Bot Eat Brain
Posts
Stable Diffusion XL 1.0 is here

Stable Diffusion XL 1.0 is here

PLUS: Altman launches Worldcoin

Dan Bogachek & Michael Parrish
July 27, 2023

TOGETHER WITH

Good morning, human brains. Welcome back to your daily munch of AI news.

Here’s what’s on the menu today:

The Frontier Model Forum 🏛️
AI chiefs team up to build “safe“ and “responsible“ AI.
Stable Diffusion XL 1.0 is here 🌄
Stability AI’s latest image-generating AI model launched yesterday.
Sam Altman debuts Worldcoin 🤑
The OpenAI CEO launches a new cryptocurrency.

Sponsor Bot Eat Brain | New here? Subscribe!

APPETIZER

The Frontier Model Forum

Safety first.

Anthropic, Google, Microsoft, and OpenAI unveil the Frontier Model Forum. It’s a partnership to develop safe, responsible AI models.

Source: OpenAI

The main points:

1/ Promote safety standards and mitigate risks.

2/ Identify key research questions and create an AI model evaluation library.

3/ Establish secure, responsible data exchange on AI safety.

The forum will establish a so-called “advisory board” for developing future AI.

Teamwork makes the dream work.

Our take: First, the “voluntary“ agreement with The White House, and now this. Seems like they want the public to know that AI safety is a priority. Too early to tell whether that’s true or not — the proof will be in the digital pudding.

BUZZWORD OF THE DAY

Amazon Bedrock

A machine learning platform by AWS (Amazon Web Services).

It eliminates the need for developers to spin up their own complex server infrastructure to build foundational models (FMs).

Developers can also deploy their FMs to other Amazon services once they’re ready, e.g. an AI-powered chat website.

FROM OUR PARTNERS

When’s the last time your newsletter wrote itself?

Probably never, until today.

Introducing Hoppy Copy’s Newsletter Creator.

It’s the newsletter writing machine that lets you go from topic, to outline, to full draft — in mere minutes.

With an AI trained specifically for long-form copy, it’s able to write your newsletter in the exact tone you choose. And Hoppy even incorporates images, that dovetail, right into that copy.

It does all the lifting you want, without tying your hands. Tweak the outline, edit the copy on a sentence level, change the images — your piece is completely steerable, every step of the way.

Join 20,000 creators who don’t know what to do with all their new free time.

MAIN COURSE

Stable Diffusion’s new AI model 🌄

Fashionably late?

Stability AI released Stable Diffusion XL 1.0 yesterday. It’s Stability’s newest image-generating AI model.

The release was originally slated for July 18th.

To be fair, a week late isn’t egregious.

Source: @StabilityAI

The 411:

1/ Stable Diffusion XL 1.0 runs on Amazon Bedrock.

2/ It’s open-source.

3/ Stability says it’s more powerful, customizable, and requires less computing power than previous versions.

4/ Stable Diffusion XL 1.0 is now on GitHub, Clipdrop, and DreamStudio.

Check it out on Clipdrop here.

Coinciding with yesterday’s UFO hearing, Stability’s Chief Edgelord, er sorry — CEO Emad Mostaque, posted this timely meme:

Source: @EMostaque

Cheeky.

Our take: Let’s address the elephant.

It didn’t take long for the internet to do what the internet seemingly does best — there’s quite a buzz around using SDXL for NSFW content. There are already tutorials online on how to create “non-consensual deepfake” porn. Even in this HackerNews thread, porn is right in the top comment.

Stability claims there are “extra steps“ in place to prevent this. But a simple Google search says otherwise… so there’s that 🤷‍♂️.

A LITTLE SOMETHING EXTRA

OpenAI’s CEO launches new crypto 🤑

OpenAI CEO Sam Altman launched Worldcoin earlier this week. It’s a new crypto designed to distinguish humans from AI, and (hopefully) provide AI-funded universal basic income to the world.

Pay up, machines.

Source: CoinMarketCap

You create your digital identity by scanning your eye with…

The Orb.

Source: Fortune | Karsten Moran

Nothing creepy here…

Worldcoin lives on the blockchain just like other cryptos, but it’s unique in that it revolves around “proof of personhood” instead of other traditional verification methods like Proof of Work (Bitcoin) or Proof Of Stake (Ethereum). The iris scans live on the Orbs themselves as “IrisHashes”, which are secured and distributed cryptographically with zero-knowledge proofs.

All this is to say: non-humans can’t join in the fun, and in theory, it eliminates the double-dipping problem. One human, one World ID.

Our take: Sam Altman’s claims are ambitious to say the least, but it’s 3 years in the making. But what are they going to do with all those delicious iris scans… we hope they really know what they’re doing.

MEMES FOR DESSERT

YOUR DAILY MUNCH

Think Pieces

A Minecraft-playing chatbot named STEVE-1. How they made it happen, how good Steve is, and how to replicate the results.

Using watermarking techniques to prevent harmful deepfakes. A new technique, dubbed “Photoguard“, developed by MIT CSAIL.

Programming with ChatGPT. A basic guide on how to squeeze coding performance out of the go-to chatbot.

Startup News

Netflix’s newest AI job offering. They’re hiring for a product management role with a salary of $900,000.

Microsoft’s shares fall after earnings report. Azure Cloud’s slow growth may be to blame.

Spotify’s upcoming AI features. CEO Daniel Ek hints at personalized experiences, automatic podcast summaries, AI audio ads, etc.

Research

Reddit thread on Stable Diffusion XL 1.0. Including instructions, prompts, and helpful insights.

“Ripple Effects“ in LLMs. How to minimize incorrect/outdated outputs in Large Language Models.

Helpful Twitter (X) thread on using Runway’s text-to-video. A straightforward, 4-step guide to get the most out of your Runway prompts and how to download.