I was playing earlier today with Triton[0], from OpenAI. Like Taichi it makes it super easy to write native GPU code from Python, but it really does feel like something very experimental for now. (I know the use case is very different)

[0] https://openai.com/research/triton

Triton is clearly a popular name for GPU access and inference[0].

[0] https://developer.nvidia.com/nvidia-triton-inference-server

>but it really does feel like something very experimental for now

Meaning?

- their approach is still bizarre and exploratory and they still don't know how to structure their APIs and are making it up as they go?

or:

- there are still some rough edges, bugs, and no full documentation yet?

as those are quite different cases...

>CPUs and AMD GPUs are not supported at the moment

CUDA-only, no mention of Metal.

I've been meaning to do some project in the browser involving the GPU and can't decide on a framework to use.

What would you recommend? The cool kids seem to be using pure webGPU

This was made in Rust using WGPU and their WGSL shading language. It still needs more documentation and tutorials, but the results are great and port to many platforms.

I'd love to read about this view of the world as the power actuated by language. It is really a popular narrative nowadays everywhere you look (people censoring "bad" words etc).

Which thinker is that from? Foucault? Marcuse?

This might be an individual reaction to fasting, but I always have headaches when I don't eat properly (skip meals, etc). Is this something that I would have to get used to when eating less?

Could be hypoglycemia, which has been linked to headaches.

One thing I've observed from going on / off IF several times is that trying to jump straight into the full fasting window can cause problems. If you normally eat on a 12/12 schedule, for instance, and try to switch straight to 16/8, your body will not appreciate it.

Try very gradually building up to the window -- add another 10 or 15 minutes between meals each day.

I got used to it by doing 12-12 then 14-10 then 16-8 then 18-6. I believe I did 10 days on each phase as 1 week seemed too "quick". It was just my moderate way of weaning myself off breakfast. I have tried the 24 hour fasts and I just can't make it and not be miserable.

Hello HN! After seeing a similar project on HN[1] I thought I would share my pet project of using AI to find citations between my books (e.g. Montaigne cites Plato, Hannah Arendt cites Dostoevsky etc.)

The ideia is to use AI to detect from the text books citing another books, and to make it "historical" by assigning the correct publication date to the target and source of each citation and graph everything!

The technical aspect is explained on this medium post[2](paywall free link!)

If you have any suggestion of improvements I will be happy to hear them :)

[1] https://news.ycombinator.com/item?id=28595967 [2] https://medium.com/mlearning-ai/graphing-citations-between-b...

Amazing results!

I did something similar with RoBERTa and my own Kindle library to graph (with D3.js) all mentions/citations between my books (which books cites another books I have). I sorted the final graph by publication date to see some cool historical patterns of books citing another older books [1]

I also manually annotated ~1000 book mentions, but I combine RoBERTa with string search (I list all titles I want to search a priori) to reduce the number of false positives. I also augumented the dataset with thousands of books titles and metadata from goodreads.

I explain all the process on a blog post[2]

[1] https://thiagolira.blot.im/_projects/book_graph/main.html [2] https://medium.com/mlearning-ai/graphing-citations-between-b...

Thank you!

The medium post is amazingly written! I basically did the same thing - and you beat me with the data augmentation piece. I tried using nlpaug [0] but it didn't improve the model performance. I'll definitely try swapping book titles around.

[0] https://github.com/makcedward/nlpaug

Can you also share the source code of the model and the site? I am starting to learn NLP, and I love indie applications like this.

I once applied for a company here in São Paulo where after the technical interview (no whiteboard) I talked on-site with the CTO and he passively-agressively asked a bunch of random questions and tried very hard to be a dick about it ("oh, you consider yourself a good self-learner? French? I bet you couldn't read some news in French if I opened any here right now. JUST KIDDING")

At the end of this 40 minute nightmare I had a bizarre homework to do at home that involved writing a list of uses to a paperclip.

They took a month to send me an email saying I wasn't accepted.

IMO you should have ended the interview the moment you detected that you were being treated poorly. Let alone not doing the homework.

I went from knowing nothing of French to reading 20 thousand leagues under the sea by Jules Verne in French (the original language)

It's not that big of an achievement since I'm Brazilian and I already speak Italian, the romance languages are somewhat similar. But I'm happy and I want to keep learning new languages because that proved to be a rewarding experience all steps of the way.

I've used Lingq and Memrise, if anyone is interested :)

Was going to ask what you had used. Brazilian too, but know only Spanish (kinda), and English. Will check out these two. Thanks and congrats!!!

LingQ is something else. The ideia is that you should feed the app with "input" from your target language that you care about[1], and listen to it again and again. Memrise is more like a Duolingo that actually works haha

[1] https://www.youtube.com/watch?v=rw-u_vVx6Is

Thanks!

I do envy you. Been busting my brain with French for years as its always been my dream to speak it. But coming from Slovenian (slavic) and not having an ear for languages does not help.

So, congrats!

Yeah, I've been applying more often to the smaller ones.