NewOldGuard
@NewOldGuard@lemmy.ml
- Comment on To USB, or not to USB 2 weeks ago:
It’s more akin to clicking on every ad you see on the web. Sure you might not get a virus from lots of them but the risk is real and it’s good practice to just not do that. It’s a real and frequently exploited attack vector, it’s just good practice for anybody with a semblance of concern about digital security
- Comment on To USB, or not to USB 2 weeks ago:
There are tons of ways to exploit a computer via a flash drive like that. Lots of viruses exist that would immediately install themselves upon the drive getting recognized. Famously Iran has a nuclear power plant taken offline by a random flash drive somebody plugged in, but aside Deon state level threats they can also just steal your financial details and personal info
- Comment on publication maxxing 1 month ago:
Lmaoo, I see your name now you’re funny
- Comment on publication maxxing 1 month ago:
We still do research and have lots of work to do before a paper even begins lol
- Comment on Can anyone scientists confirm? 2 months ago:
I’m the type of fish to brag about eating the bait
- Comment on Have you know???. 5 months ago:
It depends on the model but I’ve seen image generators range from 8.6 wH per image to over 100 wH per image. Parameter count and quantization make a huge difference there. Regardless, even at 10 wH per image that’s not nothing, especially given that most ML image generation workflows involve batch generation of 9 or 10 images. It’s several orders of magnitude less energy intensive than training and fine tuning, but it is not nothing by any means.
- Comment on Have you know???. 5 months ago:
The training is a huge power sink, but so is inference (I.e. generating the images). You are absolutely spinning up a bunch of silicon that’s sucking back hundreds of watts with each image that’s output, on top of the impacts of training the model.