• 1 Post
  • 130 Comments
Joined 3 years ago
cake
Cake day: August 27th, 2023

help-circle








  • The SR-71, designed largely to spy on Soviet countries, was made of titanium sourced from Soviet countries.

    In the decades since then I would wager the arms race against “straw buyers” has favored the buyers. Especially when it’s just bare microchips. It’s not a dinky Polish auto-repair shop ordering five thousand Tomahawk missile nose cones. Some of these smuggling operations happen by accident, because the chip someone needs happens to be used in a fancy toy car. Or if you want to get conspiratorial, a toy car can be designed to use any chip in a missile, without raising a single eyebrow.

    Amateur rocketry enthusiasts have run into tracking problems because commercial GPS and motion-tracking hardware is designed to shut off past a certain speed. That’s how mundane and widespread some of these weapon components are: if you can jailbreak a TomTom and a Wiimote, you’re left making the part that explodes slowly and the part that explodes quickly.

    And drones are still scarier.







  • I again submit the last two years where model collapse did not happen. The doom-and-gloom predictions - some rather gleeful - plainly missed the mark. The proliferation of generated content has not in fact ruined the content generators, and it’s sure not because we’re any good at marking generated content. Early symptoms went away entirely and the problem has been practically addressed.

    As for “unlearning,” universality is why it’s a made-up problem. Nobody loudly complains that x-rays make doctors worse at feeling around for lumps.


  • That’s a lot of “could” and “will” from an article a year old, primarily about concerns from two years ago, while image models to-day keep getting smaller and better. They didn’t find a second internet’s worth of JPEGs. Better training on the same data, or even better labels on less data, beats a simple obsession with scale.

    Yes, photocopying a photocopy will degrade, but diffusion is a denoising algorithm. Un-degrading an image is its central function. ‘Make it look less AI’ is how you get generative adversarial networks.

    Anyway, the grim truth is that the central concern is mistaken. Training data for cancer screening does not require the patient lived.


  • The doctors were better, until someone yanked the tool away. That’s how every tool works! Even going from a handsaw to a table saw and back will make you lose some skill with the handsaw, because your brain focused on higher-level goals and finer motions. That’s not proof a table saw is bad for woodworking. The problem is “and back.”

    since apparently AI can’t feed into AI without collapse

    Have you checked on that narrative? It’s been a while. Things stopped getting yellow. Improvements continued.