Saucy AI and Models

Hi, this is Ben’s Bites, your simple summary of what’s going on in AI. Written with AI.

Today we’ve got:

• Open-source and its role in AI

• Guided Diffusion Models - huh?! See below 😉

• Product Roundup

The Role of Open Source in AI

Swyx wrote a post titled “How Open Source is eating AI” and why open-sourcing other forms of AI might create more opportunities:

• Improving documentation. Lots of research is very technical, for widespread adoption we need to dumb down.

• Sharing prompts. People are more reluctant to share their prompts than they are startup ideas. Hopefully people get over this and more sharing of prompts helps everyone advance their skills.

• New UIs and improving accessibility. We need these things to be super easy to use! Remember how hard crypto has been for ‘normies’.

You can read the full post here.

Guided Diffusion Models

A paper was released titled “On Distillation of Guided Diffusion Models” - sounds technical, doesn’t it?! We broke it down for you:

Firstly, what is a ‘Classifier-free guided diffusion model’?

It’s a type of machine learning model that’s effective at high-resolution image generation.

But they’re expensive to run.

However, if you have a model that you've already trained, you can use that to create a new model that is more accurate.

So basically a new model is created that is simpler and requires fewer sampling steps to complete.

The result? It can generate images up to 256 times faster than the original.

It took 4 years for Michelangelo (not Michael Angelo) to complete the ceiling of the Sistine Chapel. Now the new-age AI-enabled Michelangelo would be able to do it in 5.7 days!! (if my calculations are correct)

ps. here's the link to the full 26-page PDF.

Product Roundup

Twitter translate bot. Translate any videos in any tweets, in... any language. Just tweet ‘@vidtranslator translate’ in a reply to a tweet and you’ll get it sent to you!

Check the tweet: https://twitter.com/altryne/status/1576741225226698752 

Pieter Levels is building in public again. He wrote how he turned his InteriorAI site into a money-making tool for interior design inspiration. Good job I've just finished my house 😒.

Check the tweet: https://twitter.com/levelsio/status/1577086792695975936 

Imagen Video. Essentially really really good text-to-video by Google. It even boasts “not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding.”

Check the tweet: https://twitter.com/hojonathanho/status/1577712621037445121 

A complimentary release from Google is Phenaki. This is another text-to-video piece of magic but this includes the ability to use different prompts throughout. So you can create stories, movies, comics, anime, romance, you name it!

Check the tweet: https://twitter.com/doomie/status/1577713150530445312 

AI Image of the day

That's it for today! See you next time ✌️

What did you think of this? Reply and let me know. It's a new experiment.