Bambu at it again! They sent a C&D to a developer who forked their AGPL licensed software. Can we stop supporting that company? Please, youtubers, take note.
Context: github.com/jarczakpawel/Orca…
Luis Rossmman coverage: youtube.com/watch?v=jIbpQtoz…
Another nice TED talk by @SashaMTL . Unfortunately it won't change much, but still. Only when it will be too late we will finally do something. Humans are great at doing the right thing, eventually.
youtube.com/watch?v=Bl-vPf_I…
Isn't Nano Banana fun? A lot of hallucination... but considering the little effort, it's impressive.
So tired of this DGX Spark nonsense. It's an overpriced hardware released too late. Slower and more expensive than any integrated alternative (apple or AMD). The only advantage right now is the nvidia ecosystem but we all know that they'll stop supporting it in 6months.
More experimenting. This is done by taking 2 vectors and trying to find a third in between. I also like to add noise to the embeddings, it kinda shakes the training off from the model a bit. Vector 1 is fantasy fire and vector 2 is vintage poster. PS: not comfy
More experiments in the latent space. These are done by averaging very different textual embeddings and adding a different grade of noise to the resulting tensor, not the actual latent but directly to the embeddings. I like the abstract but still quite controllable results























