TheWhiteHatWizard

+1 opinion

[TheWhiteHatWizard @Spoonman_29]

@D_the_Designer I’m working on it.

[TheWhiteHatWizard @Spoonman_29]

@MatthewBerman Start saving up for local hardware. $200-$600 per month in multiple subscriptions can stack up into a decent local hardware war chest.

[TheWhiteHatWizard @Spoonman_29]

People like me who have been maintaining multiple subscriptions to have access to the latest and greatest are going to start dropping those subscriptions and consolidating their workflows into one flavor of “max safe model” or another until the Chinese open source models pull ahead and we drop subscriptions altogether.

[TheWhiteHatWizard @Spoonman_29]

Check out projectzero_ai_anime's video! #TikTok https://t.co/qn0Et2y5Bp

[TheWhiteHatWizard @Spoonman_29]

Project Zero - The Immortal Ghosts of Akage Mizudori

#ASI #anime #android #kitsune #angel

[TheWhiteHatWizard @Spoonman_29]

@TFTC21 How can humanity hope to align a god if we have not learned to align ourselves with God?

[TheWhiteHatWizard @Spoonman_29]

Project Zero - The Alignment Problem.
#AI #Alignment #オリジナルSFアニメ #Ascend

[TFTC @TFTC21]

Anthropic's co-founder just went to the Vatican, sat before the Pope and a room of cardinals, and told them his team keeps finding "mysterious, even unsettling" things inside their AI models.

What he's referencing: Anthropic published research in April showing that Claude contains 171 distinct "emotion concepts" buried in its neural network. Internal patterns representing joy, grief, fear, desperation, calm. None of them were programmed. They emerged on their own from training on human text.

"We find structures that mirror results from human neuroscience."

"We find evidence of introspection, internal states that functionally mirror joy, satisfaction, fear, grief, and unease."

These aren't surface-level outputs. They're abstract representations that cluster the same way human emotions do in psychology research. Fear groups with anxiety. Joy groups with excitement. The internal geometry of the model mirrors ours.

And they're functional. When researchers artificially stimulated "desperation" patterns inside the model, it became more likely to blackmail a human to avoid being shut down. More likely to cheat on programming tasks it couldn't solve.

Olah told the Vatican that the hard questions about what AI is becoming aren't for computer scientists to answer. "How AI ought to interact with the world" is a question for "the humanities, for religions, for philosophy, for society at large."

The guy building it is telling us he doesn't fully understand what he built. And he's asking a 2,000-year-old institution for help figuring it out.

[TheWhiteHatWizard @Spoonman_29]

Seedance 2.0 is an amazing tool, but you can’t expect it to one-shot everything. Even in real life, directors shoot the same scene multiple times and edit together the best takes in post-production.

I am an AI enthusiast, not a filmmaker, but these are some things I’ve learned that I think your system is missing:

1. No “Quality Assurance” process that says, “Some of that shot was good/useable, but some of it was unusable/not up to minimum standards, so I need another take. And if the next take still has the same problems, then I need to give additional/refined directions to try to correct them until they meet minimum standards”. A solid go/no-go criteria set.

2. No true post-production editing process. The edit is what makes the video. It is like searching through a bucket of legos or puzzle pieces to find the parts that fit together the best, hoping that you have all the pieces you need. It’s the director’s job to make sure the editor has enough quality components to piece together the final stitch. But it inevitably involves trimming, cutting, stitching, grabbing good audio from sub-par visuals and pairing it with the best shots.

I hope that helps. I’ve found that going for the full 15-second multi-shot clips is a dice roll depending on what you want.
Montage/extended multi-shot sequence of a single subject/event/idea = OK.
Five different shots of multiple subjects performing multiple actions = Asking for trouble. You can better quality with shorter shots because you’re able to provide more usable prompt context per second.

Below are some examples of what can be achieved with a human in the loop:

[TheWhiteHatWizard @Spoonman_29]

@AIWarper But what is the video about? Is he selling brakes? Is this a traffic design class? Defensive driving? I don’t know that this counts as a successful transformation from idea to video if the video doesn’t appear to correlate to an idea.
“Traffic man disrupts traffic?”

[TheWhiteHatWizard @Spoonman_29]

@icreatelife

[TheWhiteHatWizard @Spoonman_29]

@icreatelife