Alex Vranas @Alex@vran.as

We need to talk about Sam Altman's sanpaku eyes.

My wife and I are celebrating our 10 year anniversary today. It's been a wild decade, but we're making it.

Did something stupid as root and completely bricked my laptop. Really seeing the engineering from System76 pay off as I recovered my OS is less than a minute using the recovery wizard.

I've never recovered from this big of a fuck up so fast before on Linux. I'm having to reinstall a few things, but I lost absolutely nothing.

Disappointed in tech media for not using the term "dogfooding" to describe Apple's Scary Fast keynote event produced entirely with iPhones and Macs.

Important supplemental material: https://web.archive.org/web/20000303145808/http://twinkies.com/splash.html

I don't actually have these, but you could, for $995.

https://www.ebay.com/itm/334281500193

Time to find out how shelf stable they really are.

Complaining about how cold it got so fast, only to remember that I can finally fire up my 3D printer without creating a heat problem in my small apartment.

Printing the BYU One-Piece Compliant Blaster.

https://compliantmechanisms.byu.edu/maker-resources

This screenshot will temporarily teleport your brain back to 2004. Use with caution.

They're selling your data for $2,000/month.

https://app.snowflake.com/marketplace/listing/GZTSZ290BUX40/cybersyn-inc-cybersyn-consumer-spending-dashboard

Small correction: Stable Diffusion has 2.3B 256x256 images, and 170M 512x512 images that it is trained on.

But still, that comes out to less than one byte of model data per image in the training set. There's literally no way to argue that a 256x256 image got compressed down to less than a byte.

It is worth noting to those who are in opposition to AI that the "models trained on my work contain my work" argument isn't technically or legally sound. Better arguments need to be made. If we could cram 1B+ 512x512 images into a single 2GB model, it would represent the single greatest breakthrough in data compression technology in human history. It turns out, this isn't a compression breakthrough, it's just proof that the original data isn't in the model.

I agree that some degree of mimicry isn't infringement - fair use includes some degree of derivation - but I really don't think that training and using a model to deliberately imitate another artist's work represents fair use, especially if it can be found to harm the artist.

The line shouldn't be drawn at "is it an exact copy?", it should be drawn at "would it confuse a reasonable person into thinking this was original art by another artist?".

Yesterday's lawsuits against Generative AI companies like Stable Diffusion and Midjourney were mostly dismissed, and I generally agree with that. There is one case still ongoing between Sarah Andersen and StabilityAI that I expect Sarah to lose on the grounds that her accusation isn't sound (her artwork isn't actually stored in the model).

But I am disappointed that they ruled that copying styles isn't infringement unless it is an exact copy of a work. I feel like this isn't a nuanced ruling...

Sad to remove Cohost from my bio, but I do not ever use it anymore. I don't really have anything bad to say about it, it just wasn't my style.

Still love eggbug though.

But, again, I don't fall for the "All AI art is art theft" argument, because it just doesn't make sense.

If I type "a moonlit beach, painting by [living artist]", It could be argued as infringing on that artist, but If I just type "a moonlight beach, oil painting", who does it infringe on? Every artist the model was trained on?

If a work infringes on every artist at once, it's safe to say that it doesn't infringe on any artist at all.