You mean you feel that R1 is overhyped compared to e.g. o1? It does appear that it's not quite on o1 level, but I think what made it a selling point for people was that it was 100% and showed its "work" (thinking process). More people got the taste of a reasoning model because of this, hence the current hype I think.
Yeah, mainly that R1 will prove to be somewhat less amazing than it first seemed, although I don't have any specific mechanism by which I envision this bubble bursting (if it does). By all accounts, though, it seems impressive... just maybe not quite THAT impressive, you know?
When you say "competitive" - are you talking about the actual examples I've shared in my post above? Because to me it's very clear that DeepSeek results are terrible compared to even SD3 and DALL-E 3.
I went through SD 1.1 -1.5 via automatic1111 during the time that it arrived. I have subsequently bought a 3090 to do AI inference. and pondering a 64GB Mac Mini for the same. So my idea of competitive is different to the norm.
Yes, I am talking about the images you posted, the samurai with a blaster is more on point, with a sword. Though I prefer the other one, as I like the city scene, and it's not so busy with a drone, etc.
The cat and dog is stylistic choice. again I prefer the first image.
the third image is crap, but text is a "meh?" for me. though it does attempt a offset card unlike the others, and the last image has dodgy text on the garland in the background. Live by the sword die by the sword.
I had the same thought about R1, although I know there's at least some fire to this smoke.
You mean you feel that R1 is overhyped compared to e.g. o1? It does appear that it's not quite on o1 level, but I think what made it a selling point for people was that it was 100% and showed its "work" (thinking process). More people got the taste of a reasoning model because of this, hence the current hype I think.
Yeah, mainly that R1 will prove to be somewhat less amazing than it first seemed, although I don't have any specific mechanism by which I envision this bubble bursting (if it does). By all accounts, though, it seems impressive... just maybe not quite THAT impressive, you know?
Not a hot take, but a level take, a rational take, a quantitative take, a unhypey take. Cool take man
It's a chill take, bro. So chill.
Sight unseen, (I gave up on image diffusion 2 years ago) It looks competitive for Deepseek. Is it free? Because Dall-E isn't, neither is Midjourney
When you say "competitive" - are you talking about the actual examples I've shared in my post above? Because to me it's very clear that DeepSeek results are terrible compared to even SD3 and DALL-E 3.
But yes, it's free to use: https://huggingface.co/spaces/deepseek-ai/Janus-Pro-7B (I also show how it works in the guide above).
But so is DALL-E 3 via Microsoft Designer: https://designer.microsoft.com/image-creator?scenario=texttoimage
And so is Stable Diffusion 3 Medium: https://huggingface.co/stabilityai/stable-diffusion-3-medium
And even some of the much better models like Imagen 3 (Google) and Ideogram are free:
Ideogram: https://ideogram.ai/
Imagen 3: https://labs.google/fx/tools/image-fx
I went through SD 1.1 -1.5 via automatic1111 during the time that it arrived. I have subsequently bought a 3090 to do AI inference. and pondering a 64GB Mac Mini for the same. So my idea of competitive is different to the norm.
Yes, I am talking about the images you posted, the samurai with a blaster is more on point, with a sword. Though I prefer the other one, as I like the city scene, and it's not so busy with a drone, etc.
The cat and dog is stylistic choice. again I prefer the first image.
the third image is crap, but text is a "meh?" for me. though it does attempt a offset card unlike the others, and the last image has dodgy text on the garland in the background. Live by the sword die by the sword.