Google Lays Down Some Phat Beats
Friday, December 8
Gemini Arrives Suddenly
Just like all the other AI newsletters, I got this one wrong. Many of us were referencing The Information, a very respectable source, when we shared that Google’s answer to ChatGPT, Project Gemini would be delayed until 2024. However, here it is, arriving in 2 out of it’s three flavors, with only the “Ultra” being delayed, and the Nano and Pro ready to go. If you’d like to read more about what Gemini is capable of, I found this article from the Verge really helpful.
Gemini is more than a single AI model. There’s a lighter version called Gemini Nano that is meant to be run natively and offline on Android devices. There’s a beefier version called Gemini Pro that will soon power lots of Google AI services and is the backbone of Bard starting today. And there’s an even more capable model called Gemini Ultra that is the most powerful LLM Google has yet created and seems to be mostly designed for data centers and enterprise applications.
However, Kyle Wiggers over at TechCrunch was not impressed with the prerecorded demos, the lack of concrete answers and considers it a rush job like the initial launch of Bard. While these are probably cherry-picked examples, you should probably watch this video to see what it’s capable of.
Why it matters: Google Gemini is a huge leap forward, but is too little, too late?
Google’s AI Instrument Playground
I had a lot of fun playing with Google’s “Musical Test Kitchen” last summer and used it to create the soundtracks for many of the AI film projects I’ve been working on. This tool is even more amazing because it let’s you combine from a list of 50+ adjectives with a broad range of international musical instruments. And then, after you’ve made your AI music sample, you can read all about that instrument or remix and record in the advanced tab. I’m going to have to find someone who knows more about music to explain it to me, but it sure is fun!
Leonardo.ai raises 31 million
Leonardo has been getting a lot of attention lately, and not just for it’s dynamic canvas that I discussed on Monday. As I was explaining to SCCA student Travis R. today, with it’s web based interface, multiple models, and plentiful options, it can produce Midjourney level images. With it’s generous credit system that refills everyday, you could easily create 100 images a day for free. It’s Pro plans are affordable and it’s generally a lot of fun to use once you’ve figured out it’s rules. It has prompt helping, out-painting, custom model generation, and upscaling. And now they have an extra 31 million dollars to work with. Check it out
Meta Has it’s Own Image Generator Now
Kudos to Meta for having an excellent research department and going all in on open source for it’s LLMs, but I’m not sure what purpose it serves. The quality is good, the images are not that big…and why would you use it when you have so many options. Unless they want to fold it into Facebook and Instagram as a feature so everyone can now just post made up pictures to their feed. Read about it | Try it out
Okay, I just discovered my hunch was correct, it’s being rolled out as a feature in group chat. Just think, you can generate a zany image with a prompt and your friend can modify it, and you can spend even more hours on the platform.
Pimento Looks Promising. AI Mood Boards
To be honest, I haven’t used it yet, I just signed up fro the waiting list. The premise behind Pimento is brilliant; you enter in a creative brief and a few starter images and it generates an entire mood board. Additional prompts can refine it further. I’m looking forward to tyring it out . If you’re curious, you can join the waitlist here or read the review on TechCrunch here.
Explainer Videos: AI in a Minute
How do you create clear and concise videos about complex AI topics? Ask Greta Heller (aka Greta Rose), SCCA class of ‘23 who worked on this project for Microsoft. You could read the blog post here, or watch the videos here. Congratulations, Greta!
Wow, this year is flying by. Only two weeks until Christmas. Pika Labs 1.0 and Midjourney V6 any day now. I hope you’re learning a lot and staying dry. As always, drop me a message if you have questions.
.