[CuriouslyC]

My personal theory on this is twofold:

1.) High achievers are used to pushing at things to win. Creativity is like a garden - you have to do some tending, but then you have to back off and let the plants grow. Fiddling doesn't help, in fact you're liable to end up killing the plants.

2.) Smart people actually learn too fast. Creativity requires very broad neural connectivity, and I think fast learning tends to produce neural networks with sparse connectivity to different areas. This is supported to a degree by learning in artificial neural networks. When the learning rate in artificial neural networks is too fast, this can cause the network parameters to converge prematurely. This premature convergence typically results in poor generalization performance. It is also worth noting that human brains mature more slowly than those of previous hominids and great apes.

[nostrebored]

Using neural networks as a machine learning concept to explain actual concepts in Neuroscience is a huge overstep. People have always thought the most advanced technology can be analogous to cognition, see steam engines and clocks.

[CuriouslyC]

Artificial neural networks are much closer to neurons and brains than economic models are to markets, but we give economists a pass. Perhaps we should stop listening to them and revoke their publishing privileges?

Anyhow, the point wasn't to say Brains == ANN, rather it was to show how learning too quickly has consequences in other models of learning. This was mostly so people wouldn't just dismiss the idea out of hand as having no justification (which people seem to want to do regardless).

[nostrebored]

Really? I think that we have a much better understanding of markets than Neuroscience. The human brain is an immensely complex system. We just figured out that lungs may produce the majority of our blood cells -- do you really have so much faith in a field that routinely misrepresents results to fit their hypotheses?

[OisinMoran]

Your second sentence is definitely true and very interesting but maybe you'd agree that the gap between our most advanced technology and the brain is decreasing to the point where this will eventually be right.

[princeb]

> gap between our most advanced technology and the brain is decreasing

first we need to understand the brain.

[psyc]

And those were progressively better analogies. These are great terms to think in, and it's not an overstep. It doesn't have to be absolutely true to be very useful for understanding.

[Retric]

2) This falls into the just world fallacy. Intelligence does not seem to come with real drawbacks. Highly intelligent people often have upper end reflexes, are unusual attractive, highly creative, social, and or content. They are more often taller and live longer as well.

The trappings of intelligence are different, but very high raw intelligence is seen across most walks of life including actors, athletes, and salesmen.

[CuriouslyC]

You might not believe intelligence comes with any drawbacks, but when those neurons are wired well to perform one task, they're by definition wired poorly to perform other tasks. There is no universal neuron wiring that does a great job at everything. Barring things like malnutrition or lack of environmental enrichment, being bad at IQ tests just means your neurons wired differently. There is almost certainly something lower IQ people are better at. Take autistic savants for instance - their abilities arise because their brains are wired very differently from normal. They're not functional in a lot of ways, but there are some things they are freakishly good at.

As for your uber race, maybe you want to use the word "talent" instead. The only thing that connect these highly talented peopled is that they seem to excel at pretty much everything they put their minds to. If you tested them you'd find that their IQs have a fairly broad range. The median would certainly be above average but that isn't what makes them different. It is more about learning to focus and apply yourself than having a knack for manipulating symbols in your head.

[Retric]

IQ tests are proxies for intelegence but not direct measurements. They are fairly good at that from around 60-140, and less useful at the top and bottom.

As to low IQ scores people with an IQ of 60 generally don't have anything that they preform very well. The problem is everything we think of as takes even something as simple as holding your hand steady is really complicated. Getting better at even simple motor functions takes feedback loops, effort, and optimization. Further, intelegence generally allows people to optimize more quickly even as an infant. Which means intelegent people not only get to optimize more quickly they have more time to optimize complex tasks or tasks they are worse at.

Specialization of course kicks in, but the sills a six year old has are fairly universal and things 98% of people can do. Intelegent people tend to be better at.

[CuriouslyC]

An IQ of 60 isn't difference, it's indicative of either malnutrition, severe neglect, or some form of developmental disorder. In fact, I'd be surprised if there are many people of below average intelligence that don't suffer from one of those things. I'd even wager that if you took any random infant that didn't have a developmental problem and raised them in a loving affluent western family with good food and an enriched environment odds are good their IQ would measure >110 after a few years.

[Retric]

That's your own bias talking, genetics and environmental toxins can both cause significant issues. But, so can non genetic birth defects resulting from disease or what amounts to bad luck.

Further sub 110 IQ's don't indicate a problem. Natural variation is simply much wider than you assume.

[j9461701]

This is only true up to a point. Past ~125 IQ, you start running into serious social problems that mean such individuals are much less likely to obtain leadership positions, join intellectually challenging professions, or obtain advanced degrees.

https://polymatharchives.blogspot.ca/2015/01/the-inappropria...

http://psycnet.apa.org/psycinfo/2017-14279-001/

[Retric]

This get's into questions of what is an is not a good proxy for actual intelligence. IQ tests are more likely to become arbitrary rankings once you looking at ever smaller population sizes and thus relative poor proxy's for g.

Still, Roughly one third to one half of the billionaires (45.0%), Fortune 500 CEOs (38.6%), Senators (41.0%), and federal judges (40.9%) attended a school requiring standardized test scores that likely places them in the top 1% of ability. https://www.psychologytoday.com/files/attachments/56143/wai-... Further, why that's only a proxy for intelligence, those with different backgrounds may also have unusually high IQ's.

Granted we are now using a proxy for g to estimate IQ. But, as I was saying intelligence not scores on IQ test IMO it's a valid approach.

[fjdlwlv]

Obviously if you define intelligence as "successful" than intelligence correlates with success. That's a circular void claim.

[Retric]

I am taking about different types of success. Otherwise you can just define a specific IQ test as success.

My point is if you look at Intelligence and assume they should be more successful with more IQ you need validate the IQ test to verify it's still valid at the outer edges. But, if you look at the most successful people and find IQ > 140 to be vastly more common than the general population then intelegence unlikely to be actually harmful.

[ziikutv]

I would also like to point out that while there might not be any direct drawbacks, think about others. How many kids put down those smarter to feel better about themselves?

Another disagreement: Not all intelligent people are highly social. Many are introverted.

Intelligent or introverted kids get bullied the most.

[Retric]

Your assuming kids are accurately measuring intelegence and then picking on the smartest person. Instead they are picking on an outlier who may be less intelegent than other people in the population with different interests.

[ziikutv]

Fair point.

[hueving]

>are unusual attractive

Source? I've never heard of intelligence being correlated with visual attractiveness.

[Retric]

Like height the link is probably though health and generally freedom from birth defects. It's not a 1:1 correlation, but it is a positive one.

Consistent with such views, meta-analyses (Jackson, Hunter & Hodge, 1995; Langlois et al., 2000) show that there is a small but significantly positive correlation between intelligence and physical attractiveness https://personal.lse.ac.uk/kanazawa/pdfs/I2011.pdf

That said, selected populations may show a negative correlation such as when more attractive people are more easily given a job etc.

[hueving]

>That said, selected populations may show a negative correlation such as when more attractive people are more easily given a job etc.

How is that possible? If intelligent people are attractive, wouldn't they be the ones getting the jobs?

[fjdlwlv]

Retric has made many posts conflating above-average intelligence with the extremely high intelligence that the article talks about.

[walshemj]

As a hi IQ Dyslexic id have to disagree

[Retric]

People of average or below average intelegence can also be Dyslexic, so what's you point?

[deepnotderp]

Repeat after me:

Artificial Neural Networks/DL =! Biological Neural Networks

[CuriouslyC]

You'd have to be pretty ignorant to assume they are exactly analogous, but to assume that there is no transfer is equally ignorant. Artificial neural networks are much closer to neurons and brains than economic models are to markets, yet we bet the farm on forecasting (with pretty bad results, I might add).

The repeat after me meme is really condescending by the way, might want to avoid using that.

[deepnotderp]

What are the analogous components? Backpropagation is biologically implausible.

[CuriouslyC]

Dopamine released from task success primes long term potentiation, effectively acting as the analog of the cost function's gradient. The larger the dopamine hit (generally) the bigger the learning step.

You are correct that neurons can't do backprop. Keep in mind that the networks of the brain aren't straight feedforward, they're recurrent. In order to provide temporal control of activation propagation throughout the network, inhibitory neurons are needed. Thus, instead of "tweaking weights" backwards in the network, the brain learns to activate inhibitory neurons to provide forward feedback against activation.The mechanism is different but the learning effect is pretty similar.