Daniel Dvorkin @medigoth@qoto.org

@lulu_powerful Yeah. I usually *don't* pick apart the functions unless (a) something doesn't look right and I want to check, (b) I want to implement a faster and/or more stable version, which doesn't happen nearly as often as it used to, or (c) I'm just curious. But knowing that I *can* is reassuring. Not to mention implementing new tests, where I *always* do the math on paper first!

There's already way too much #ML code out there that basically says "this seems to work." My background is about equal parts #biostatistics and software engineering, so I find this just as frustrating as I do the very large amount of statistically sound but *horribly* inefficient and poorly documented code that makes it into the wild.

Interesting tidbit from FiveThirtyEight:

"Sad news for us data nerds: According to respondents of a recently released December YouGov poll, #statistics was ranked the least interesting college major, with 42 percent of adults calling it 'not interesting.' Criminal justice had the lowest percentage of respondents who called it 'not interesting,' at just 18 percent. When respondents were asked which majors they would pick if they were pursuing a college degree today, a plurality (20 percent) chose #computerscience."

The full poll results are here: https://docs.cdn.yougov.com/8nv9pr6ke6/results_College%20Majors.pdf

Almost surely, a lot of those students who think computer science is interesting, but statistics isn't, are planning on careers in "#datascience," #ML / #AI, etc. So here comes yet another generation of computer scientists who will badly reinvent statistics instead of learning the field from the ground up. Great.

@jhertzli Ha! Like you, I'm reluctant to accept that conclusion, but, well ... [ianmalcolm]there it is[/ianmalcolm].

I guess the #awesomebros will find it ...

...

...

...

... hard to swallow!

Thank you, thank you, I'm here all week. Try the #pachycephalosaurus.

https://markwitton-com.blogspot.com/2023/03/new-paper-fresh-evidence-and-novel.html

@b3n I'd be a lot more willing to accept that framing if he didn't present himself as an AI researcher, and pull stunts like challenging actual researchers to debates. He bears much the same relationship to computer science as Kent Hovind does to evolutionary biology. Unlike Hovind and other prominent creationists, though, Yudkowski very often manages to fool people with a genuine interest in the field into taking him seriously.

Every #tool is dangerous, and the more powerful the tool the more dangerous it is. Of course. Is #AI as dangerous as #nuclear #weapons? Probably not. It might be in the same league as, oh, say #internal #combustion #engines—and those have done a hell of a lot of damage. But they haven't done it by ushering in the #apocalypse. Instead the damage is from slow, creeping, cumulative change where the effect of any one individual event is too small to measure.

So I really think the focus on world-ending scenarios takes away from the conversations we need to be having. This reminds me a lot of the simmering "how far is too far" #genetics debate, especially the kibitzing from "#ethicists" with no understanding of the #biology and an #ethical sense that isn't nearly as developed as they think it is. There are conversations on that topic I'd like to have without the constant Greek chorus of "#Frankenstein! #Gattaca! #JurassicPark!"

This is your regularly scheduled reminder that Eliezer #Yudkowsky is a complete fraud. He has no qualifications of any kind, in any field. Nor has he contributed any substantial research, in any field. His only credentials are membership in a "research institute" he founded himself, and starting a #cult. None of his opinions are worth a moment of your time.

If you know, you know. If you don't, you're probably better off. Carry on.

@Pat Heh. Most of that was probably infant mortality, to be fair. And older children and adults at any point in their lives were a lot more likely to die by misadventure, infectious disease, or environmental causes, than we are now. I suspect the lifespan for people who manage to avoid *external* causes of death really hasn't changed much throughout the history of the species.

But yeah. Just because they ate a certain way, based on what they could get at the time, doesn't mean that was the ideal diet! Same for other species, really, e.g. all the "this is the natural diet for your dog or cat" pet food.

In fairness, I should add that I do know people who have found certain aspects of the #paleodiet work well for them. Weight loss, lower blood sugar, etc. Okay: I think the ideal #diet is very much an individual thing, and people who are measurably healthier when they eat a certain way should absolutely eat that way! But when it becomes an #evangelical #religion, it ceases to have any possible medical value.

This makes a great deal of sense to me. Our ideas about which #foods and methods of preparation are tasty vs. disgusting are *deeply* cultural, even in today's relatively homogenized world.

Much more so when our #ancestors lived in widely separated groups, in vastly different environments, and making use of whatever they could get was the only alternative to #starvation. Humans are the omni-est of #omnivores, and we assign cultural significance to practically everything we do.

Also, I'm always just happy to pass on any news that demonstrates how much the modern idea of the "#paleo diet" is bullshit. That's a pretty strong preference on my part, but I don't claim it as a cultural universal. 😉

https://www.sciencenews.org/article/meat-rotten-putrid-paleo-diet-fire-neanderthal

Via an old #USAF #medic friend: a picture from my much younger days. My head looks weirdly small. Perhaps I was hoping the jolt would blow it back up to normal size.

I can date this one pretty precisely. It's at the #RAF #Upwood #clinic, where I arrived in June 1990, and shaved off the mustache within a few months. Also, I'm wearing whites, which I only did at the Primary Care Clinic, where I was first assigned. In the Acute Care Clinic, where I was from October of that year on, we wore BDUs. So it has to be somewhere in that time frame.

That kid, he's not me. Better in some ways: energy, optimism, idealism, openness to new experiences. Worse in others: temper, stubbornness, a certainty of his own rightness taken to absurd extremes. It takes some growing up to learn the difference between opinion and truth. Uncle Sam was not always pleased with the process.

But I remember him fondly, and wish him well. Given the number of people who were close to him then and are still here now, I guess they do too. I'm glad.

@realcaseyrollins @freemo "Thanks for letting me know when to stop reading" is a rhetorical device, of course. I did read your entire post. And *if* I saw evidence of what you describe in #1, I'd agree with your final point.

But so far I haven't. What I have seen is a whole lot of conservatives labeling as "progressive activists supporting child pornography and involving children in sexual performances" things that really aren't, e.g. the tempest-in-a-teapot "drag queen story hour." If you present actual evidence—from reliable sources, not right-wing propaganda sites, to be clear—I'll try to evaluate it fairly.

BTW, there *are* people making children put on sexual performances ... in the child beauty pageant world. I don't claim to know, but I'd guess most of the adults who put those events on aren't "progressives."

@freemo It occurs to me that it might be fairly easy to get the data, at least as far as *convicted* offenders go. I don't actually know what information is available on sex offender registries, but if they include empoyment at the time of conviction, one could download the data and gather the relevant statistics. Otherwise it would be a hell of a research project.

I'm curious enough about this that in the first case, I might actually do it, and if so I'll post the results here. But really I should probably get back to the kind of statistics they pay me for. 😀

@realcaseyrollins @freemo No. If you genuinely believe what you wrote in item #1, provide some evidence.

@realcaseyrollins @freemo Casey, thanks for letting me know when I could stop reading with the first few words of your first "point."

@freemo I don't have the data. I was hoping maybe you did. :)

If neither of us has the numbers, of course to some degree we're just blowing smoke. But I will say I personally know an alarmingly large number of people who suffered exactly that type of abuse, and never got justice for it.

@freemo "religious conservatives of all denominations," I should have specified.

@freemo How about youth pastors, teachers at church schools, et al.?

I agree that the Catholic church has been scapegoated (probably shades of ancient anti-Catholic prejudices in general). They've covered up plenty of abuse, but they're certainly not the only offenders, and probably not the worst ones. My point is that these days, religious conservatives in the US label things like teaching LGBTQ kids to be comfortable in their own bodies as "grooming" ... while having an *extremely* well-documented history of covering up the real thing.

Most school shootings are inexplicable to anyone but the killer or killers. This one may not be: https://twitter.com/xxclusionary/status/1640598560876920832

As soon as a friend sent me the link, before I even clicked on it, I just knew what it was going to be. Churches and church-affiliated schools covering up systematic sexual abuse of children is very nearly as common as the abuse itself.

Expect an intensification of the "groomer" rhetoric in the coming weeks. They still think they can cover themselves that way.

#church #school #shooting #schoolshooting #sexualabuse #molestation #grooming #conservative