【Hacker News搬运】Meta Llama 3

hackernews

Title: Meta Llama 3

Meta Llama 3

Text:

Url: https://llama.meta.com/llama3/

Post by: bratao

Comments:

dang: See also <a href="https://ai.meta.com/blog/meta-llama-3/" rel="nofollow">https://ai.meta.com/blog/meta-llama-3/</a>and <a href="https://about.fb.com/news/2024/04/meta-ai-assistant-built-with-llama-3/" rel="nofollow">https://about.fb.com/news/2024/04/meta-ai-assistant-built-wi...</a>edit: and <a href="https://twitter.com/karpathy/status/1781028605709234613" rel="nofollow">https://twitter.com/karpathy/status/1781028605709234613</a>

dang: 另请参阅<a href=“https://；&#x2F；ai.meta.com#xx2F；blog&#x2F！meta-llama-3#xx2F”rel=“nofollow”>https://&#x2F；ai.meta.com&#x2F；博客&#x2F；meta-llama-3x2F</a> 和<a href=“https://；&#x2F；about.fb.com&#x2F：新闻&#x2F！2024&#x2F？04&#x2F“meta-ai-assistant-build-with-llama-3&#x2F”rel=“nofollow”>https://&#x2F；关于.fb.com；news&#x2F；2024；04；meta-ai助手构建的wi</a> edit:和<a href=“https://；&#x2F；twitter.com&#x2F：karpathy#xx2F；status&#x2F！1781028605709234613”rel=“nofollow”>https://&#x2F；twitter；karpathy；status；1781028605709234613</a>

whereistimbo: How do they plan to make money with this? They can even make money with their 24K GPU cluster as IaaS if they want to. Even Google is gatekeeping its best Gemini model behind.<a href="https://web.archive.org/web/20240000000000*/https://filebin.net/nnwafdpsyxynuq5z/p_pic.zip" rel="nofollow">https://web.archive.org/web/20240000000000*/https://filebin....</a>
<a href="https://web.archive.org/web/20240419035112/https://s3.filebin.net/filebin/04bb7233f6d9d040a6ba22706400b3d4d2c09ff14ddf0c3d190851559f564875/f99fdf2dcba024f178b2ae357f3c5086fe7ae0f773f205a8ded8716f79f15675?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=7pMj6hGeoKewqmMQILjm%2F20240419%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20240419T035111Z&X-Amz-Expires=300&X-Amz-SignedHeaders=host&response-cache-control=max-age%3D300&response-content-disposition=filename%3D"p_pic.zip"&response-content-type=application%2Fzip&X-Amz-Signature=6a35f8e7a20dd432cb0a331398defeec91977775e7b9600208b9dccea5eee805" rel="nofollow">https://web.archive.org/web/20240419035112/https://s3.filebi...</a>

whereistimbo: 他们打算如何用这个赚钱？如果他们愿意的话，他们甚至可以用24K GPU集群作为IaaS赚钱。甚至谷歌也将其最好的Gemini模型抛在后面 <a href=“https://；&#x2F；web.archive.org&#x2F！web&#x2F：20240000000000*&#x2F”https://；#xx2F；filebin.net&#x2F，nnwafdpsyxynuq5z&#x2F；p_pic.zip”rel=“nofollow”>https://&#x2F；web.archive.org&#x2F；web；20240000000000*x2F；https:&#x2F&#x2F；filebin</a><a href=“https://；&#x2F；web.archive.org&#x2F！web&#x2F：20240419035112&#x2F”https://；#xx2F；s3.filebin.net&#x20F；filebin&#x20F：04bb7233f6d9d040a6ba222706400b3d4d2c09ff14ddf0c3d190851559f564875&#x209fdf2dcba024f178b2ae357f3C5086fe7ae0f773f205a8ded8716f79f15675？X-Amz-Algorithm=AW S4-HMAC-SHA256&amp；X-Amz-Credential=7pMj6hGeoKewqmMQILjm%2F20240419%2Fus-east-1%2Fs3%2Faws4_request&amp；X-Amz-Date=2020419T035111Z&amp；X-Amz-Expires=300&amp；X-Amz-SignedHeaders=主机和响应缓存控制=最长使用期限%3D300amp；响应内容处置=文件名%3D%22p_pic.zip%22&amp；响应内容类型=应用程序%2Zip和X-Amz-Signature=6a35f8e7a20dd432cb0a313398defeec91977 775e7b9600208b9dccea5ee805“rel=”nofollow“>https:&#x2F&#x2F；web.archive.org&#x2F；web；20240419035112；https:&#x2F&#x2F；s3.filebi</a>

bbig: They've got a console for it as well,
<a href="https://www.meta.ai/" rel="nofollow">https://www.meta.ai/</a>And announcing a lot of integration across the Meta product suite,
<a href="https://about.fb.com/news/2024/04/meta-ai-assistant-built-with-llama-3/" rel="nofollow">https://about.fb.com/news/2024/04/meta-ai-assistant-built-wi...</a>Neglected to include comparisons against GPT-4-Turbo or Claude Opus, so I guess it's far from being a frontier model. We'll see how it fares in the LLM Arena.

bbig: 他们；我也有一个控制台，<a href=“https://；&#x2F；www.meta.ai&#x2F”rel=“nofollow”>https://&#x2F；www.meta.ai&#x2F</a> 并宣布在Meta产品套件中进行大量集成，<a href=“https:&#x2F；&#x2F关于.fb.com.&#x2F新闻&#x2F 2024&#x20F；04&#x2F元-ai-assistant-build-with-llama-3&#x2F”rel=“nofollow”>https://&#x2F；关于.fb.com；news&#x2F；2024；04；meta-ai助手构建的wi</a> 忽略了包括与GPT-4-Turbo或Claude Opus的比较，所以我想它是；It’这远非前沿模式。我们；I’我来看看LLM竞技场的比赛情况。

nathanh4903: I tried generating a Chinese rap song, and it did generate a pretty good rap. However, upon completion, it deleted the response, and showed
> I don’t understand Chinese yet, but I’m working on it. I will send you a message when we can talk in Chinese.I tried some other languages and the same. It will generate non-English language, but once its done, the response is deleted and replaced with the message

nathanh4903: 我试着制作一首中国说唱歌曲，它确实制作了一首相当不错的说唱歌曲。然而，在完成后，它删除了响应，并显示&gt；我还不懂中文，但我正在努力。当我们能用中文交谈时，我会给你发信息 我试过其他一些语言，也一样。它将生成非英语语言，但一旦完成，响应将被删除并替换为消息

eigenvalue: I just want to express how grateful I am that Zuck and Yann and the rest of the Meta team have adopted an open approach and are sharing the model weights, the tokenizer, information about the training data, etc. They, more than anyone else, are responsible for the explosion of open research and improvement that has happened with things like llama.cpp that now allow you to run quite decent models locally on consumer hardware in a way that you can avoid any censorship or controls.Not that I even want to make inference requests that would run afoul of the controls put in place by OpenAI and Anthropic (I mostly use it for coding stuff), but I hate the idea of this powerful technology being behind walls and having gate-keepers controlling how you can use it.Obviously, there are plenty of people and companies out there that also believe in the open approach. But they don't have hundreds of billions of dollars of capital and billions in sustainable annual cash flow and literally ten(s) of billions of dollars worth of GPUs! So it's a lot more impactful when they do it. And it basically sets the ground rules for everyone else, so that Mistral now also feels compelled to release model weights for most of their models.Anyway, Zuck didn't have to go this way. If Facebook were run by "professional" outside managers of the HBS/McKinsey ilk, I think it's quite unlikely that they would be this open with everything, especially after investing so much capital and energy into it. But I am very grateful that they are, and think we all benefit hugely from not only their willingness to be open and share, but also to not use pessimistic AI "doomerism" as an excuse to hide the crown jewels and put it behind a centralized API with a gatekeeper because of "AI safety risks." Thanks Zuck!

eigenvalue: 我只想表达我是多么感激Zuck和Yann以及Meta团队的其他成员采用了一种开放的方法，并分享了模型权重、标记器、训练数据信息等。他们比任何人都要对像llama.cpp这样的开放研究和改进的爆发负责，现在可以让你在消费硬件上本地运行相当不错的模型，从而避免任何审查或控制 我甚至不想提出与OpenAI和Anthropic（我主要将其用于编码）实施的控制相冲突的推理请求，但我讨厌这种强大的技术被隐藏在墙后，并由守门人控制如何使用它的想法。显然，有很多人和公司也相信开放方法。但他们不；t拥有数千亿美元的资本和数十亿美元的可持续年度现金流，以及价值百亿美元的GPU！所以它；当他们这样做的时候，影响力要大得多。它基本上为其他人设定了基本规则，所以米斯特拉尔现在也觉得有必要为他们的大多数模型发布模型权重 无论如何，Zuck没有；我不必走这条路。如果Facebook是由“；专业的”；HBS/F的外部管理人员；麦肯锡，我认为它；他们不太可能对一切都如此开放，尤其是在投入了如此多的资本和精力之后。但我非常感谢他们，并认为我们都从他们不仅愿意开放和分享，而且不使用悲观的人工智能中受益匪浅；末日论”；作为隐藏皇冠上的珠宝并将其放在一个有看门人的集中式API后面的借口；人工智能安全风险&quot；谢谢祖克！