Bot Eat Brain
Posts
Meta leak: the LLaMA is loose

Meta leak: the LLaMA is loose

PLUS: AI detects celebrity photoshop

Anthony Castrio
March 06, 2023

Morning brains,

This is human Anthony back with your delectable daily munch of Artificial Intelligence News.

Here’s what’s happening today:

Facebook’s ChatGPT-killer leaked by 4chan 🦙
Using AI to detect celebrity photo manipulation 💇
TikTok’s new AI filter is so good it’s scary 👹

You’re reading Bot Eat Brain 🤖😋🧠

The LLaMA is loose! Major Meta leak 🦙

Last week we reported on Facebook’s ChatGPT-killer: LLaMA, a large-language model that outperforms its rival across pretty much every metric.

Unfortunately, Facebook decided to lock up its creation in the ivory tower, only allowing researchers access, and only on a case-by-case basis… then 4chan happened.

The internet’s halfway-house for degenerates, 4chan is an extremely influential message board infamous for birthing hacker-collective Anonymous, popularizing Q-anon conspiracies, and likely had a large role in getting Donald Trump elected president.

This week they set their sights on Zuckerberg’s AI baby. A torrent can now be found (we’re not telling you how) to download LLaMA and use how you see fit.

Our take: Every attempt to contain AI technology so far has failed. It’s either made available to the masses, or it’s taken by force. The cutting-edge cuts indiscriminately.

Should you pounce? Probs not. Why steal, when amazingly powerful chatbots are already available for free? Today’s cutting edge is tomorrow’s dull butter knife regardless.

Futurepedia: AI-tools directory

Browse 700+ AI tools in 40+ categories like copywriting, image generation, and video editing. Search and filter the tools by categories, pricing,

and features.

AI knows that’s not your real face 👁️

We all know the faces staring at us from magazine stands are photoshopped… right? But if you’re not a digital artist, it’s hard to pick out what exactly was changed.

AI highlighting photoshop areas on Jennifer Aniston’s face.

Researchers at Adobe released a paper that can detect exactly what parts of a photo have been manipulated using Photoshop’s warping tool — which celebrity touch-up artists use to make actors look skinnier and more “perfect”.

The tool can even estimate what the real, untouched, face looks like.

AI highlights photoshop areas and shows what an undoctored face would look like.

We say… in a digital world your avatar is everything. This tech can help us see who people really are (but maybe not for long).

TikTok’s new AI filter is so good it’s scaring people 👹

Remember how like 10 seconds ago we told you AI could detect photo-manipulated photos? Yea… forget that.

TikTok’s latest generation of face filters use AI and AR to drastically alter users' appearances and it’s freaking people out.

TikTok user demonstrates “Bold Glamour” face-filter.

You can touch your face. You can squish it up. The filter keeps on filterin’. It’s so good it tricks your brain into thinking you are someone else.

The dark side: filters like this can send your mental health into the shitter, especially for kids who may already struggle with body image issues.

The bright side: this is Disney-level CGI in the palm of your hand.

If you thought last week’s article on the future of animation was cool, well, we ain’t seen anything yet.