very useful - I find a lot of audio SR (compression) algos to sound really bad - likely just because of the loss functions and/or eval metrics are 'inhuman'.
this is false - despecialized intentionally puts back original artifacts that were "cleaned up" by later editions (including garbage mattes, VFX errors etc).
The way to think about carbon offsets is to simplify the problem as analogous to "excess mortality" - if something doesn't actually increase excess mortality, it's not as bad as you think it is (and vice versa).
The same applies to carbon offsets: if their use produced an excess (or dearth) of CO2, we would notice.
that doesn't make sense. you keep the excess mortality and then pay a con company for mortality offsets which then gives you lower tax on your mortality bound business.
Models are hard and causation near impossible, but feel pretty safe being on the "arrow of causality pointing from lead to low IQ" rather than the hilarious inverse...
1. Make up some scientific "evidence" that exposure to some common, harmless substance is actually harmful.
2. People believe it. Smart people make an effort to avoid it. Dumb people don't care.
3. Measure it 30 years later, and now there's a correlation between higher IQ and lower exposure to the substance. Higher IQ causes lower exposure to the substance.
I'm not saying this is the case with Pb, it's just a funny idea.
There are longitudinal studies that track people from childhood before the dangers of lead were widely known. There are also comparative studies where populations are studied with respect to occupational exposure.
Not nearly as hilarious as the idea that lead is more harmful at levels <5 mcg/dL than it is in the 10-20 mcg/dL range. But that's the only way to explain the available data while attributing the observed correlation to Pb neurotoxicity. And indeed, it's the mainstream view among lead-IQ researchers.
Back when lead levels were mostly in the 10-20 mcg/dL range, the observed correlations with IQ were typically around 3-5 IQ points per standard deviation of lead exposure. Now that lead levels are an order of magnitude or two lower, the observed correlation is still... 3-5 IQ points per standard deviation of exposure. Blood levels of other environmental toxins have also been correlated with IQ. These studies typically come in around... wait for it... 3-5 IQ points per standard deviation of exposure.
The common-sense explanation, in my opinion, is that developmentally delayed children eat more dust and dirt.
Poverty is easy to control for. Controlling for it brings the Pb-IQ effect size down considerably, but not down to zero.
My view is that poverty, low IQ, and lead exposure are all caused by intractably complex webs of genes, behaviors, beliefs, and parenting styles. This occurs in such a way that high IQ, high socioeconomic status, and low lead exposure are all strongly correlated, but not perfectly correlated.
As a spherical-cow sort of model, imagine that there's a "I don't give a fuck" (IDGAF) gene. People with the IDGAF gene have higher levels of lead exposure (because their houses aren't as clean), lower IQs (because they're not as motivated to think hard about the IQ test problems), and lower socioeconomic status (because they're not as motivated to get high-paying jobs) relative to those without the gene. But some IDGAF-positive people will get lucky and end up with high-paying jobs. Some IDGAF-negative people will be impoverished due to bad luck, etc. In other words, socioeconomic status is a noisy indicator of IDGAF. The key point is that IDGAF-positive people who have good jobs still probably don't keep their houses as clean as IDGAF-negative people. So when you control for poverty, you've killed off most of the IDGAF-mediated correlation between lead and IQ, but not all of it.
(as a reminder, this was just a spherical-cow model for illustrative purposes... reality is intractably more complex!)
reply