WanderPanda

Afaik libcurl is for clients only, but I also need to implement a server. Otherwise it sounds like exactly what I wish for from the governance perspective

Don‘t you need a http server to bring about any websocket connection through the upgrade mechanism? Otherwise it would only be a websocket client library

Damn I just yesterday was pulling my hair out to get boost::beast to run (how unergonomic can an api be?!). Now I‘m considering to switch but what I really need is long term stability. I‘d rather like to not ever touch this code again, but it should compile over the next 5-10 years. I‘m not sure which one would be preferable in that regard

Couldn‘t there be something between chaotic and completely random, let’s call it correlated, where e.g. (conditional) independence structures are similar in real-world problems?

So obscure that in a field as important as optimization we still think in terms of „escaping from local minima“. Also (as a total outsider) the progress in general optimization algorithms/implementations appears to be very slow (I was shocked how old ipopt is). I was wondering if all the low hanging inductive biases (for real world problems) have already been exploited or if we just have no good ways of expressing them? Maybe learning them from data in a fuzzy way might work?

My great confusion is why docker —-platform linux/amd64 is so much faster (almost native performance) than x86 UTM VMs. Can docker somehow leverage Rosetta?

I just want a C with templates (and function overloading)?

It is free for me from here: https://link.springer.com/book/10.1007/978-3-031-45468-4

The ebook seems free from the first link you mention (https://link.springer.com/book/10.1007/978-3-031-45468-4)

I agree with you but I think the argument is flawed. If you think about it like this h265 also just steals 10% (or whatever the compression ratio is) of an artifact

I was wondering the same! I just recently came across the restrict keyword (thanks to someone mentioning it on HN) but I have never seen it being mentioned before, even though I was looking into all sorts of optimizations (also ChatGPT would never bring it up on its own).

Since with templating everything seems to be going header only anyways, I feel like we should give the compiler more power to reason about the memory layout, even through multiple levels of function invocation/inlining.

I wrote some library where basically all shapes can be infered at compile-time (through templates) but the compiler only rarely seems to take advantage of this.

Oh no I'm just realizing that arxiv will be increasingly spammed with what should have been a blog post. I hope I'm wrong in assuming that in a few years the level of credibility that comes with a paper being on arxiv will have entirely worn off.

I know that in theory arxiv, being a pre-print server, shoulnd't give any credibility but practically that is the case and it still is a good quality/bs filter compared to e.g. Medium articles.

Depends on how you interpret the mission statement of building ai for all of humanity. It’s questionable that humanity is better off if ai only accrues to one or a few centralised entities?

You can still be a force for decentralization by creating actually open ai. For now it seems like Meta AI research is the real open ai

“never” is a strong word. I believe in the RL era of OpenAI they were quite aligned with the mission/values

So now Sam and Greg have Satya in their pocket, probably a more powerful position than being in the OpenAI management

The robotic hand Rubiks cube manipulation using RL was no joke either!

I agree with the sentiment but "big things" != broader audience

It was supposed to be a non-profit

I think it is still true that they have a higher dynamic range. They can build on the highest level of quality (iPhone, DJI) but also produce loads of cheap but dangerous stuff.

Shouldn‘t the NAT take care of the outgoing frames as it is operating on the tcp/udp level?

This! Who really needs to tie together business logic components written in different languages?!

I don’t think the margins here are so considerable that we can assume revenue = profit

No computational photography/videography?

If funded by private donors it seems legit but if this is funded by tax money it smells like a grift

By asymmetric capabilities I guess

Agreed, the moat are the models (as an extension of the instruction tuning data)

Yep I feel like they solved the problem that Apple never managed to solve with Siri: How to interface it with apps. Seems like this was an LLM-hard problem

Does anyone have an idea why they are so open about Whisper? Is it the poster child project for OAI people scratching their open source itch? Is there just no commercial value in speech to text?

I think hill-climbing rejects updates that would worsen the objective while gradient descent usually doesn‘t