😮 https://t.co/Y8WOODbhIT

#GenerativeAI for Txt2Vid being unlocked. Less than a year since txt2img turned our world upside down.

Thanks to Shutterstock for involuntarily sponsoring it. You can’t buy this kind of publicity. 🦙 https://t.co/ucohEFDBEB

@lvminzhang I never generated any image with adversarial noise. Am I missing out? When did it happen to push you to build this?

T2I Adapters for style transfer are the next big thing after ControlNet. I was supposed to test them this weekend but higher duties took all my time. Now. Can somebody make me a man completely made of tarantulas before I try myself next weekend? https://t.co/glrqsgnkcV

@Hairyears @Ninetail_foxQ Which is exactly why I started writing Synthetic Work, @Hairyears. I did not want to talk about the AI tooling or techniques. I wanted to document the impact of AI on human labor, the economy, and our society. Because this trajectory concerns me greatly.

I find amusing that we all talk about all things that #GPT4 can and cannot do and how it will impact humanity as if this is a moment frozen in time. If you look at the timeline, we have a new *major* AI model trained every every (possibly as part of a tick-tock release cycle)

The preoccupation should not be about GPT4 limitations but how many of those will disappear with GPT5 in 12 months. Unless OpenAI either hits a wall (as @GaryMarcus has been saying all along) or they see something too concerning to keep the established pace.

The problem is that now OpenAI is not alone anymore. It’s with Microsoft. And Microsoft wants its monopolistic profit back, and a place in history as winner far bigger than one as the creator of Windows or Office.

👀 https://t.co/h12yaHLyG9

@amli_art Yes. I am referring to the expected model/s that Stability AI will announce next week. I assume they will announce @DeepFloydIF, but it’s also possible they will announce a truly open LLM (LLaMA is not).