Hola! This issue will be a bit longer than usual! So many things are happening and I want to share them all with you.

Get ready, because you'll be both perplexed and amazed by what you're about to learn! Let's dive in! 🤿

Prompt engineering might be a short lived "gig"

DeepMind has been working on a pre-training method called Algorithm Distillation which makes the technology behind our beloved GPT3 to be able to learn and improve through trial and error.

In other words, we won't need to do any prompt engineering or fine-tuning for this new technology. 

Will prompt engineering be a short-lived gig?

If we keep up this rate of innovation, we'll find out soon enough.

Easy for me, hard for you

What's easy for humans, it's hard for machines, and vice-versa. We can easily find a familiar face in photo of a crowd of people, but this kind of task it's incredibly difficult for computers.

That's because some core aspects of intelligence are missing from today’s AI models, and it takes a lot of data to train them.

The big news from Generally Intelligent is that they are open-sourcing Avalon, the world's fastest 3D simulator for AI agents.

This system allows researchers to engineer AI agents that can learn and understand like us, by exposing them to a set of tasks that are easy for humans to do.

AI-generated animation

To understand what's happening on the video above, you need to understand how Stable Diffusion in-painting works.

Wanna take out your ex from your holiday photos? No need to retouch your photo in Photoshop, you can just use editing prompts.

"take out boyfriend", and "add Hugh Jackman".

Try it here.

What these folks have done is truly impressive - they use in-painting and other AI systems to generate their frames in a short motion without using any animation software!

If you want to do a similar experiment, or if you want to generate extra heat with your GPU - check their post.

What else is new

• Researchers at Microsoft Research have developed a system that can generate 3d designs from basic speech descriptions of objects. Speech to VR is soon to be reality (pun intended).

• Sergei can show you how he transferred a physical toy in the digital world by using AI, in just one hour.

• Katrina created an amazing app that lets you shop by "prompting". Type down "animal print on a dress" and the app will show you the exact search, and a link to purchase it.

• This Videobook by Daniel, lets you "watch" audiobooks. Yeah, you read that right

• Google shared an AI-generated super-resolution video using Phenaki to generate long, coherent videos from text prompts. AI-Generated movies with our avatars are next?

• Nick launched an MVP for HairstyleAI, an easy way to try out new hairstyles and get inspiration with AI.

This photo does not exist

No biggie, just the current and ex CEO of Twitter hugging each other.