You could even have options to only allow replies from such verified accounts, greatly reducing bots & spam on your profile. Maybe even "show replies from only verified accounts" so you straight up rarely, if ever, see bots.

I think it'd be good, especially on sites like Twitter, Facebook, or YouTube.

Imagine a world where you can verify the validity of who you're talking to, and the website or others don't store your information. The government already has your IDs in databases, it'd be (mostly) seamless and harmless.

for account creation and usage

bonus pts: Verification logs are not permanently stored, for both security and safety purposes (i.e. hacks, and abuse by the Gov. in question, tho highly unlikely this happens)

Would make it easy to determine who is and isn't a bot, while partially ensuring validity.

tbh i support ID verification for large-scale social media, but if and ONLY IF the following happens:
* Verification is handled via a government portal, not a third party
* Ability to differentiate between verified and not
* Only applies to large-scale, centralized platforms
* Isn't mandatory...

(they do not care that they are being lied to by the people they listen to)

...?

The website tried to stop me with a rate limit (Tried to ban me)
I just switched up the proxy, then got back in it (Back in)
This shit is questionably sourced, but I gotta get it fed (It's hungry)
Training up a model 'til it's painting shit inside my head (In my mind)

This was exciting. I don't think a score below 20 is physically possible yet. www.aipuritytest.org

Part of my research involves training, generating images (very rarely 'video'), and I fucked around with making shitpost music for a bit a year ago. The most embarrassing one for me is saying 'thank you', but I swear I only do it to encourage the idea that it's on the right direction for a process 🤗

yeah it is btw

i just moved to Zen browser (firefox but like vertical tabs focused or w/e) like 2 days ago stop watching me thru my window <3

bonus image where i turned the guidance up

me when i study distributions only to realize that training on your own outputs just increase the already-existing underlying signals

bonus meme: kid named reinforcement learning:

kid named Sex Murder Party by Gorillaz on the album Humanz:

Bonus meme: Use it in addition to Tangential Dampening (any model) & SkipLayerGuidanceDiTSimple (if using a DiT model, like Flux, Chroma, SD3, and others). They prevent CFG overburn & speeds up sampling with a slight increase in structural quality, respectively.