In the AI race, so far I’m hoping, assuming what they say about them is true, that Anthropic’s Claude wins. I’ve read that the founders were former OpenAI employees who were concerned about macrohard funding and that AI should value safety first and foremost. That line of thinking is something you never see anymore when large companies try to financially back new tech.
Probably the biggest downside is that the article I found talking about them is that they mention getting funding from gøøgl€, which is always a little worrying.
Didn’t OpenAI start getting massive investments from Microsoft and then became ClosedAI?
Edit: Not that I necessarily think this will happen here. If anything, Stable Diffusion might run much better on ARC cards (or whatever their Neural Processing Unit becomes) and give a viable alternative to Nvidia
OpenAI is known for ChatGPT, which is a text generator that works for things like writing a haiku. Though OpenAI also has a image generating AI (DALLE), it doesn’t seem nearly as popular as its competitors Midjourney and SDXL. People just assume ChatGPT is OpenAI’s main product, but they are actually competing on everything.
I mentioned to someone that I ask ChatGPT things all the time and they were like, “Don’t you know it doesn’t actually know facts? It just spews bullshit that sounds plausible.”
The joyous thing for me is that’s why I’m using it: To generate plausible sounding nonsense for dungeons and dragons. That, to me, has been one of the biggest use cases for me. Name generation is fantastic through it. “List 10 suggestions for epic sounding names for a tavern built into a cliffside in a deep elven rain forest” and then work shopping it from there.
As a programmer, I also make pretty consistent use of GitHub Copilot… Because half of programming is boiler plate that LLMs are really good at generating. Super useful for explaining what kind of statically defined array I want without having to type out the whole thing myself. Or, and I think this is my favorite use, any time I need to translate from one data format to another, just describing my input and my desired output gets me a great starting point that I can refine.
I’m also a programmer. I’ve found it’s pretty useless except for code that is very repetitive (test cases) or for documentation… But been there it’s a coin flip as to if I’ll have to go in and correct it.
And there’s no indication that it’ll ever be better than that tbh. No matter what articles on MSN say.
At 49% ownership and being 100x the value of Open AI that is effectively the same as full control. Open AI cannot blink without Microsoft getting right of first refusal.
Good, it’s going to be important for this tech to be open. Stability has been hurting for support in comparison to OpenAI, Meta, and Google. Hopefully this can really help them moving forwards.
StableLM is a new open-source language model released by Stability AI. The Alpha version of the model is available in 3 billion and 7 billion parameters, with 15 billion to 65 billion parameter models to follow.
Add comment