HN user
Yeah, but this is partly due to there being a shortage of entry level GPUs for consumers. NVIDIA has literally stopped manufacturing them.
There are massive numbers of data centre GPUs sitting in hyperscaler warehouses waiting to be deployed in a data centre. They may never be deployed because there’s more GPU than DC space and you want your most efficient GPUs in the active slots.
KV cache compression, so how much memory the model needs to use for extending its context. Does not affect the weight size.
Isn’t there also basically 0 American DRAM?
The 80/20 rule always wins
I just wanted deterministic outputs and was curious how you were doing it. Sounds like probably temp = 0, which major providers no longer offer. Thanks for your response.
Wait sorry how did you use and expose seeds? That’s the most interesting part of your post
It might have been explicitly targeted, but they did say that there were older versions of Notepad ++ with ""insufficient update verification controls" so it might have just been there was only one subset of users actually susceptible to this.
I believe the what the parent comment was referring to is the advice not to praise character, but instead praise hard work.
“You’re so smart” leaves room for failure when they encounter something that challenges their image of being smart. Praising the amount of effort they put in is not something that is taken away or challenged regardless of the outcome.
I think it’s also largely driven by the apparently cheapness of turning the CapEX of server buying to the OpEX of cloud renting. Less up front investment and auditing/access controls for SoC2 compliant are so much easier m.
Unfortunate name collision on that one
Obviously its not at the scale of the top auto-regressive models yet but there are some OSS models https://github.com/dllm-reasoning/d1
It may be that it was time for the hardware that was previously running Arxiv to be retired and this is just another Capex -> Opex decision being made by so many tech companies.
I'd like to know if GCP is covering part of the bill? Or will Cornell be paying all of it? The new architecture smells of "[GCP] will pay/credit all of these new services if you agree to let one of our architects work with you". If GCP is helping, stay tuned for a blog post from google some time around the completion of the migration with a title like "Reaffirming our commitment to science" or something similarly self affirming.
I believe they are using scalable TTC. The o3 announcement released accuracy numbers for high and low compute usage, which I feel would be hard to do in the same model without TTC.
I also believe that the 200$ subscription they offer is just them allowing the TTC to go for longer before forcing it to answer.
If what you say is true, though, I agree that there is a huge headroom for TTC to improve results if the huggingface experiments on 1/3B models are anything to go off.
The LMSYS leaderboards are crowdsourced and would be hard to fake, it showing a pretty strong performance in terms of human preference.
Captcha solvers as a service are already well developed. The end result is going full circle to in person applications only.
Tragedy of the commons at work once again
There’s black sand! Volcanic sand from Iceland is perfectly black and would be a great way to distinguish them
I think right now they lose more money with each user. But maybe their value lies in training data
Not as much as meta, no. But AI21 labs is partnered with Amazon and did a ~$200M funding round last year IIRC so still plenty of funds for training big models
As another commenter said, this has no GGUF because it’s partially mamba based which is unsupported in llama.cpp
We’ve all had moments like that
Check out Nougat from meta
Sell them and invest the money