hwers

10 ms·

61.
by hwers 5y ago
It's hard for me to see "X gets more funding" as something to celebrate. Funding is essentially zero sum so that just means something else is now less funded as a consequence.
62.
by hwers 5y ago
I know even publishing the code for stuff like this is generous but it'd be so cool if the standard practice included publishing a little wasm binary blob as a library independent network with weights to play with without having to go
63.
by hwers 5y ago
I agree. The limit encourages creativity and brevity
64.
by hwers 5y ago
The spam scale problem should be solvable with a one-shot locally evaluated NN trained to classify behavior/tweets from their massive massive ground truth dataset. If it’s evaluated locally then I don’t see how scale is an issue.
65.
by hwers 5y ago
Anyone have a clue what happens next?
66.
by hwers 5y ago
I've come to think of this as just a culture problem. We think of share-everything and recent social changes as innovation and "the future" but not everything that is modified at t+1 vs at time t is necessarily a pareto impro
67.
by hwers 5y ago
Making claims without scientific backing probably shouldn't be a bannable offence. The spirit of science is about discussion and counter arguments and being open to being wrong or even the idea that the consensus might be invalid, not
68.
by hwers 5y ago
I think we're not discussing the real point here which isn't whether "full completely unregulated free speech" should be the norm on twitter or not. Elon isn't proposing allowing CP or whatever the worst stuff you s
69.
by hwers 5y ago
Write a paper about it. Post it on arxiv.org. Contact some open minded researchers on twitter or here (show HN) for critique.
70.
by hwers 5y ago
I think this is really cute and wonderful. I think there's some room to improve here, e.g. an AR ghost overlay of previous frames as well as ML extracted frames from pure paper without having to print a template (both of which I believ
71.
by hwers 5y ago
Autonomous cars and fusion are very different creatures than the growth of the internet or computers in general. The latter was widely distributed and had low hanging fruit that just about everyone in society could pluck and have fun with,
72.
by hwers 5y ago
Are you sure? I'd like to hear the take of someone in their early twenties on this. My feeling is that a lot of them are terrified by it themselves too.
73.
by hwers 5y ago
Would you say The Recognitions is readible? I've wanted to read it for a while but it felt intimidating and it was lumped into the 'post modern' category in book stores etc which has made me afraid it's one of those samu
74.
by hwers 5y ago
Imagine any diffusion-style text-to-image model on specialized ASIC hardware.
75.
by hwers 5y ago
Can't wait for DeepMind to take a stab at outcompeting dall-e.
76.
by hwers 5y ago
Not a good look for Parag as an early move as the CEO.
77.
by hwers 5y ago
Who even decides what's best for the company? One could argue him openly criticizing twitter is in its best interest. Like pouring alcohol in a wound, it hurts and kills off some bad parts of the system but it'll be stronger at th
78.
by hwers 5y ago
They should probably be transparent about it in latter case though (which they aren’t always)
79.
by hwers 5y ago
Feels like both the comments in here are paid for by Notion somehow.
80.
by hwers 5y ago
Hang on to NP complete problems instead, that one will stick around for a while I expect.
81.
by hwers 5y ago
They'll just be A/B tested from out of a collection of alternatives. (After all isn't that what the artistic filter is supposed act as a proxy for in the first place.)
82.
by hwers 5y ago
"We've made the scarce resource abundant, finally the scarce thing is democratized!" "Wait why doesn't anyone care about the scarce thing anymore?"
83.
by hwers 5y ago
You could just use ffhq/faces, it's fine. You'd think it gets stuck in a local minima but it really doesn't. Transfer learning from any 'finished' checkpoint is much much much faster than doing it from scratch.
84.
by hwers 5y ago
I learned everything I know about stylegan (prior to experimenting myself) from this channel: https://www.youtube.com/c/ArtificialImages/videos For stylegan this playlist was useful: https://www.youtube
85.
by hwers 5y ago
> "The computation for the 512x512px training set was run for 20 days on an IBM AC922 POWER9 server running RHEL7, utilising all four Nvidia Tesla V100 SXM2 16GB GPUs." That's super overkill. If you just transfer learn fro
86.
by hwers 5y ago
I've always wanted to read that book but it always felt so intimidating. Would you say it's readable? It feels like it'd be really dense.
87.
by hwers 5y ago
I strongly disagree, I loved it
88.
by hwers 5y ago
The correct response here from the artists point of view should be a widespread coming together against their art being used as training data for ML models. With a quickly spread new license on most major art submission sites that explicitl
89.
by hwers 5y ago
They're also not censored on the dataset front and thus produce much more interesting outputs. OpenAI has a low resolution checkpoint for similar functionality as this - called GLIDE - and the output is super boring compared to communi
90.
by hwers 5y ago
Yeah if I were in a similar position in a future scenario and I found this story and it ended with public shaming and prison etc, then I'd just burn the thing instead of trying to return it.

More ›