters are numbers lalal i dont know the alphabet
67 or something, each time you double and you the current index of the number

>when we can literally design a mathematical function to solve them.
wait how does that in any way contradict that how well you do in IQ tests is genetic?

this isn't some fast board like you're used to, you can read the posts in the thread

In a way, yes. I definitely have mental-health problems, but the fact that I have scored 130-170 in tests has at least changed the docs attitude towards helping, and the same with teachers..
..or some of them:

oops

"ACHE"?

Man I'm up to 5 minutes and I still can't think of one. To be fair english is not my native language, but I thought I had a pretty rich vocabulary.

What is Veeky Forums used online IQ test? I did Mensa one but it doesn't give a nice meme number.

ills

>poops
kek

BUT THERE ARE SIX LETTERS WHAT THE FUCK

I think the real answer is ills since oops and boops is stretching it a bit.

urbandictionary.com/define.php?term=boop

nice user, i was thinking ears, "stears" doesnt exist though

>We should treat two functions that behave identically with identical inputs and identical outputs differently because super special magic that you can't see is happening in one but not the other

Shit, I discounted plurals and inflected forms even though it said nothing about that in the question.

Everybody knows IQ only works on children (detecting learning disabilities). Only retards think it measures an adults intelligence

It's not that a machine shouldn't be able to show intelligence. The problem is that intelligence as measured by the test can be boiled down to being approximated by a formula and repeated retraining of a model.

Essentially, this means you can study for an IQ test and get better results, like people have been saying for decades. It's bunk because it boils down to who studied the most for the recognizing the various matrices on the test. It's not necessarily anything innate or genetic, unless you're doing a blind test on someone who has never encountered recog tasks, like a child, and you're timing them.

Intelligence as we currently define it through IQ is bunk.

Look up what an algorithm is and you'll find out why it's not an adequate show of intelligence.

Performing an Optimization Algorithm and outputting a correct answer despite the programmer not explicitly programming the instructions != a show of intelligence.

Any problem you can model and then go ahead and solve with heuristics or something like Simplex LP and giving you a good answer does not at all mean machines are thinking, conscious creatures that are showing intelligence. It just means the CPU ran the generalized optimization algorithm.

What definition of "intelligence" automatically excludes algorithms?

>thinking, conscious creatures that are showing intelligence.
What exactly is then, faggot? You're just doing exactly what I said you guys always do, which is move and shift the goalposts away until "not being a computer" just becomes part of the definition.

Then why does an electrical signal travelling through your synapses = a show of intelligence?

the test can be trained without directly doing IQ tests
there's tons of skills you acquire throughout the education process that make you better at the test

not to mention it not measuring how good your memory is, how fast you learn new things - it's just not its intent

You information is outdated by some 50 years.

That depends on what we want to consider intelligence to be. Is this supposed to be a measure of something innate or intrinsic to a person, that states what their potential to solve problems is, that is non changing and permanent as an adult with the exceptions of decrease due to brain damage?
Or is this just a measure of the success rate of the current model you have thus far, and given that the model can be tinkered with and score improved, that the test doesn't really matter when it comes to ranking or judging people or even animals?
Because if a computer can do it, it seriously means anything can do it. Animals aren't biologically inclined to do it, as their brains are predisposed to do other things, but they can, as they have more raw computing power than a computer. All you have to do is train the net.

It doesn't necessarily show intelligence to just pump a signal. For all we know that might show someone pumping 10K Volts through a dead brain.

But I suppose why most people would consider it to be a show of intelligence is because the being which is pumping signals and backpropagating is conscious. A computer isn't conscious when it's optimizing routes or playing chess. It's running simulated neural nets or optimization algorithms and is essentially no different from an abacus. The difference would be if the computer was programmed with consciousness while doing all of that.

>A computer isn't conscious when it's optimizing routes or playing chess.
[citation needed]

>But I suppose why most people would consider it to be a show of intelligence is because the being which is pumping signals and backpropagating is conscious. A computer isn't conscious when it's optimizing routes or playing chess. It's running simulated neural nets or optimization algorithms and is essentially no different from an abacus. The difference would be if the computer was programmed with consciousness while doing all of that.
>Conscious
>aware of and responding to one's surroundings;
So how exactly is a computer not conscious? A chess-playing program is perfectly aware of the state of the board, just as we are aware of what we are given by our senses (the program has only one sense, which gives it the state of the board).

So if you're using a different meaning of consciousness, and I suspect you are, you'll have to define more specifically what you mean.

common sense

You are also essentially no different from an abacus friend. Consciousness is a meme

>conscious: aware of and responding to one's surroundings; awake.

Computer programs only respond to a combination specified input once a model has been approximated, regardless of individual parameters and what other environment they have access to.

A chess playing program can see the board but it can't purposefully move it's piece into a wrong spot just because, or to respond to the board alone. It can only move into the legal areas (and among those legal areas, it optimizes to move to the optimal spot).
Another example is a program having access to numerous environmental things concerning the system it's running on. Again, you don't see programs exploring these options and finding solutions like cutting Internet connection to brute force wins against players, due to there being explicit algorithmic programming to limit their ability to actually be fully awake and responsive. We don't have any programs as far as I'm aware that are fully awake and allowed to explore and respond to their environment or individual subsets of their parameters beyond the training stage of model building.

IQ test were invented to find retards to filter in European eugenics you stupid fucking nigger.

We also can't visualize 4 dimensional objects, despite having very logical and understandable mathematical models of what they would be like, no matter how hard you try you can't imagine a 4 dimensional object, no not it's 3 dimensional projection, not its net, but the way it would actually look in 4 dimensions.

Because, just like that chess-playing program, it can't disconnect for an easy win because its internet connection was never part of it's environment so none of it's strategies can involve it. Since 4 dimensional objects have never existed in our environment, we never evolved to think in that way, our visual cortex is literally algorithmically incapable of acting in that fashion.

A chess program can't do anything beyond the rules of chess. Humans can't do anything beyond the laws of physics. It's a matter of degree only.

>IQ isn't a great measure of success/intelligence
The data says you're wrong

>What does this imply about the true relationship between "intelligence" and IQ scores?

Intelligence can't not be quantified.

listen my friend, not saying iq tests are complete bullshit, but the euros who made those eugenics programs weren't very good at their jobs

as it's been shown itt, iq scores can be improved both by training the types of questions directly and indirectly by being employed/educated with things that are similar to what the tests are aiming for
iq is a measure of academic success because you actually get better at solving the tests when having a math-y eduaction
of course it's a shitty measure then

>iq is a measure of academic success because you actually get better at solving the tests when having a math-y eduaction
But when given to young children with very little math education, it gives very good predictions about their future academic achievement and economic success.

FYI, real IQ tests are nothing like the ones you see online.

We can however respond to individual stimuli in this universe and are not bound to some function which only allows us to make our next move as a combination of input parameters. We can continue to operate after losing one of our senses (input paramters), and can even continue to think and process when we lose a combination of them. A program which is a mathematical model would crash if you expected it to continue without giving it all the needed parameters.
We truly respond to the environment, while many programs are hard models that respond to explicit combinations of prespecified input ranges.

>and are not bound to some function which only allows us to make our next move as a combination of input parameters

muh quantum consciousness

No, we literally are not some fixed function of
Y=sight*weight+hearing*weight+feeling*weight+taste*weight+smell*weight
That would crash if one of the input parameters was missing. Blind and deaf people can lead productive lives, and a person's brain doesn't crash if they lose a sense.

It doesnt imply anything because your frame of reference, questions and understandings are all incorrect or incoherent.

It's really easy to make a program that works on mostly optional parameters (so it can continue to function even if it lost some of its "senses"), it's harder, but still possible to make it work ENTIRELY on optional parameters, but then things get weird since it has no idea what sort of effect its actions are causing, similarly for humans.

just stop posting lad

...you understand what neural nets and machine learning models construct, right?

8 or 2, not sure
Either the double or half of 4

you understand what "we are not bound to some function which only allows us to make our next move as a combination of input parameters" is implying, right?

You seem to think that they're significantly more crash and error-prone than they actually are.

However, unlike a computer, a person with no senses can still respond to its environment. They can still think in the void, and they can still output. They can still have food put in their stomachs and digest it. We could likely put a neural cap on them and monitor signal firings. They can scream endlessly, and we can all hear it, though they wouldn't know we can all hear it.
A machine learning model, which is simply a mathematical fomula which expects to continously give solutions based on input, would crash and be nonresponsive. We explicitly program the things to solve problems, and they don't fully respond to or act in their environment. They just pop into existence, do work, and die. They aren't conscious. Maybe if we programmed one that was always capable of output into the environment, despite having variable input access, and continued to backpropagate infinitely in the absence of new information.

We literally aren't explicitly bound to input only. Machines as we use them now are. No chess playing or route solving program is conscious as a result. They could be one day but they aren't right now.
How so? You will get nothing out of a statistical learning model if you give it no training input, or no further input to test after training.
Meanwhile I can theoretically have a baby born with no senses that just screams and cries, and has a functioning brain. We don't have machines that can interact with their environment completely blind and additionally their response to their environment is a fixed equation which is a sum of weighted input, instead of being able to truly interact with elements on a whim at all times. It's a fixed model.

Really, these models/machines are more like the ideas we have after we interact with the world, rather than an actual constantly developing consciousness.

Nigga, what did I just say? Optional parameters mean it can be there or it can not be there, the program still runs, it doesn't crash. Humans are probably closer to all optional parameters, but you need at least a few to function at least sort of normally, and again, you could easily make a program like this.

And no, a person with no senses can't respond to its environment. Constantly screaming isn't responding to its environment, that's just screaming regardless of what's going on, digestion happens irrespective of what else is going on (provided you have a stomach). A response means acting based on a stimuli, an input.

>We literally aren't explicitly bound to input only.

>what are the laws of physics

Unless you are suggesting consciousness is a result of non-deterministic quantum processes, then we are absolutely a function of our inputs.

>How so? You will get nothing out of a statistical learning model if you give it no training input, or no further input to test after training.
Nice moving the goalposts. Now that you've suddenly agreed that it wouldn't crash we can move on.

A baby screaming and crying isn't interacting with its environment, its just constantly creating an output. And yes, we have machines that can interact with their environment blind, and no, the whole point of neural networks is that they don't have "fixed" equations, they're completely flexible, with the freedom to change any thing in how it acts in order to better optimize its fitness.

There are computer programs that can run with no input. Or at least, a lack of input doesn't cause them to crash. Maybe they just halt immediately for example.

Also idk if you know this but blind and deaf people do face some difficulties due to the lack of input

Doesn't say much at all. Still just weak-ai bullshittery.

it's the double
the upper slice is the lower one in terms of numbers
the lower slice is the doubled one

You're assuming we're merely conscious androids, lacking free will. Given how thoroughly unexplained qualia are don't you think that conclusions a bit hasty?

Sure there's a constant correlation between qualia and neural (even atomic) activity, but what in physics then goes on to explain a one directional causal arrow between the physical brain and qualia?

It's not unreasonable to hold your breath for a model beyond classical and quantum.

>qualia implies free will
???????

If you're argument relies on free will to begin with you should just give up now. Go play with your popsci faggot friends and jerk off about how we've all got magic in our heads that makes us special.

faggots STOP arguing about consciousness
STOP, i've programmed a computer to play chess, it's not very smart at all, but that's irrelevant
main topic of the thread is iq tests and how they are a little useful, but also flawed

ok we are back on track

>it's not very smart at all
because your program probably sucked

Premature friend

here's how a comp plays chess, in broad terms
for each move, check all possible outcomes to a certain depth, pick the move that maximizes how many pieces you have
it's literally one routine + added heuristics of which alpha-beta pruning is enough
this simple algorithm of search + big amounts of parallelism is how a computer beats a chess grandmaster

i'm kind of a softie, so i get sad when i think of the sorry state of ai and machine learning or at least the algorithms i've learnt about

Free will goes against all current understanding of physics.
Are you claiming physics is wrong? If so provide your evidence.

>t. Doesn't know what's under the hood of heuristic and learning algorithms

If consciousness and free will aren't relevant, then there's the answer to op, the northwestern network model implies nothing about the relationship between intelligence and iq because the only difference between the model and a human brain is complexity and manufacturer.

I'm holding my breath.

Are you really so arrogant as to think we've figured out the fundamentals of reality to an extent that we can close the book on this? Then provide evidence that in any way explains how qualia are caused.

>the northwestern network model implies nothing about the relationship between intelligence and iq because the only difference between the model and a human brain is complexity and manufacturer.
That's right, the human brain is orders of magnitudes more complex than some matlab or python script, and yet the latter has a higher measured IQ than the former, isn't that strange?

you're a fucking retard