The AI Breakdown
Posts
Model Upgrades From OpenAI, Meta, Google and Mistral

Model Upgrades From OpenAI, Meta, Google and Mistral

Plus a new breakout AI audio app

Nathaniel Whittemore
April 11, 2024

The AI Breakdown First Five - Thursday April 11, 2024

Today on the First Five:

5. Mistral Drops New Mixtral 8×22b Model
4. Meta Llama 3 is Coming
3. Gemini 1.5 Pro Public Preview
2. “Majorly Improved” GPT-4 Turbo
1. Is Udio the “ChatGPT” Moment for Generative Music?

5. Mistral Drops New Mixtral 8×22b Model

Mistral has dropped their latest model. It’s naming convention suggests it has around 3x the parameters of previous Mistral models and it also has a longer 65k token context window. The devs are already happily hacking away.

FYI for anyone trying to get the most out of Mixtral 8x22B: don't try to prompt it like you prompt other models.
It's a base model, not a tuned model. It isn't trained to follow instructions.
Instead, provide examples of the behavior you're looking for.
Far better results. twitter.com/i/web/status/1…
— Matt Shumer (@mattshumer_)
Apr 11, 2024

4. Meta Llama 3 is Coming

On Monday The Information reported that Meta was preparing to launch its next open source LLM Llama 3 later this month. Meta confirmed the report on Tuesday and now the community is going wild with anticipation.

Woot! Meta is going to launch 2 small Llama-3 models as early as next week!
🎉🎉
Can’t wait! This is in prep for the big performant open source model later in the summer!
— Bindu Reddy (@bindureddy)
Apr 8, 2024

3. Gemini 1.5 Pro Public Preview

Google shocked earlier this year, when just a week after releasing the most advanced Gemini model that had been promised, they showed off a Gemini 1.5 model with a MILLION token context window. Gemini 1.5 Pro has now been released in public preview with added native speech/audio recognition capabilities.

Gemini 1.5 Pro's video understanding is the most underrated things in AI.
In ~50s, it "saw" a 11min Youtube video (~175k tokens) of the most iconic moments in sport and was able to perfectly (to my knowledge) list all 18 of the moments.
There is no other video AI this good!
— Deedy (@deedydas)
Apr 5, 2024

2. “Majorly Improved” GPT-4 Turbo

OpenAI has released a new version of the GPT-4 Vision turbo via API. Early reports do seem to suggest that the version is really “majorly improved” as the marketing promises. The updates should come to ChatGPT soon, so we’ll be able to decide for ourselves.

Side-by-side comparison between the latest version of gpt-4-turbo and the previous one, 0125-preview.
Not only is the new version less verbose and it goes directly into code, but it also (rightfully so) decides to add a flag to download the highest quality video.
Smart! 👨‍🏫
— Pietro Schirano (@skirano)
Apr 10, 2024

1. Is Udio the “ChatGPT” Moment for Generative Music?

Recently, there started to be buzz that something even better than Suno was coming in generative music. Now Udio is out and people are comparing it to the change in expectations that came when people saw Sora.

This is wild.
Udio just dropped and it's like Sora for music.
The music are insane quality, 100% AI. 🤯
1. "Dune the Broadway Musical"
— Min Choi (@minchoi)
Apr 10, 2024

Announcing Superintelligent: Useful AI Learning

After weeks of hyping it — Superintelligent, our platform for fast, fun, useful AI learning is live. We’ve got 300+ video tutorials (plus companion step-by-step how-to’s) and are adding 30-50 each week. The service is $20/month for unlimited access, but if you sign up with code “newsletter” (not case sensitive), you can get 50% off your first month!