3 ms·

> The challenge is the human proteins needed to make each of these to happen can conflict with each other. The proteins — created by the genes — needed for risk

by gwern 3y ago

> The challenge is the human proteins needed to make each of these to happen can conflict with each other. The proteins — created by the genes — needed for risk taking may be the ones that dampen good listening. There are genetic trade offs, in that you cannot optimize all traits. In other words there are genetic costs for each trait. To raise IQ might cost lowering something else, such as empathy. It’s not that there is a zero-sum quantity being conserved, it’s that genes cannot do all that is possible. They are hugely constrained by each other. It is like trying to design a machine: it cannot optimize all properties; it cannot be fastest, lightest, strongest, and cheapest at the same time. Everything is a trade off. Performance, reliability, speed, cost — all are trade off between them. Genes and traits also operate under the same regime of trade offs. There are no free lunches. Untrue.

Purplehermann 3y ago
While the author's framing is wrong, aren't there genes that have positive dominant effects and negative recessive effects?
inglor_cz 3y ago
"Untrue." Can you be more specific? This kind of smug response is worthy of Twitter, incapable of starting any deeper conversation on a pretty deep topic.
- Purplehermann 3y ago
  When things are just.. wrong, there's not much reason to be specific. Especially when gwern.net has multiple long-form blog posts discussing these topics
- gwern 3y ago
  Kelly is making extremely sweeping broad claims based on no evidence and flimsy analogies which are obviously wrong - humans are not carefully-engineered cars competing in a free market where cars could exist on a Pareto frontier (which itself would be a rather questionable claim...), they are organisms under a bombardment of mutations constantly happening and those mutations being very slowly purged by natural selection over thousands of generations. It is not surprising that there could be 'free lunches', anymore than it is surprising that there might be typos left in a file after a cat walked across your keyboard and you tried to fix it. I don't need to provide any references because Kelly didn't provide any of his own: what is presented without evidence can be dismissed without evidence. I shouldn't be held to a higher standard than he is simply because he said his claims first. He's not a geneticist or psychologist or otherwise working in this area, so there shouldn't be even a presumption here that he is probably right if he isn't going to provide any evidence or sources or more convincing arguments than 'there just can't be free lunches, there can't!'. I am being rude here because I am fed up with decades of people making these sweeping assertions about genetic correlations while simply completely ignoring the large literature actually measuring these FUD bugaboos. Kelly, and everyone making those arguments, deserves nothing but mockery until they show any sign of catching up to the literature as of, let's say, 1990 (https://en.wikipedia.org/w/index.php?title=Genetic_correlation&oldid=863326253#Intelligence https://en.wikipedia.org/w/index.php?title=Genetic_correlati...), and while we're at it, I could wish that they would understand the idea of 'adding two variables before doing selection' (I mean, index selection in genetics only goes back to the 1940s or so, you can't expect people to understand such cutting-edge concepts). And he can't provide those references because it's not true: the genetic correlations between IQ and other traits are pervasively positive, and Kelly would know that if he spent any time looking into it. In fact, you would have to go thoroughly through the list of known genetic correlations with IQ before you would find any that are bad, contrary to his assertion that there is always a strong tradeoff in a myriad of ways. (I'll save you the effort: out of hundreds of reasonably reliable genetic correlations of IQ/gf with other traits like diseases, the bad ones worth mentioning are myopia, possibly allergies (I forget if that was just phenotypic), anorexia, and autism spectrum disorder checklist self-reports. And I think the last one is probably bogus anyway due to measurement issues.)
  - inglor_cz 3y ago
    Thank you. I agree that the author acts recklessly and presents unwarranted conclusions. That said, in case of widespread genetic engineering, I would expect potential parents to make more specific choices that will result in some real tradeoffs. For example, it seems that longevity and fertility are somewhat negatively correlated, though this is precisely a field of study that requires a lot more research, because longevity was, until recently, a very underfunded topic. Most people, if they had a choice, would probably opt for longevity, given that you can now "outsource" reproduction to IVF and similar techniques. But over several generations, any small effects may be amplified.
flobosg 3y ago
I would say it is partially true. Expression of a gene (or even changes in its expression levels) may affect detrimentally the product of other genes. But, on the other hand, these changes can be compensated (for instance) by “beneficial” epistatic interactions coming from additional genes, allowing the individual to keep its phenotype. Biology is tangled and messy, after all.

by gwern 3y ago

Purplehermann 3y ago

While the author's framing is wrong, aren't there genes that have positive dominant effects and negative recessive effects?

inglor_cz 3y ago

"Untrue." Can you be more specific? This kind of smug response is worthy of Twitter, incapable of starting any deeper conversation on a pretty deep topic.

Purplehermann 3y ago
When things are just.. wrong, there's not much reason to be specific. Especially when gwern.net has multiple long-form blog posts discussing these topics
gwern 3y ago
Kelly is making extremely sweeping broad claims based on no evidence and flimsy analogies which are obviously wrong - humans are not carefully-engineered cars competing in a free market where cars could exist on a Pareto frontier (which itself would be a rather questionable claim...), they are organisms under a bombardment of mutations constantly happening and those mutations being very slowly purged by natural selection over thousands of generations. It is not surprising that there could be 'free lunches', anymore than it is surprising that there might be typos left in a file after a cat walked across your keyboard and you tried to fix it. I don't need to provide any references because Kelly didn't provide any of his own: what is presented without evidence can be dismissed without evidence. I shouldn't be held to a higher standard than he is simply because he said his claims first. He's not a geneticist or psychologist or otherwise working in this area, so there shouldn't be even a presumption here that he is probably right if he isn't going to provide any evidence or sources or more convincing arguments than 'there just can't be free lunches, there can't!'. I am being rude here because I am fed up with decades of people making these sweeping assertions about genetic correlations while simply completely ignoring the large literature actually measuring these FUD bugaboos. Kelly, and everyone making those arguments, deserves nothing but mockery until they show any sign of catching up to the literature as of, let's say, 1990 (https://en.wikipedia.org/w/index.php?title=Genetic_correlation&oldid=863326253#Intelligence https://en.wikipedia.org/w/index.php?title=Genetic_correlati...), and while we're at it, I could wish that they would understand the idea of 'adding two variables before doing selection' (I mean, index selection in genetics only goes back to the 1940s or so, you can't expect people to understand such cutting-edge concepts). And he can't provide those references because it's not true: the genetic correlations between IQ and other traits are pervasively positive, and Kelly would know that if he spent any time looking into it. In fact, you would have to go thoroughly through the list of known genetic correlations with IQ before you would find any that are bad, contrary to his assertion that there is always a strong tradeoff in a myriad of ways. (I'll save you the effort: out of hundreds of reasonably reliable genetic correlations of IQ/gf with other traits like diseases, the bad ones worth mentioning are myopia, possibly allergies (I forget if that was just phenotypic), anorexia, and autism spectrum disorder checklist self-reports. And I think the last one is probably bogus anyway due to measurement issues.)
- inglor_cz 3y ago
  Thank you. I agree that the author acts recklessly and presents unwarranted conclusions. That said, in case of widespread genetic engineering, I would expect potential parents to make more specific choices that will result in some real tradeoffs. For example, it seems that longevity and fertility are somewhat negatively correlated, though this is precisely a field of study that requires a lot more research, because longevity was, until recently, a very underfunded topic. Most people, if they had a choice, would probably opt for longevity, given that you can now "outsource" reproduction to IVF and similar techniques. But over several generations, any small effects may be amplified.

flobosg 3y ago

I would say it is partially true. Expression of a gene (or even changes in its expression levels) may affect detrimentally the product of other genes. But, on the other hand, these changes can be compensated (for instance) by “beneficial” epistatic interactions coming from additional genes, allowing the individual to keep its phenotype. Biology is tangled and messy, after all.