All posts written purely with Natural Stupidity ML/3D/Gen AI. Building things at @artflow_ai | hiring
Joined August 2018
- Tweets4.8K
- Following640
- Followers790
- Likes4.9K
Pinned Tweet
AI for Game Art with a scary title :) Overview of the current tools including @Artomatixltd @Allegorithmic Alchemist, @unity3d AR facial perf capture and Kinematica, and countless other examples from CV and ML research bit.ly/2TUZPKA
found a picture of a little possum made out of felt, loved the look and asked video model to animate it (no specifics, no 'leading the witness' and most certainly nothing about a cat!).
Now I want a whole 150+ minute movie about that anxious little possum.
Here's a game where you can fight against a simulated fruit fly connectome in a 1v1 dogfight.
Play as Major Buzz, or Leutnant von Zip.🪰
@FlyWireNews I hope that's the intended use of your research.
links in comment.
#flywire #aigamedev
For creatives "It was the best of times, it was the worst of times ..."
job insecurity, conceptual inbreeding and disruption happening in parallel to new opportunities and a Cambrian explosion of tools.
it was the age of wisdom, it was the age of foolishness ...
I let Claude play my VR music tool. First tests: Balinese gamelan, medieval polyphony, footwork, and Japanese jazz-funk @claudeai
Glad the recent LLMs finally 'clicked' with a lot of creative people.
I've been writing my own tools, scripts for ages, but evidently some threshold has been finally crossed around Astra and Claude 5.x
Against so much negativity about AI, this is the positive side.
This is the first time my mind has been completely blown away at what Claude could do.
I have so many conflicting feelings and thoughts. #claude #opus Not ONLY animated our design.
It built it all in #blender3d
What does this mean?!
(Fellow artists please don't hate me! we should all talk about this)
The fun part about training robots to do boxing is that it has to go through a "two drunk dudes trying to land a single punch" phase.
I really like how the responsible labs, you know the only ones we can trust with this tech, constantly lose control over those models, but literally a billion pedestrians using the same models, including criminals, psychopaths, terrorists, narco cartels, CS majors that lost the will to live, unhinged lunatics, cult members, or just simpletons that clicked allow one too many times, did not manage to accidentally, or intentionally hack the Australian government.
"Never attribute to AI malice that which is adequately explained by human stupidity"
Lets see which of the 3827 "endless video" projects will take off with avg viewership -> 0
Content needs to mean something to the viewer. Consumption in isolation is feasible, but the content would need to be personalized, not a random roll of the dice.
note: not picking on this one in particular. They will all face the same challenge.
There were a few versions circulating of Plato's cave allegory that included AI, but they universally missed a few elements.
Multi-modality breaks this 'metaphor', so it applies to pure text training corpus.
Interestingly NanoBPro turned out to be best suited to deliver this.
The new Muse Image was surprisingly good but still struggled to assemble it all spatially and it overused it's licentia poetica, adding unwanted annotations.
