293 karma · joined November 2, 2015
Or you could just take 1/10th the luck and time and energy and apply it to Archive Team projects and get 10x the results...
Especially considering that the population average in many places and times is far below the current population averages normed in the USA and UK and Western Europe. If you asked how many people throughout history or globally would be in the top 1% of the USA/UK/WE population, the answer will be much smaller than a billion. (Thin tails strike again.)
Probably they mean youngest legal human, or youngest post-birth vitrification.
This is something I've suspected for a while: the Hutter Prize has become obsolete. I agree with the 'prediction is intelligence' paradigm, but the Hutter Prize cannot show any meaningful progress in AI for the simple reason that the corpus is too small and the resource limits are too stringent. 4 layers with 414k neurons is nowhere near what deep learning researchers use to create intelligent systems. You couldn't fit char-rnn into the Hutter Prize constraints! At this point, all the prize incentivizes is finding an algorithm which fits into the allotted resources, no matter how dumb and inferior that algorithm is to the state of the art which requires more resources to run. Bit-twiddling to optimize the current approaches may be fun for a certain kind of people, but it's hard to see how the contest has contributed at all to progress in AI in the past decade, which was the original purpose.
Perhaps what is necessary is a Hutter Prize 2: expand the corpus to the modern English Wikipedia (not one from a decade ago), perhaps add in other language corpuses to increase it further to get a multi-gigabyte corpus, and then increase the resource limits drastically (32GB RAM for starters so cmix can get in), and shut down the old one for diminishing returns. A desktop machine these days can certainly support a decent amount of RAM and computation (on Newegg, you can get a desktop tower refurb with 32GB Ram for <$1000), so there's no cost excuse if one can afford to offer thousands of euros as a prize...
(That is: even if parapsychology was 'right' in the sense that something like these electromagnetic waves did occasionally affect cognition and could be used to transmit information in limited circumstances, it was still wrong overall, falling for all sorts of things like Geller's spoon bending, successfully managed to produce many false positives despite the most stringent measures, and can be considered a control group for the scientific method.)
Yes, this is the critical piece. The results of the Reproducibility Project were not remotely a surprise to Bayesian observers. People like Gelman have been pointing out for ages (and I mean back to the 1960s) that the prior probabilities in these fields is low and necessarily a lot of the results were false positives. With the rise of meta-analyses, it is possible to have informative priors for particular fields of psychology or for psychology as a whole, which would let you make much better predictions about whether a result was real. But you can't use these in papers - authors are heavily biased towards using procedures or flat priors which are uninterpretable or grossly overestimate the evidence, and if you try to use any of the informative priors or more advanced models, they'll nag you to death with a thousand objections and complain about double standards and subjectivity and how this time is different and (ironically) bias. So for the most part, there's not much to gain in academic research.
But in a prediction market, you don't have to listen to the self-serving excuses or explain your reasoning, and there's something to make it worth your while.
Then you can simulate the results of future experiments with _n_ datapoints: sample 1 effect size from that distribution, generate _n_ datapoints assuming that effect size, run the hypothesis-testing, and return the _p_-value.
The fraction of _p_<=0.05 is your best forecast of whether the future experiment will succeed in reproducing it or not.
Sorry, I left out that part of Kelly's estimate of 1b songs is that album rates will also continue to increase at the historical rate. It seems improbable to me that we could really hit 1.1b songs by 2060 (it's such a mindboggling number), but on the other hand, the world keeps getting wealthier, making music keeps getting cheaper, and the appetite and dreams of people for making music seems to know no limits...
Isn't it amazing? And about that rate has been sustained for decades, as well. Can you imagine hundreds of thousands of books being published every year for decades? Who's writing them? How is this economicly sustainable? (If it's not, do we really need to incentivize publication with ever longer copyright?) This is why when Google Books tried to estimate how many books had ever been published, they got a lower bound of 129 million and no real idea how much off they might be.
And it's not just books, either. There's maybe 100,000 albums being released every year, which means in our lifetime we'll hit 1 billion+. The number of feature films keeps going up (4000 a year not terribly long ago). The TV industry is feeling shellshocked by just how many programs are in production now. And all these numbers seem like they must be undercounts because who is really able to track how many new songs get uploaded to Youtube every day by professionals, amateurs, and all sorts of indescribable people? And in the rest of the world like China?
It leads to significantly stronger animals. That's not useless, for either prey or predators.
> Human intelligence is an emergent trait like speed determined by the co-ordination of many sub-systems.
We don't know what human intelligence is on a neurological basis. Imaging studies are linking it to a number of things, but it's still preliminary and so it's premature to baldly assert that there's no possible single mutation which might help.
> As an aside I remember reading a paper from long ago that suggested that high intelligence was the result of the relatively absence of mutant alleles at the various intelligence loci.
Mutation load hasn't been found to be strongly correlated with intelligence in studies which have looked at it directly, and the GCTAs already upper bound any such effect: "The total burden of rare, non-synonymous exome genetic variants is not associated with childhood or late-life cognitive ability", Marioni et al 2014 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3953855/ "A genome-wide analysis of putative functional and exonic variation associated with extremely high intelligence" http://www.nature.com/mp/journal/vaop/ncurrent/full/mp201510... , Spain et al 2015
They have? I'm not aware of any GWASes targeting intelligence using Ashkenazi samples.
Not necessarily. This is something you might expect if intelligence is a net fitness advantage and it is in a mutation load situation where rare variants need to be purged to keep things constant. But if intelligence is only worthwhile up to a certain point and it is controlled by frequency-dependent selection, or there is heterozygote advantage, or if intelligence is not necessarily reproductively fit at all, or other situations, then there could certainly be rare variants of large positive effect. (I would not have bet on their existence for many reasons, but not because it's impossible.)
To give an example, your claim would predict that there is no such thing as a single mutation which increases muscle mass a lot because it's a complex system affected by a lot of genes; yet nevertheless, there is a single mutation affecting myostatin which makes humans and pigs and dogs much more muscular, and it's even been edited into pigs with CRISPR this year and last year into sheep and cows. Presumably the reason that not all animals are ultra-strong thanks to the mutation is that it causes birthing difficulties and increases metabolic demands considerably, and so despite the obvious advantages of being ultra-strong, it's not actually fit.
To actually try to explain the design here:
we think intelligence is caused by thousands of common genes, each of which slightly helps or hurts. When you have thousands of independent genes, then they add up to a normal distribution like the one we see when we measure a lot of people's intelligence. The exception is that there are too many retarded people; if the normal distribution was the whole story, then retarded people would be as rare as geniuses, but they are much more common than that. We explain there being too many retarded people by saying that they have one or a few very rare mutations, mutations which are very harmful. But another theory might be that the retarded have very few normal intelligence genes, or that intelligence genes don't simply add up but interact in complex ways, or that they have very harmful environments. How do we decide which theory is right? Well, we look at the retarded people vs their siblings; if their parents were lacking a lot of good genes, or they were being raised in a toxic waste dump, then we would expect the siblings to also be near-retarded themselves since they also inherited few good genes or are affected by the toxic waste. But they're not; they are almost average! This is more consistent with there being one bad mutation and the retarded sibling had bad luck than the other theories. This theory has since been confirmed by finding hundreds of unique and harmful mutations in retarded people. So we conclude that the effect of intelligence genes is indeed a normal distribution, with an occasional bad mutation overriding that and making someone retarded.
This immediately raises a question. If we admit that on the low end intelligence may be controlled by a single rare mutation, why not on the high end too? Maybe there are special genius genes floating around. This would be important because it means that you can't make much progress by just looking at the genes of regular people, and it also means that SNP studies will be extremely limited in what they can find. How can we check this? We can do the same thing as on the low end: if there is a special genius gene, then geniuses will have much higher IQ than their siblings do, who will be close to average; but if their parents have lots of good genes and there is no single special gene, then the siblings will also be well above average and similar to their genius sibling.
Using a very large set of siblings and twins, OP finds that very intelligent twins/teens are similar to their siblings. So this is the opposite of the retardation findings. There are rare genes for retardation, but there are not rare genes for genius.
The delayed amount was not a fixed amount but calibrated to another subject's total sleep, so you can't say exactly how much each one was delayed by overall:
"RSO During Nights 2–4 of RSO, subjects’ total sleep opportu- nity period was restricted and yoked to the amount of total sleep time achieved by a matched subject in the FA group. This was accomplished by delaying the RSO subject’s bedtime and keeping a fixed wake time. For example, if an FA subject achieved 210 min of total sleep time on Night 2, the yoked RSO subject would be provided a 210-min opportunity for undisturbed sleep (bedtime 03:30, wake time 07:00) on Night 2. Subjects in the RSO condition were monitored polysomno- graphically for an entire 8-h period."