xyzsparetimexyz

No, everyone just watches Nolan movies because he's seen as a director who can do no wrong. It says nothing about CGI.

Time to get a VPN bucko

Cool poem!

broken statistical text generation software

I find this framing very annoying. Yes I understand why leftist hacker types hate LLMs. I don't disagree that their economics is precarious. I'm also annoyed by ai boosters.

But the whole 'stochastic parrot' thing is also very tired at this point. Idk what best describes the system that is outputting all these mathematical proofs but it is at least doing more than just parroting.

Interesting. This dooes lend credence to the distillation idea. Good for them!

Is Accra the hotspot for AI data contracting?

Train station: :\

Train station, Japan: :D:D:D <3

I think its more a case of, when/if this kind of LLM can be run on a normal desktop then it'll become a big deal. But its also not the kind of work thats economically transformative at all

Cheap is a relative term. How many people would pay $5 to reverse engineer a device in their home?

There's no way to make a profit off reverse engineering home devices so I expect while this is likely to be one of the most groundbreaking impacts of LLMs, it won't see a ton of adoption until token costs come down.

Boring ragebait

What did the bisexuals do to deserve that?

Okay. Who's gonna pay me to work on it? I'm looking for a new job. Can start tomorrow

Even when training both models together in a novel way?

Well yeah. You start on as medium model, it does the digging says says 'this task is too easy/hard/just right for me' and go from there.

Currently. I'm sure that you could make a system where the cache values are a superset C of e.g. models A and B where C is probably bigger than max(A,B) but smaller than A+B

Well yeah, it'd require the models to be loaded on the same system and the cache to be shared between them somehow

Couldn't someone build an adaptive system, where the llm is frequently judging the difficulty of a task and switching to a more/less powerful model?

Its the currency of the future.

It's fairly simple. Sell GPU compute + extra margins as only some GPUs can load the models + extra margins based on how much better closed source models are from open source ones + hopefully reduced cost due to batching

esim

Wow! Does it accept a foreign alipay/wechat pay account?

Stackoverflow did it to themselves by having incredibly unhelpful users

Moores law is dead for memory and other things

I seriously doubt that we'll be running 1TB ram setups any time soon. I was hearing about people running 64GB back in 2014 or so

committed to enslaving the Uyghur people

What?

Any updated Pareto frontier graphs? https://paraplouis.github.io/llm-pareto-frontier/ is quite out of date now.

This is like meat eaters who wish they were vegan but state that they're not mentally strong enough or whatever. Incredibly annoying. Either shut up or convert.

No archive of previous games?

I'm sure I could.

Consumer choices are only going to get you so far