scarmig

17 ms·

241.
by scarmig 1y ago
Speculative, more than imagined. You're free to offer other reasons for why objective statistics indicating male disadvantage in certain contexts don't count as male disadvantage.
242.
by scarmig 1y ago
Women attend university at significantly higher rates than men (3:2 ratio). Men's suicide rate is around 3x that of women's. The ratio of the unhoused is even worse, more like 4:1. Men work tens of thousands more hours during thei
243.
by scarmig 1y ago
That thread was skeptical, but it's still far more substantive than what you find here today.
244.
by scarmig 1y ago
It's caught in a kind of feedback loop. There are only so many times you can see "stochastic parrot" or "fancy autocomplete" or "can't draw hands" or "just a bunch of matmuls, it can't repli
245.
by scarmig 1y ago
"Marginalization" is a systemic issue; its reality isn't dependent on anecdotes or your particular happiness with your lot in life. You've got to look at differential statistics between men and women to see if either (or
246.
by scarmig 1y ago
Your "ideas" are just regurgitations of things you read off the internet; you have no coherent theory of "creativity" beyond some ineffable reference to the sanctity of the human soul.
247.
by scarmig 1y ago
We don't need theoretical market socialism; we had actual existing Yugoslavia. Although it functioned better than centrally planned socialism, it still had lots of issues related to prices, particularly of capital. Who provides capital
248.
by scarmig 1y ago
Ironically, I suspect a UBI not only can coexist with inequality but might substantially increase it (not a bad thing in my book). The vast majority of Americans already have incomes above a UBI level, especially when current government ben
249.
by scarmig 1y ago
Socialism didn't fail because of a UBI, which it never attempted; it failed because it couldn't calculate prices accurately, because it was bad at finding and processing information, political economy, and deeper computational com
250.
by scarmig 1y ago
The complaint is that laws are structured in a way that only affect people who have something material to lose. There's no law of nature that says laws have to be that way, and, historically and cross-culturally, they haven't been
251.
by scarmig 1y ago
The diversity is itself indicative, though, that intelligence isn't bound to the particularities of the human nervous system. Across different animal species, nervous systems show a radical diversity. Different architectures; different
252.
by scarmig 1y ago
What does Mondrian's estate owe to Oliver Byrne's estate, since he basically stole Byrne's color scheme and style from his version of Euclid's Elements? (And what does Byrne's estate owe to Euclid's estate?) IP
253.
by scarmig 1y ago
You'd still want to hedge.
254.
by scarmig 1y ago
Of course. But that's not relevant to the point I was responding to suggesting that LLMs may lack consciousness because they're deterministic. Chaos wasn't the argument (though that would be a much more interesting one, cf &q
255.
by scarmig 1y ago
If you somehow managed to perfectly simulate a human being, they would also act deterministically in response to identical initial conditions (modulo quantum effects, which are insignificant at the neural scale and also apply just as well t
256.
by scarmig 1y ago
Did you investigate other search processes besides SGD? I'm thinking of those often termed "biologically plausible" (e.g. forward-forward, FA). Are their internal representations closer to the fractured or unified representat
257.
by scarmig 1y ago
> DeepBrain (the people who did all the AI work at Google) "DeepBrain" isn't a thing. Google acquired DeepMind, which was the second biggest research lab in AI at the time. The first biggest was Google Brain. They existed
258.
by scarmig 1y ago
All sheer speculation--the reference to cyberattack was from a now dead comment suggesting it was ransomware. It doesn't seem likely to be a physical obstruction on the tracks, though, as the entire system is down and trains aren'
259.
by scarmig 1y ago
Oh, that just updates the AAAA record for the Waymo. Totally harmless.
260.
by scarmig 1y ago
I think there should be options for low income residents to ride for free or heavily discounted rates (which exists now). But it's all about implementation: simply letting anyone jump on turns the system into a madhouse. Making sure ev
261.
by scarmig 1y ago
The gates are less about direct fare recovery and more about limiting the externalities that gate jumpers impose on other passengers. BART is definitely calmer and cleaner than it was a year ago.
262.
by scarmig 1y ago
I don't know the last time the entire BART system got shut down. There was a loud explosion earlier this morning, which may or may not be related, but there aren't any power outages. No information as to the actual cause right now
263.
by scarmig 1y ago
Identity politics of all sorts get their salience by having enemies to coalesce around. Who would be the enemy of your broadened woke coalition? White coastal elites?
264.
by scarmig 1y ago
Inevitably, companies would arise that don't sell your data, and charge a higher fee to compensate the loss in revenue. Then, they'd go out of business as customers decide they prefer to sell all their data for a couple extra buck
265.
by scarmig 1y ago
Key point is that it's a system that you shouldn't trust, not any individual algorithm designer or implementer. Bugs that cost Google money will be found and fixed really quickly; bugs that make Google money will linger and go unn
266.
by scarmig 2y ago
Check out feedback alignment. You provide feedback with a random static linear transformation of the loss to earlier layers, and they eventually align with the feedback matrix to enable learning. It's certifiably insane that it works a
267.
by scarmig 2y ago
Dopamine modulates long term potentiation and depression, in some complicated way.
268.
by scarmig 2y ago
No one does anything useful with a $3 spend. That's not anything particular to TPUs, though.
269.
by scarmig 2y ago
This will be controversial, but w3schools isn't bad for the niche it targets today. It, however, has a long, well-earned reputation from being likely the worst site on the net for a decade, back in the early days. Even if I have a ques
270.
by scarmig 2y ago
Perhaps true, but the strongest privacy protections in the US are still pretty weak. The biggest penalty I know of is Anthem 2018, where they leaked HIPAA-qualifying records on 80 million customers. Their financial penalty was a whopping...

More ›