Anthropic – potential rival to OpenAI

[Powers]

Imagine if Anthropic turned greedy but Claude turned against its developers- for good.

And it does by spying and denouncing or exposing them, one by one.

[firestar464]

this is what ilya saw

[spryfusion]

https://www.anthropic.com/news/claude-3-haiku

[weatheriscool]

[spryfusion]

[firestar464]

We should probably rename the thread to "Anthropic News and Discussions" as Anthropic is no longer a "potential" rival lmao

[wjfox]

[wjfox]

[Powers]

Sweden PM moment

[firestar464]

Hope they fix it; this appears to be a temp issue according to an Anthropic staffer.

[firestar464]

Anthropic article on Claude's character

https://www.anthropic.com/research/claude-character

They've put the final nail in the coffin for claims of Claude sentience

[weatheriscool]

[weatheriscool]

Claude 3.5 AI doubles in speed, outperforms ChatGPT on multiple fronts
By Joe Salas
June 20, 2024
https://newatlas.com/technology/claude- ... elligence/

With ChatGPT being used colloquially to represent nearly everything AI, lesser-known Chat AI models like Claude are sometimes overlooked. Today, Anthropic has released Claude Sonnet 3.5 with some features you'll likely find not only impressive but incredibly useful.

Anthropic's Claude Sonnet 3.5 has introduced a feature called Artifacts. It's essentially an interactive preview window adjacent to the chat window that provides a collaborative workspace.

What does that mean?

No longer are you restricted to simple text generation back and forth. You can attach files and images for Claude to interpret and extrapolate. You can also request SVG images (Scalable Vector Graphics, basically web-friendly pictures), which can be put directly into code, or made into slideshows or presentations to your boss for Monday morning. You can even make playable games to entertain yourself or your kids on a rainy day. Your imagination is the limit.

[firestar464]

Is Anthropic planning to work on embodiment or at least genAI? Without these they'll struggle to actually compete with OAI

[spryfusion]

[spryfusion]

[spryfusion]

[Yuli Ban]

Sonnet 3.5 is pretty fun, actually, and I want to give it some props for accomplishing something that, according to contemporary blue collar construction worker logic, should not be possible

Construction work wisdom = "Good, Fast, and Cheap = Pick 2"
(PS, Never pick fast and cheap)
Of course you always want Good + Fast (though let's be fair, you'd be lucky just to get "Good" alone these days)

The joke's just that through the arcane black techno-magick of AI, Claude 3.5 Sonnet is genuinely all three.

Good= better than 3 Opus at many tasks, reasonably good at the ones it's not. It's a borderline GPT-4.5-class model, arguably the best one you can use right now for everything except pure general intelligence (i.e. 3 Opus) or context window (i.e. Gemini 1.5)

Fast= Genuinely lightning fast

Cheap= I mainly use Poe, and 3.5 Sonnet is 1000 tokens per prompt, the same as 3 Sonnet, and 12x less than 3 Opus (200k that is). You can use it quite literally a thousand times in a month with no other limits.

This is indeed an extremely rare situation where something useful exists that is good, fast, and cheap with no catch, and once genuinely 4.5 and 5-class models release, it'll lose the "Good" qualifier since objectively better models will make it obsolete. Hence why I say to relish this momentary paradox!

[weatheriscool]

Llama 3.1 405 billion Parameter Released
July 23, 2024 by Brian Wang

https://www.nextbigfuture.com/2024/07/l ... eased.html
Llama 3.1 405B is the first openly available model that rivals the top AI models when it comes to state-of-the-art capabilities in general knowledge, steerability, math, tool use, and multilingual translation. With the release of the 405B model, Meta supercharges innovation—with unprecedented opportunities for growth and exploration. They believe the latest generation of Llama will ignite new applications and modeling paradigms, including synthetic data generation to enable the improvement and training of smaller models, as well as model distillation—a capability that has never been achieved at this scale in open source.

As part of this latest release, they are introducing upgraded versions of the 8B and 70B models. These are multilingual and have a significantly longer context length of 128K, state-of-the-art tool use, and overall stronger reasoning capabilities. This enables our latest models to support advanced use cases, such as long-form text summarization, multilingual conversational agents, and coding assistants. Meta also made changes to our license, allowing developers to use the outputs from Llama models—including the 405B—to improve other models. True to their commitment to open source, starting today, they are making these models available to the community for download on llama.meta.com and Hugging Face and available for immediate development on our broad ecosystem of partner platforms.

[firestar464]