SDXL OMG + AI Furniture + Google Robots... OH MY

Your Weekly Dose of Fun AI News & Tools

It’s that time of the week again… let your hair down, loosen your tie, pour yourself a Tab soda — for some reason we’re all odd middle-aged men from the year 1976 — now let’s GET INTO SOME AI GOODNESS!

Please Clap with Your Very Human Hands

Welcome back to the AI For Humans weekly newsletter, which is something very different than your average AI newsletter. They’re out here trying to tell you how this or that app is going to make a difference in life and we’re just out here to have some fun.

OFC you’re learning stuff along the way through being curious, honestly, isn’t that the point?

Let’s get to the C O N T E N T baby!

🌭 HOT LINKS 🌭

SDXL 1.0 Available To Everyone

Stability.AI’s new image model SDXL has finally come out and we’re seeing some pretty good results with is so far.

Sad Tiger Astronaut is Sad

So why would you want to spin this up instead of Midjourney?

Well, let’s ask a few questions to get that answer:

1) Do you like tinkering with things or would you rather just get a “good enough” result?

2) Do you have some basic skills with computers and/or the patience to learn some very basic terminal (non-windows) work?

3) Do you hate the idea of Discord as a UI so much you’ll do literally anything to stop using Midjourney?

If the answer is yes to any of these questions, SDXL is for you.

In all reality, we adore Stable Diffusion (and prefer it over MJ) for the ability to get granular with what you can create. The world’s best plug-in for AI image generation is called ControlNet which allows you to generate more images with the same shape & SDXL is optimized for that out of the gate.

ControlNet in action

You can try it out for yourself on ClipDrop and if you have the tech chops, you can install it from here: https://github.com/Stability-AI/generative-models

Adobe Photoshop Adds ‘Generative Expand’ For Real

Adobe took the world (and a lot of our TikTok feeds) by storm with its Generative Fill technology which brought AI art generation to a TON of new artists who use Photoshop daily.

One of the weirdest use cases (and one we’ve loved to see be exploited for fun on TikTok) is expanding out the field of view of a particular image and seeing what gets filled in.

@pictelate

Using Photoshop’s generative fill tool to extend a scene from The Office #photoediting #videoediting #generativeai #theoffice

Adobe, smartly seeing this get used a lot, just went ahead and made it a feature of the tool-set.

Wayfair (yes, the furniture company) Makes Cool AI App

In the “def weren’t expecting that’ category this week, Wayfair, the semi-cheap but now ubiquitous online furniture store has launched Decorify, an AI powered app that let’s you see new furniture in your space.

🚨 PAY ATTN TO THIS 🚨

Large Language Models (LLMs) Are Finally Making Robots Smart?!

Very good tech reporter & Hard Fork podcast host Kevin Roose from the New York Times had an amazing piece this week on Google’s new RT-2 robotics model which is integrating LLMs into their robot training. This has proven a major break-through for getting robots to learn and act closer to how people do.

Google’s Robot Can Pick Up Toy Dinos

Why should you care? First and foremost, if we ever want robots to really be able to do things like help us around the house or maybe more importantly get our medicine for us when we get old (those damn kids will be too busy trading AlphaWorldCoins in Metaverse4), we need them to get better at understanding the real world.

Of course, this also significantly ups the potential for a robot uprising so there’s that as well.

Unrelated, but not entirely, check out this AI4H Tok showing how Minecraft robots are learning, just like humans!
@aiforhumansshow

Meet STEVE-1, a breakthrough Minecraft AI that's more than just a game-changer—it's a future-shaper. STEVE-1 responds to real-time text an... See more

🪤 DUMB THING TO DO WITH AI 🪤

Ever see something on your iPhone and you can just never remember what in the hell it was?

Well worry no more! The Rewind App is here to save you.

Rewind basically lets you search through screenshots, and everything you’ve read in Safari (it uses your browser history) then helps you find that one thing. The team says they put privacy first and they won’t see any of your info whatsoever.

To be honest, this is less a dumb thing and likely MUCH more useful than we’re including but in name it sounds a lot like the generally pretty terrible Adam Sandler movie “Click” so we’re including it here.

🤖 AI FOR HUMANS LINKS 🤖

This week on the Pod we went deep into RunwayGen2’s Image-To-Video service in a very special way: We used it to create the cutest little Animal Influencers you’ve ever seen.

The clip (separated from the larger show for both ease and view-count purposes HINT HINT) is available below:

We’ll see y’all next week for another one!

Please don’t forget to share this newsletter, or our pod, with the world!

-Gavin & Kevin. And Gash, in spirit. But Cassidy in practice. IYKYK.