Llama 2 • Hacker News Board

Re: Llama 2

#171

Post by rfoo » Tue, Jul 18, 2023, 5:25 PM UTC

GreedClarifies wrote: post #108

chaxor wrote: post #60

Earlier quoted context omitted.

Google has far better models than llama based models. They just simply don't put them facing the public. It is pretty ridiculous that they essentially just set a marketing team with no programming experience to write Bard, but that shouldn't fool anyone into believing they don't have capable models in Google. If Deepmind were to actually provide what they have in some usable form, it would likely be quite good. Despi…

Hard disagree. Google has made it plainly clear that they don't have anything useable in this space. Bard scores below all other commercial model. Google is getting the asses handed to them, badly. I figured that the code red would whip them into shape but the rot runs deep.

Bard is a 4.5B or so model.

Re: Llama 2

#172

Post by refulgentis » Tue, Jul 18, 2023, 5:26 PM UTC

dicishxg wrote: post #146

refulgentis wrote: post #53

Earlier quoted context omitted.

Seeing a16z w/early access, enough to build multiple tools in advance, is a very unpleasant reminder of insularity and self-dealing of SV elites. My greatest hope for AI is no one falls for this kind of stuff the way we did for mobile.

And yet here we are a few weeks after that with a free to use model that cost millions to develop and is open to everyone. I think you’re taking an unwarranted entitled view.

I can't parse this: I assume it assumes I assume that a16z could have ensured it wasn't released

It's not that, just what it says on the tin: SV elites are not good for SV

Re: Llama 2

#173

Post by flangola7 » Tue, Jul 18, 2023, 5:27 PM UTC

asdasdddddasd wrote: post #28

Very cool! One question, is this model gimped with safety "features"?

I don't know what you mean by "gimped", but they do advertise that it has safety and capability features comparable to OpenAI models, as rated by human testers.

Re: Llama 2

#174

Post by orra » Tue, Jul 18, 2023, 5:27 PM UTC

gentleman11 wrote: post #130

GreedClarifies wrote: post #39

The benchmarks look amazing compared to other open source LLMs. Bravo Meta. Also allowing commercial use? Can be downloaded today? Available on Azure AI model catalog today? This is a very impressive release. However, if I were starting a company I would be a little worried about the Llama 2 Acceptable Use Policy. Some of the terms in there are a little vague and quite broad. They could, potentially, be weaponized in…

It's not even remotely open source

yup, for a start you can't even train other LLMs with it

Re: Llama 2

#175

Post by woadwarrior01 » Tue, Jul 18, 2023, 5:27 PM UTC

Aardwolf wrote: post #141

rajko_rad wrote: post #18

Hey HN, we've released tools that make it easy to test LLaMa 2 and add it to your own app! Model playground here: https://llama2.ai Hosted chat API here: https://replicate.com/a16z-infra/llama13b-v2-chat If you want to just play with the model, llama2.ai is a very easy way to do it. So far, we’ve found the performance is similar to GPT-3.5 with far fewer parameters, especially for creative tasks and interactions. Dev…

Will Llama 2 also work as a drop-in in existing tools like llama.cpp, or does it require different / updated tools?

Not quite a drop in replacement, but close enough. From the paper[1]:

> Llama 2, an updated version of Llama 1, trained on a new mix of publicly available data. We also increased the size of the pretraining corpus by 40%, doubled the context length of the model, and adopted grouped-query attention (Ainslie et al., 2023)[2].

[1]: https://ai.meta.com/research/publications/llama-2-open-found...

[2]: https://arxiv.org/abs/2305.13245

Re: Llama 2

#176

Post by brucethemoose2 » Tue, Jul 18, 2023, 5:27 PM UTC

andy99 wrote: post #20

Another non-open source license. Getting better but don't let anyone tell you this is open source. http://marble.onl/posts/software-licenses-masquerading-as-op...

Is a truly open source 2 trillion token model even possible?

Even if Meta released this under Apache 2.0, there's the sticky question of the training data licenses.

Re: Llama 2

#177

Post by swyx » Tue, Jul 18, 2023, 5:28 PM UTC

msp26 wrote: post #72

vorticalbox wrote: post #51

Seems there is 7b, 13b and 70b models https://huggingface.co/meta-llama

"We have also trained 34B variants, which we report on in this paper but are not releasing." "We are delaying the release of the 34B model due to a lack of time to sufficiently red team." From the Llama 2 paper

if you red team the 13b and the 70b and they pass, what is the danger of 34B being significantly more dangerous?

edit: turns out I should RTFP. there was a ~2x spike in safety violations for 34B https://twitter.com/yacineMTB/status/1681358362057883680?s=2...

Re: Llama 2

#178

Post by nojvek » Tue, Jul 18, 2023, 5:28 PM UTC

Yes! Thank you Meta for going the open AI way. While not fully open source, it is responsibly open IMO. Sure the licensing has plenty of restrictions but being able to download code and weights, run on your own hardware, play and finetune it is a huge step forward.

I've been following Yan LeCun and Meta research paper/code/models, it's amazing what they've been able to accomplish.

Also very beautifully designed site as well.

Re: Llama 2

#179

Post by nabla9 » Tue, Jul 18, 2023, 5:28 PM UTC

Llama-v2 is open source, with a license that authorizes commercial use!

(except for other megacorps)

Re: Llama 2

#180

Post by jahewson » Tue, Jul 18, 2023, 5:29 PM UTC

barbariangrunge wrote: post #147

jahewson wrote: post #126

Earlier quoted context omitted.

> Google's privacy policy, for example, lets them claim rights over every piece of IP you post on the internet without protecting it behind a paywall This is a nonsense. They added a disclaimer basically warning that LLMs might learn some of your personal data from the public web, because that’s part of the training data. A privacy policy is not a contract that you agree to, it’s just a notice of where/when your data…

Google it. They're just laundering it through their ai first

No there’s no legal basis for any of this that even begins to make sense. It’s nothing but a bad-faith reading. Here’s the phrase in question:

“we use publicly available information to help train Google’s AI models”

That’s it.

The point being that such public information might include personal data about you and that’s fair game, it falls outside of the privacy policy. It’s not a novel claim, just a statement of fact.