Imnimo

50 ms·

631.
by Imnimo 4y ago
Sharecropping originally refers to a process where a landowner would let people farm their land in exchange for a portion of the yield. It has negative associations with post-slavery exploitation of black farmers as well as poor farmers in
632.
by Imnimo 4y ago
>Traversing the trade-offs between state-of-the-art performance and the ability to serve our user base at scale, we implemented (in its latest iteration) an EfficientNetv2-based binary classifier: a convolutional network that has faster
633.
by Imnimo 4y ago
What is the correlation between distrusting the media and believing that Trump won the election? If that correlation is very high, does this statistic say more about the media or about the public?
634.
by Imnimo 4y ago
>Copilot introduces what we might call a more selfish interface to open-source software: just give me what I want! With Copilot, open-source users never have to know who made their software. They never have to interact with a com
635.
by Imnimo 4y ago
You have to use a lot of judgment to pare these down, or you'll do more harm than good. "My car is very fast" -> "My car is breakneck" "It's going to be very cold tomorrow" -> "It's
636.
by Imnimo 4y ago
The way I think of it, we have a 512x512x3 target, so that's 48x the information. I don't think it's unreasonable to say that far less than 1/48th of the space of 512x512x3 outputs are natural images (meaning an image t
637.
by Imnimo 4y ago
I'm not a copyright lawyer or anything, but the way I look at it is that the big concern is that you can't easily prove that a particular AI output is not just a memorized copyrighted training example. So even if we assume that i
638.
by Imnimo 4y ago
I'd be curious to know how many additional "wanted" recommendations the system is able to serve by not treating a dislike as a strong negative signal. It seems like there's a tradeoff here, where YouTube figures that if
639.
by Imnimo 4y ago
I feel like this may not be the best time to open with a cutesy nickname like "Twilions"...
640.
by Imnimo 4y ago
Is there someplace that spells out exactly what an unstoppable domain is and how it works? I looked at their website but it's hard to find clear technical details.
641.
by Imnimo 4y ago
But this one requires only the primary course - why would you assume that you would need "all 6+ earlier boosters in sequence" when this one doesn't even require any previous booster?
642.
by Imnimo 4y ago
Is "has anyone ever made X?" really a "research question"?
643.
by Imnimo 4y ago
It is interesting that there are at least a few images in the dataset that were generated by previous diffusion methods: https://laion-aesthetic.datasette.io/laion-aesthetic-6pls/im... (Surely there are many more that
644.
by Imnimo 4y ago
Maybe you can paint adversarial patterns on the bottom of your pool, like those shirts that fool people detectors ( https://arxiv.org/abs/1910.11099 ) This probably wouldn't actually work in practice to avoid the ta
645.
by Imnimo 4y ago
Does this AI veganism extend to every model with training data that is not in the public domain? Every language model, every translation model, every classification and detection model build on top of Imagenet weights?
646.
by Imnimo 4y ago
>What if 80% of the public follows the political issues and is well informed? What if 40% have rational reasons to vote for the right-wing, and 40% have rational reasons to vote for the left-wing? Then it is the most ignorant, badly info
647.
by Imnimo 4y ago
Your interpretation also seems reasonable to me. I would say that "utilitarian calculus" in the second level represents possible disagreements over what the best calculus is, like "is measuring QALYs a good idea or should we
648.
by Imnimo 4y ago
I interpret it to mean something like, "you should put a lot of thought into figuring out what the best way to allocate your personal charitable donations is." I draw a contrast against other approaches to charity which are basic
649.
by Imnimo 4y ago
>Every single time I see an article about a new AI model that has a section called "societal impact" I know immediately they are not releasing the model, the training set, nothing... This section is just a requirement for some
650.
by Imnimo 4y ago
If the advice given by EA were simply, "here is our charity that tries to make good use of their money, you should donate to them", I think it'd look a lot different as a movement, and many common objections to it would be mo
651.
by Imnimo 4y ago
I would say that using GiveWell is a perfectly fine approach, but it is not the one implied by the phrase "strongly consider".
652.
by Imnimo 4y ago
The base level of the tower contains "we should strongly consider how much effort we devote to this". Isn't this the pillar that leads to all the crazy "Kill predatory animals" (I know this particular example is no
653.
by Imnimo 4y ago
> Interestingly, in the end I was able to withdraw all my funds from AAVE through direct contract calls as illustrated above Will this just result in the next wallet being tainted, and so on? I guess maybe you can try to cash out the cr
654.
by Imnimo 4y ago
In the original Prophet paper ( https://peerj.com/preprints/3190.pdf ) they claim that Prophet outperforms ETS (see Figure 7, for example). And in the NeuralProphet paper, they claim that it outperforms Prophet (but do
655.
by Imnimo 4y ago
>The opposing player has 1-turn to defend, or remove, the king from danger. Does this mean that if I am put in check, regardless of what I roll on the die, my king must be out of danger after my first move?
656.
by Imnimo 4y ago
I'm not copyright expert, but I would have imagined the answer is relatively straightforward: if an image would be infringing if drawn by a human, it's also infringing if drawn by a model, and similarly for non-infringing. Does t
657.
by Imnimo 4y ago
It's basically Pascal's Wager but for tech billionaires. Once you imagine of a suitably large magnitude of potential harm, it doesn't really matter how likely it is, you can convince yourself that you should worry about that
658.
by Imnimo 4y ago
Some bits of the website really set off my alarm bells. Having Stanford Medicine and Cornell University logos front and center on the landing page gives the impression that these organizations endorse the product. I know it says, "mad
659.
by Imnimo 4y ago
The crypto scam industry is going to need to tighten their belts, it sounds like.
660.
by Imnimo 4y ago
I think a lot of DALL-E 2 outputs fall into the category of "extremely impressive that a neural network made this" and also "not quite up to the standards of a human expert". Like if you show me an output and told me a

More ›