naillo

This is gonna be great to finetune on. There's only so many boards of canada/aphex twin songs out there but I wish there were more and this will let us generate more.

I keep thinking back to when we didn't have stabilityai and it was just google and meta teasing us with mouth watering papers but never letting us touch them. I'm so thankful stability exists.

Scientific sleuths spot dishonest Spellcheck use in papers

There probably is a cool mix of both that is better than either one separately. I'm thinking something like

  func = llm('function that sorts two input argument lists')

where llm calls openai or a local llm (cached for later use). This way you don't lose the benefits of a coding interface (e.g. make all code with openai and the maintainability mess that can come with that). And you get readability through the prompt etc etc. (I mean this project is sort of in that direction already.) It's basically like writing code with a framework that is 'filled out' automatically by a coworker.

Worth being creative with ideas like this at least.

Literally gaussian splatting (and nerfs)

I'll be quiet and good

The things people are able to do these days with low rank approximation is crazy

Fun to imagine the lives of the developers making these things. Hardware/software coding in 90s japan.

The lockin attempts begin

7b runs on my 4gb vram machine (8gb memory). I.e. quantization helps a lot too

The most plausible explanation I've seen (other than the carmack 'sudden grokking' beyond the cutoff idea) is that they're planning to release llama3 sooner than later with some arcitecture changes for even better performance, so it makes sense to dedicate resources there instead.

Python etc drama has historically been strong arguments about the future of various parts of the language. This is way different, this is bickering, personal attacks etc. I have nothing against healthy criticism and debate focused on making the project stronger, but this is unprofessional and not something I want to be adjacent to.

I'm assuming it is unserious, the use of the word "fiasco" is a tell. The fact that it's unserious but still made into a big fuss is an even bigger warning sign in my eyes to steer clear.

Makes me want to distance myself from ever delving into the language knowing this is the drama the culture is engaging in.

A lot of people were just born into this too. People under 27 or so have never known anything other than this pretty much.

Asking to send notifiactions straight away on page load is a huge red flag.

OpenAI had access to the twitter database since forever ago anyway with the old terms of service.

xkcd on the HN front page, tech is healing

That's true for driverless cars but there's many areas where you don't need 100% to be useful. Stable diffusion and LLMs are 90% there but still very useful and cool.

8x4090 machines should be a lot cheaper than that. 1599*8=12,792. I.e. closer to 16x4090 machines. Sure labor is worth something but he has it in cash, and also there's barely a lot of people (yet) paying high prices for research (non ML ops) labor hours for this yet so the salary wouldn't be that high anyway.

Just a pet peeve when ppl under exaggerate how many gpus/flops a dollar can buy (if you're slighly smart with your money).

A16z is getting 'mad respect' from me for taking this line of the 'ai wars' and generally their attitude on all this. Really cool that they're going straight into 2010 era startup energy on this new wave. Idk I just find them really cool and inspirational.

What is tinder doing to facilitate this? I assumed bots on there were just financial scam bots not of the trafficking kind

What a crazy choice of background color. On the other hand it kinda signals that the developers themselves designed it which is generally a good sign of other characteristics of a project.

I think it's a generally practiced strategy at this point to spew out blog dev posts on company blogs to build SEO or act as an ad, regardless of quality.

This feels like bordering on flamebait

Que substack becoming medium and 'stacksub' popping up with a new clean interface to take the users. Rinse and repeat and we never learn.

Tangent but if you 'value' your time but don't use the time to entertain yourself I'm not sure you're using it right.

You can probably beat the game in two weekends though, just FYI. I haven't felt that much need to replay it since then (though I'll probably try Space Age) but it's great fun.

It's funny how this is faster and more responsive than my actual windows desktop window manager experience (speed of opening apps etc).

The only one I subtly disagree with is "comments should be substantive". What it discourages I think is comments like "thanks" or other really 'unsubstantive' comments. It's true that maybe it adds noise, and in many cases are maybe supposed to be inferred without explicitly saying. But I think discouraging this slightly leans behaviour towards snark vs not. (If you see comments like "thanks" you're less likely to be snarky than if you see 'substantive' but maybe too harsh critiques in the comments that appear because "cool project!" isn't allowed.)

Personally I like to make it a point to break this rule from time to time to reduce this pattern.