Whether or not true Artificial General Intelligence is something that can be achieved in the near future is debatable, but the fact is that these companies are all actively trying to build it, and they don’t even have a plan on how it could be controlled once built. We need this shit regulated yesterday. (not to mention all the less existential AI junk that’s already causing lots of harm.)
Notably, a few Chinese companies have been at the forefront of AI video for most of this year, and some experts speculate that the reason is less reticence to train on copyrighted materials, use images and names of famous celebrities, and incorporate some uncensored video sources. As we saw with Stable Diffusion 3’s mangled release, including nudity or pornography in training data may allow these models achieve better results by providing more information about human bodies. HunyuanVideo notably allows uncensored outputs, so unlike the commercial video models out there, it can generate videos of anatomically realistic, nude humans.
including nudity or pornography in training data may allow these models achieve better results by providing more information about human bodies.
It may? So, ok, however they were censoring the data set in SD3 plainly did them no favors. But artists study nudes because it helps them understand how human bodies are put together and move, whatever happens to be draped over them afterward. And these models don’t. They don’t even have a concept that humans exist in three dimensions. They just match images together. So the more source data the better, but does it actually help to put in things you don’t want to generate?
Just a gentle reminder of what “the square of” means; that green line to the left is the square of the input cost of N items. Cost is the vertical axis. Linear scale shown.
What’s worse, that appears to be just the runtime cost. The training cost is N5 or worse, according to my colleagues who use these models.
I don’t care how many NVidea GPU’s you have running beside your newly refurbished 1950’s nuclear plant, it sure as hell ain’t N2.
Ha. As a quantum guy phrased it to me at lunch, at a conference where he was desperately trying to get another job: “quantum is an application waiting for the hardware to catch up.”
A Nature paper from Google with revolutionary claims in AI-enabled chip design was heralded as a breakthrough in the popular press, but it was met with skepticism from domain experts for being too good to be true and for lacking reproducible evidence.
Now, crosschecked data indicate that the integrity of the Nature paper is substantially undermined owing to errors in conduct, analysis, and reporting. Independently, detailed allegations of fraud and research misconduct in the Google Nature paper have been filed under oath in California.
Nature has been slow to enforce its own policies. Delaying retractions of problematic publications is distorting the scientific process. Swift and decisive action is necessary to maintain the integrity and credibility of scientific research.