ThenElection
No bio...
User ID: 622
ThenElection
1yr ago
More or less. It's essentially a binary search, where you're looking for a local maximum instead of an exact value. It's optimal.
Usually competitive candidates see this immediately. If someone suggests a linear scan, I'll ask if we can do better and they eventually get to the binary search. That's not in itself something that results in a NH recommendation, but in practice the people who get the binary search right away tend to do better on later parts.
ThenElection
1yr ago
I've been playing with a theory for awhile: Google became Saudi Arabia. After it solidified its ads product, it had an infinite money spigot. It could throw lots of money around internally, and it was unlikely for what would otherwise be successful products to significantly improve its bottom line. Internal politics became the main driver of who got resources for vanity projects, and the general transformation of its culture (including the Wahhabist DEI initiatives) followed from that.
ThenElection
1yr ago
If a query on the list just gives you a single bit of information, you're right, giving us the log2 bound. I have seen the claim made in an interview that a quantum speedup is possible to improve on that. (Cf https://web.mit.edu/rsi/www/pdfs/papers/2003/2003-brianj.pdf for a quantum search on an ordered list).
Would love for someone well-versed in quantum algorithms to confirm.
ThenElection
1yr ago
I don't think HBCUs have worse instructional quality or students than colleges that recruit students with similar SAT scores. An expectation that Howard grads perform similarly to MIT grads isn't a realistic expectation.
ThenElection
1yr ago
TheMotte wasn't banished; it was a self-imposed exile in response to an increasingly opaque, arbitrary, and hostile environment. Perhaps a distinction without a difference, but my impression is that it had one or two years left in it before the coup de grace.
The general sense I get from e.g. stupidpol is that they strongly dislike the reigning neoliberal order, and the neoliberal order is supporting Ukraine. The near-enemy is more salient than a distant entity, and so they perform their political identity by expressing skepticism of everything the near-enemy says.
China support is a bit more complicated and substantive, but it offers an alternative ideology that someone can map their own political values to with only a moderate amount of mental gymnastics and curated blind spots. It's comforting to imagine you have a powerful counterweight on the horizon to a political system you hate.
This week's revolutionary AI advance:
It's not really revolutionary, as people have been pointing out this is the obvious next step for ages months now. But it still is a milestone worth noting.
As for this:
While our internal testing suggest much of explicit and violent content can be filtered out, there still exists social biases and stereotypes which are challenging to detect and filter. We have decided not to release the Imagen Video model or its source code until these concerns are mitigated.
Google's made a habit of this. They announce an amazing advance, and then say no one can have access to it because it can be used for Evil. No matter: Stable Diffusion will have something comparable out in a couple months.
ETA:
Actually, this out of DeepMind might be the bigger advance today, if less flashy:
Press: Discovering Novel Algorithms with AlphaTensor
Paper: Discovering faster matrix multiplication algorithms with reinforcement learning
ThenElection
1yr ago
As far as the training set goes, it used more images than videos, treating images as single frame videos. Although you're right that very similar videos likely existed in the training set, what's interesting is that there was substantial knowledge transfer from the images to the videos. E.g. the Van Gogh cat eating, and the paper has further examples (a drone flying through a 8-bit pixel city, or a rotating 3d model of a car made out of sushi). I don't see videos regurgitating the training set being any more an issue than it is for images.
ThenElection
1yr ago
From a profit-maximizing perspective, Google should push for an extremely regulated environment for generative models, to protect us from the fake news/deep fake revenge porn/violence/Russian bots/racism/sexism/copyright violation. Put up large barriers to entry, have the government restrict access to training data and maybe even GPUs, and then claim all the profits for themselves.
ThenElection
1yr ago
It's also possible that manufacturers could nerf GPUs for the purpose of ML except for customers with whom they have a special relationship. See e.g. the rate limiting NVIDIA did for crypto mining while still selling a higher priced card without the nerfing.
Does FSDP help at all here? My very naive understanding is that its approach allows sharding of the model parameters so that they don't have to all fit into VRAM, though I wouldn't be surprised if it couldn't scale down to arbitrarily low VRAM or scale up to arbitrary numbers of parameters. Perhaps a similar strategy could be used for wide scale distributed learning on consumer hardware.
ThenElection
1yr ago
Every straight dating app has a large excess of men and a large deficit of women. Most fail before ever achieving a critical mass of users.
I'd be curious to see the gender balance in a socialist/far left dating app. Apparently there once was one called Red Yenta, but they rebranded into the personals section of a Jewish lefty lifestyle magazine.
ThenElection
1yr ago
Rings of Power is far from Bezos' largest expense in the venture of his dating life. It's comparatively a bargain.
ThenElection
1yr ago
Maybe I'm used to it, as it doesn't really break the fourth wall for me. But at some point we're going to have Idris Elba signing the Magna Carta.
ThenElection
1yr ago
To make my point clearer and plainer: Amazon Studios did not originate as a way to make a meaningful profit for Amazon, but as a way for Bezos to buy entry into Hollywood circles and increase his status. People wondering about why it does seemingly irrational things are mistaking its intention.
Inherently it's a bit gossipy, but some tea leaves:
Does Amazon Studios make a profit? This is a tricky question. If it does, it does so far behind not just AWS (the main profit driver) and online retail, but also ads. In earnings reports, it's not even broken out separately from subscription services. It's not exceptionally fast growing: compare subscription service revenues (+10% YoY) to AWS (+33% YoY) and ads (+18% YoY), though online retail is stagnant. It's also unlikely that Amazon Studios is even a primary driver of subscription revenue.
Anecdotally, I know of no one who subscribes to Amazon Prime for Amazon Studios, though I wouldn't be surprised if it did increase subscriber retention by some amount. 5%?
Also consider how Amazon is thinking about the $1B RoP:
If we can't make [RoP] successful, why is Amazon Studios even here?
On the other side of things, what Bezos has gotten out of it:
Sanchez and her ex-husband, Whitesell, were photographed with Bezos in 2016 at a holiday party for Manchester by the Sea. The film was produced by Matt Damon (one of Whitesell's clients) and distributed by Bezos' Amazon.
But obviously we're not going to see a line item on an SEC report detailing the YoY increase of Bezos getting laid.
ThenElection
1yr ago
Earlier today I reread Neutral vs. Conservative: The Eternal Struggle, which seems useful background. My question is, you say:
Then there's what amounts to a very foolish, but understandable strategy.
Why? There's really no hope for conservatives to get fair representation in ostensibly neutral institutions, and nearly all the relevant voices on the Right have abandoned them. By ideology and cultural background, the participants in those institutions are incapable of thinking of alternative viewpoints as anything except motivated by evil. And we've reached a point where in 2021 only 12% of adults had a lot of trust in national news organizations, which is a record low.
At some point in the recent past it was probably true that national news organizations were more accurate/fair in their reporting than the explicit partisanship of right-aligned media. I don't think that's the case anymore. Even if you disagree, it's likely that the Right can convince an outright majority of voters that institutions are just a mirror image of Breitbart. Discrediting and delegitimization seems like a winning move.
ThenElection
1yr ago
Gotcha. I agree that this will only marginally push the needle on delegitimization, albeit positively. Perhaps someone will go check out Pushaw Trust Journalists; perhaps the monoculture of who interacts with those institutions will make a couple more people skeptical of what they say or print.
ThenElection
1yr ago
That's a fair point, as it just refers to national media organizations and not particular organizations.
Reading the tea leaves, though, the high trust ratings are mostly sustained by Democrats in the same poll. It's unlikely that the organization Democrats are expressing high trust in is Fox, and the decline over the past several years is driven heavily by Republican and Republican-leaning adults, who are the primary targets of this strategy.
I watched the first episode last night to have a more informed take on it. I don't care about the deviations from lore, and the casting choices don't bother me. It's pretty.
The big issue is that... it's boring. Aside from Galadriel, all the characters are completely flat, and there's nothing that makes me want to keep watching. It's probably unfair to compare it to E1 GoT, but it's weak compared to HoD or even WoT. My GF (who has watched the Peter Jackson films half a dozen times but has no other connection to Tolkien or fantasy lit) checked out and went to bed halfway through. I persevered but won't be watching further episodes. I'd probably get more out of generating hobbit images on Stable Diffusion for hours.
ThenElection
1yr ago
I'll give it another shot, based purely on the recommendation.
ThenElection
1yr ago
Or take this piece of ridiculousness, such that it is.
Oh, my.
The Japanese derided the Chinese as "yellow". As Michael Keevak points out, Japan saw itself on par with Western powers. Its imperialism mirrored the imperialism of white colonisers. In the West, the Japanese were still seen as "coloured people", Keevak says, but "maybe not as yellow as the Chinese." For the past three centuries, power and whiteness have been synonymous. From the British Empire to the American century, white nations have exported violence, committed genocide, stolen land and made it all legal. China, like so many other non-white nations, has felt the sting of white imperialism.
Uh... is the claim here that the Greater East Asian Co-prosperity Sphere was a white supremacist project?
the Chinese Communist Party itself mirrors whiteness. The irony is Xi has also become what he opposes. He is a Han nationalist — his idea of Chinese power is ethnic Han superiority — persecuting non-Han, non-white people in his own country. If whiteness is power, Xi Jinping is its champion. The continuation of white power, in darker skin.
And that the CCP and Xi are white supremacist? Like, is the idea that anyone who's not a helpless victim and who has agency in the world is white?
ThenElection
1yr ago
I've noticed that Google's and OpenAI's showcases of really awesome (but totally gated off) AI media generation systems are typically about a year or two behind the open-source implementations of those, and if you haven't noticed, Google's image generation systems have gotten really really good. The clock is ticking.
Minor point: not to pick on Google too much, but keep in mind that Google's examples are all cherry-picked. Today's Imagen/Parti models are not leaps and bounds beyond the public models. Maybe, generously, a couple months ahead.
ThenElection
1yr ago
IANAL, but the relevant statute may be here:
104.16 Voting fraudulent ballot.—Any elector who knowingly votes or attempts to vote a fraudulent ballot, or any person who knowingly solicits, or attempts, to vote a fraudulent ballot, is guilty of a felony of the third degree, punishable as provided in s. 775.082, s. 775.083, or s. 775.084.
Knowingly seems relevant. Particularly, if election officials in the state of Florida told them that they were eligible to vote, as appears to be the case, it seems like the State has a high hurdle to climb to argue that they knowingly voted a fraudulent ballot and that they are guilty of illegal voting.
It also seems like even the state of Florida didn't even know whether they were eligible to vote or not.
People applying, definitely.
Getting hired usually is a signal that someone knows basic data structures (nothing too exotic) and can make vaguely accurate gesticulations about complexity.
And yeah, you should. Not as intellectually stimulating as being a bricklayer or plumber, but definitely much better compensated.
More options
Context Copy link