lairv

20 ms·

241.
by lairv 4y ago
That could only generate constructivist [0] proofs, and there are many things done in modern maths which are not constructivist. Maybe a better approach would be to use Curry-Howard [1] correspondence to directly get proofs from generated p
242.
by lairv 4y ago
It will be interesting to see how well this method will perform on other type of algorithms. Matrix multiplication was quite convenient: thanks to the divide-and-conquer approach you only need to find algorithm for 4x4 matrices (or other sm
243.
by lairv 4y ago
I wouldn't have thought we'd have these text-to-video models before the end of this year, so we might actually not be that far from minutes-long video generation
244.
by lairv 4y ago
I think AI now beat humans at poker https://www.sciencedaily.com/releases/2019/07/190711141343.h... https://www.cs.cmu.edu/~noamb/papers/19-Science-Superhuman.p...
245.
by lairv 4y ago
Reading the paper, what's interesting is that they really build upon text-to-image models. Instead of generating 1 image with diffusion model, they generate 16, and then use several upsampling/interpolation methods to increase the
246.
by lairv 4y ago
I think it depends a lot on what you mean by "better output" DALL-E is very good at conceptually representing complex prompt. Something like "a bear with a diving mask surfing in the ocean, a pelican is sitting on its shoulde
247.
by lairv 4y ago
Oh.. yeah I was talking about a web app, should have just said website I guess
248.
by lairv 4y ago
True although I have no idea how much people are willing to pay only for the ad-free feature (which they can get with any adblocker)
249.
by lairv 4y ago
Hard to estimate traffic for now. I use ad-blockers as well, but on the other hand I don't like the idea of hiding core features behind a premium-paywall
250.
by lairv 4y ago
Is that sarcasm ? Anyway I'd like my app to stay account-free (no account required to access any feature) so I don't really know how I could collect data
251.
Ask HN: Are ad revenue still a viable way to monetize an app in 2022?
19 points by lairv 4y ago|19 comments
252.
by lairv 4y ago
Does OCaml matches Go speed though ?
253.
by lairv 4y ago
> Chess engines aren't like car engines are to sprinting. I think this is questionable. While we can understand the physical limitations of a human compared to an engine, we tend to alleviate the intellectual limitations. Just like
254.
by lairv 4y ago
Put them in a room without window or any exterior interactions, 4 cameras in each corner, wifi/radio signal jammer, 10 minutes delay with livestream, I would honestly pay respect to Hans if he can still cheat in those conditions
255.
by lairv 4y ago
I get your point, thought it requires to already have a network
256.
by lairv 4y ago
So how do you go about seeking clients without having an actual product, do you just randomly contact people asking them if they would be interested ?
257.
by lairv 4y ago
> it would be reasonable to only treat the eventual on-chain withdraw as a taxable event wouldn't that be an easy way to bypass taxes ? make all your transactions off-chain and just withdraw once a year ?
258.
by lairv 4y ago
I don't really understand the hate either
259.
by lairv 4y ago
Nope, I must say I have only used a small subset of languages from the FP-language-zoo, so my opinion on OCaml might be biased But as an example when writing Typescript, I feel so frustrated of not having a clean way of doing pattern-matchi
260.
by lairv 4y ago
Every time I use recent "functional" languages (Rust, modern Typescript) I realise how great OCaml is
261.
by lairv 4y ago
"worthless" huh, not everyone can afford inference of a ~500gb models, depending on the the speed/rate you need you might definitely go for smaller model But maybe your sentence was more about "after BigScience model, op