Hacker Newsnew | past | comments | ask | show | jobs | submit | keonix's commentslogin

If I regurgiate something I read in copyrighted book without proper license that also would be theft, no distinction there.

I'm not distributing my brain, at least same (but probably more restrictive) should apply to models - training is okay, but using and distributing should be limited by copyright


Explaining anything publicly based on my understanding I got reading books would be illegal following this logic. I'm not sure this is how it works.


They want to muddle the distinction between ideas and expression. You can't copyright ideas. Everyone is entitled to copy ideas.


It would not be illegal based on fair use (though you have to be careful there also), but if you try to regurgiate large portions of the book then it would be. And we do know that models regurgiate training material verbatim (Copilot)


Wait until you hear about frankenmodels. You rip parts of one model (often attention heads) and transplant them in another and somehow that produces coherent results! Witchcraft

https://huggingface.co/chargoddard


>somehow that produces coherent results

with or without finetuning? Also is there a practical motivation for creating them?


> with or without finetuning?

With, but it's still bonkers that it works so well

>Also is there a practical motivation for creating them?

You could get in-between model sizes (like 20b instead of 13b or 34b). Before better quantization it was useful for inference (if you are unlucky with vram size), but now I see this being useful only for training because you can't train on quants


> With, but it's still bonkers that it works so well

Ehhhh…


I assume it's because such large context takes lots of memory, so you might as well have smarter model if you are not gonna fit in small vram anyway


Personally, I have found that Mistral 7B (with its native 8K context, and decent results stretched out even more) is performing much better than llama 13B tunes for storytelling, where that long context is really important.

And I think the optimized backends should implement that sliding 16k context soon...

Anyway, point is a huge context really helps certain types of queries, and VRAM usage is reasonable with a 7B model.


Mistral 7B ~ 8 GiB

StableLM 3B ~4 GiB

You could go even lower with smaller quantization if necessary. I personally wouldn't use anything smaller than 7B and Mistral already pushing it in coherence. Overall it depends on your use case, not everyone needs smart models, or large context that sometimes takes half of required memory.

Codellama is also surprisingly good even for non-coding tasks


For what it's worth, I think OP has a point in showing inconsistencies in your logic


> US isn't donating weapons to Ukraine, it's leasing them

Could you kindly support your statement with the source? AFAIK most aid provided to Ukraine is not under lend-lease but donated to them


https://en.m.wikipedia.org/wiki/Ukraine_Democracy_Defense_Le...

"Any loan or lease of defense articles to the Government of Ukraine under paragraph (1) shall be subject to all applicable laws concerning the return of and reimbursement and repayment for defense articles loan or leased to foreign governments."

Can you provide a source for your statement that these were donations?


That's disingenuous. You provided link to a legal framework, but not any use of it. Your source doesn't even mention a single delivery under this act.

> Can you provide a source for your statement that these were donations?

You made the first claim, it's on you to prove it because disproving something is harder than making false claim. But you can have my source anyway.

https://www.politico.com/newsletters/national-security-daily...


Most aid takes the form of cash for humanitarian support, ~disposing of~ donating old hardware to Ukraine, or a loan for the express intent of buying a modern system.

The vast majority of the "billions" given to Ukraine is made up of stuff that we actively are trying to throw away.


IIRC temperature is proportional to velocity squared so increase in simulation speed will result in squared increase in temperature. But even if we account for that we are not dealing with ideal gas in real world, there bound to be collisions that break molecules apart or even start nuclear fusion at certain speeds


Sure! It's just a classification of events that each person has.

Testible prediction? Socially close people will have similar classifications as opposed to randomly chosen ones (I will leave defining metric space over possible classifications as an exercise to the reader)


So it’s completely subjective to you, AKA not anywhere near the testable physics that I was looking for.


> So it’s completely subjective to you

No? My prediction holds even if I didn't exist.

> not anywhere near the testable physics that I was looking for.

And now you are moving the goalpost.


GPL restricts only developers. As an end user restrictions don't even apply to you


Nothing changes, it is still a violation


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: