Perhaps not, I’m just playing with the free version.
You can try to replicate the examples in the blogpost to see if you have the new model.
Dogs (Loyal Project) - eh? What does this have to do with rapamycin? Current AI efforts are a clown show. What’s the point, when you can’t trust anything this thing puts out? Old rule of computing applies: GIGO - Garbage In Garbage Out. Personally I lack the time and inclination to deal with garbage.
I don’t recognize those two, who are they?
Matt Kaeberlein and Peter Attia?
Uploading images of both in the prompt might add the facial features.
I think based on this model, it looks like image generation is pretty much solved soon (able to create or recreate any image possible). That it doesn’t know what should be in the text isn’t surprising to me, the text generation part of the model might not have that many parameters, e.g a smaller GPT 3 equivalent.
Video won’t be far off, then after that real time video (like video game or worlds).
Supposed to be Attia and Kaeberlein. I took a couple of photos off the internet, but it’s hard to keep distinctive features while become elongated stone heads.
I was thinking the bald guy looks like Shrek or Joe Rogan
There is literally zero resemblance to those two. As to keeping distinctive features, that’s the essence of the talent of human caricaturists. The ability to severely distort the form and still have it be recognizable. I bet you a human cartoonist illustrator would be able to draw these stone heads so you’d instantly recognize whom they’re supposed to represent. Spectacularly clear that AI has a very long way ahead of it in this regard.
Read the full report:
U.S.'s AI lead over China rapidly shrinking: Stanford report
https://www.axios.com/2025/04/07/china-ai-race-stanford-report
AI race in 2025 is tighter than ever before
State of the industry report also shows that 2024 was a breakthrough year for small, sleek models to rival the behemoths.
https://www.nature.com/articles/d41586-025-01033-y