Thursday, August 25, 2011

Video of Google talk on cognitive genomics



This is video of the talk I gave at Google.

I haven't watched it yet -- I'm worried that the audio is a bit patchy because I kept stepping away from the microphone on the podium. Probably not one of my best performances, but I think I got the main ideas across :-)

25 comments:

TheGuyFromEarlier said...

Well presented and easy for laymen to digest.  Thanks.

James D Miller said...

The audio is pretty good except that you can't hear the questions being asked.
Have you considered asking for data from 23andme customers who have high IQs?  Perhaps you could lower your IQ standards a bit for this group.

MtMoru said...

You might have gone further than calling PCA reasonable. 

If one is determined to have a single number for intelligence and n-vectors really do have a multivariate normal or elliptical pdf for a given population then if one knows nothing about an individual other than that he belongs to that population PCA gives THE BEST single number.

But of course it musn't be interpreted, as psychometricians are wont to do, as an intrinsic property of the individual but rather the mean of the pdf giving probabilities for the individual's n-vector when that n-vector is unknown and the only other thing one knows about the individual is the population to which he belongs.

Sam H said...

What % of people in the audience at Google would quality? The guy who introduces you kinda assumes everyone could qualify. :-) 

Bobdisqus said...

Hello Steve
Why only the right tail? Isn’t there just as much to be learned by looking at the left tail as well?

MtMoru said...

He might be afraid that the left has too many pathological cases. The right has none. The smart are mentally and physically healthier than the rest. The height analogy falls apart here as both giantism and dwarfism are often symptoms of pathology.

But if only the control group were replaced with healthy stupid people BGI would have an easier job. What fraction of <=55 IQs are  symptoms of pathology?

Yan Shen said...

On the other hand, finding a bunch of idiots isn't that hard to do. Just go to Stormfront or any one of these other pro-white blogs.

ben_g said...

Posted some notes on the BGI IQ study at gnxp: http://www.gnxp.com/wp/2011/08/26/thoughts-on-the-bgi-iq-study/

MtMoru said...

BTW, Adderall IS speed.

Christopher Chang said...

He might be afraid that the left has too many pathological cases.

That is correct.  Mutations of large negative effect are a lot more common than the other direction.

Christopher Chang said...

Before you dismiss a group for perceived inferiority, you should make sure you can present their arguments at least as well as they can.  I don't think you are currently capable of this.

Yan Shen said...

I criticized neo-Nazis for being idiots. I never criticized any racial or ethnic group. Why you would defend the neo-Nazi crowd at Stormfront/Mangans is beyond me. I'm far from the only person who thinks that they're idiots. In fact the mainstream opinion probably agrees with me.

Yan Shen said...

Chris, I completely agree with you. Thank you for teaching me this important life lesson.

esmith said...

The left tail is fairly well studied already. And that's precisely the problem - people end up at the left tail because of genetic malfunctions (something that does not work the way it should). Pick any major autosomal abnormality:

http://en.wikipedia.org/wiki/Chromosome_abnormalities

Almost each one will have "mental retardation" or "developmental delays" among symptoms.

And often there are intra-uterine / early childhood environmental conditions. Something as simple as 2 minutes without oxygen during birth, or living in a house where walls are painted with lead paint, can land you at the left tail.

botti said...

You present really well Steve, nice work.

MtMoru said...

I don't think you are currently capable of this.

Nobody expects the po-mo po-po.

But technically you're right, because they don't have any arguments. Do I have to learn artspeak before I can dismiss contemporary art as bullshit?

mm m said...

Great work.  I am all for conceptual simplicity, but I do
not understand why a program that weights variants according to maximum
predictive power is not value-added and within our capabilities with computers.

Ene Dene said...

Great lecture.

I have 2 questions.

1.What's the percentage of male compared to percentage of female above 3SD?
As I understand men have a wider curve (more extra smart/stupid)

2.If there is a difference, should the 0.6*nSD be different for men and female parents/kids?
For example if woman has 160 and man has 140, should not be the same as if we had 140 woman and 160 man. And it should be different for woman/man offspring. That is if having IQ>145 is much more common for men, than woman who has it should have "better" genetic material than man who has it and that should reflect the offspring. And if it's more probable to have a more male offspring above 145, than the 0.6 factor should be different for male and female offspring.

MtMoru said...

When the mother is Jewish are the children smarter than when the father is Jewish?

Ene Dene said...

Cases:
1. smarter mother, IQ of female offspring?
2. smarter mother, IQ of male offspring?
3. smarter father, IQ of female offspring?
4. smarter father, IQ of male offspring?

steve hsu said...

The male SD is perhaps 10% greater than the female SD. You can correct accordingly.

esmith said...

This is a rather curious statement if you look at it from genomics perspective. It requires one of two basic premises:

- Either there is a gene or a group of genes which reside in the Y chromosome, which contribute around 10% of all variation in intelligence in males (seems improbable, since there's no a priori reason to think that males and females end up with the same value of mean. Or do they? Perhaps they don't, but our tests aren't sensitive enough.)

- Or the entire 10% increase in SD among males is environmental / cultural.

Ene Dene said...

They end up the same because IQ tests are set up to give the same mean for male and female. It's well known that female are better at some types of tasks for example verbal test, while men are better at other, for example manipulation of objects in space. You could easily make a test that would favor female (higher mean) by putting more tasks that female are better at. I don't remember the source, but I've read that first IQ tests gave women 10 points more on average, then they've adopted it to give the same number, but they can't adopt for variation.
All psychometric tests have wider curve for men.

It could be environmental/cultural but it would be interesting how it makes a contribution on both ends of distribution (there are more stupid male as well). Is dyslexia environmental/cultural in origin? 5x more men are prone to dyslexia than female. Men and female brains and bodies differ physically, why wouldn't there be differences in variation?

LaurentMelchiorTellier said...

Nicely done, Steve. Pleasant reactions and questions, too!

I suppose the diverse comments on youtube are a preview of "normal" audience reactions. But that's inevitable.

Luke Nguyen said...

Great talk professor.  Looking forward to seeing the results of the study.  

Blog Archive

Labels