- The Renaissance Times
- Posts
- Mistral's "Le Big Model" Is an Open Source Beast That's Partnering with Microsoft
Mistral's "Le Big Model" Is an Open Source Beast That's Partnering with Microsoft
Welcome Renaissance Creator,
You know what they call a Big Mac in France? Le Big Mac.
What do they call a state of the art, open source, Large Language Model in France? Le Big Model.
Mistral does anyway. And it’s ready to compete with the big boys, to the tune of a partnership and a $16 million investment from Microsoft.
Let’s dive in.
Hot Off The Press
Mistral releases its most powerful model yet, Mistral Large aka “Le Big Model”
Inkitt, a self-publishing platform using AI to develop bestsellers, raises $37M led by Khosla Ventures
Former Twitter engineers are building Particle, an AI-powered news reader
Qualcomm’s on-device AI models will be hosted on Hugging Face and GitHub
Nvidia launches RTX 500 and 1000 Ada Generation laptop GPUs for AI on the go
Bing’s Sydney is back with a bang
Microsoft and Mistral AI announce new partnership to accelerate AI innovation and introduce Mistral Large first on Azure
Microsoft made a $16 million investment in Mistral AI, drawing attention from EU regulators
Google is planning to release Gemini image generation back to the public within a few weeks
Alibaba Invests in Monolith AI, a Chinese AI Firm, at a $2.5 Billion valuation
The One Big Thing
Le Big Model is Ready for Prime Time
Mistral, the French AI company at the forefront of the open source revolution, is upping their game and releasing their largest and most powerful model yet: Mistral Large.
The Basics:
Mistral Large is available through la Plateforme (Mistral hosted) and Microsoft Azure, their first distribution partner
32K token context window
Natively fluent in English, French, Spanish, German, and Italian
Outperforms every publicly available model aside from GPT-4 at coding and math and shows powerful reasoning capabilities
Also released today: Mistral Small, optimised for latency and cost. Mistral Small outperforms Mixtral 8x7B and has lower latency
One of the more interesting parts of this release is the built-in distribution of Microsoft’s Azure that Mistral is targeting.
Open source models are no longer going to just be used by hobbyists but by professionals alike. We are entering a whole new world where models become freely accessible and the only limitation becomes the cost of compute. We can reasonably expect to see developers operating in one cloud environment and working with whichever model is best suited for each specific task in each instance.
This creates a new type of business model for open source companies and startups in general. While it seemed for the longest time that GPT-4 would be the model of choice for every use case, this may no longer be the case any longer. The power of open source comes in the ability for anyone in the world to improve upon it.
While Mistral Large may be slightly behind GPT-4 in terms of performance at the moment, the power of open source means that the entire model and weights (parameters) are open to the public.
People can tweak and improve upon them as they choose. Open source unleashes the power of the collective toward a common goal. While I don’t see Large being able to leapfrog GPT-4 anytime soon, this is a baby step towards a more decentralized AI future.
A beautiful sight to behold.
The Gallery
Fix bugs simply by taking a video of them
The future of fixing bugs?
Just record them.
I filmed 3 separate bugs in an app and gave the videos to Gemini 1.5 Pro with my entire codebase.
It correctly identified & fixed each one.
AI is improving insanely fast.
— Mckay Wrigley (@mckaywrigley)
6:00 PM • Feb 26, 2024
AI assisted social media engagement
I have created a monster 🤯
Engaging has never been easier.
— Nilan Saha (@nilansaha)
8:15 AM • Feb 27, 2024
ChatMusician: Understanding and Generating Music Intrinsically with LLM
ChatMusician
Understanding and Generating Music Intrinsically with LLM
While Large Language Models (LLMs) demonstrate impressive capabilities in text generation, we find that their ability has yet to be generalized to music, humanity's creative language. We introduce… twitter.com/i/web/status/1…
— AK (@_akhaliq)
4:51 AM • Feb 27, 2024
New Sora videos dropped and look incredible (Beehiiv, please fix Reddit embed links)
Tools
Must have tools for every Renaissance creator to add to their toolkit:
Lmsys: Chat with any language model of your choice
StickerBaker: Make stickers with AI
Archer: Tackle 1000s of questions, real SAT tests, and enjoy personalized learning with Archer, your AI math guide.
Myko AI is a conversational AI for sales and revenue teams that instantly answers questions and builds reports from a user’s existing data sources
STORM, a system that writes Wikipedia-like articles based on Internet search
Pmndrs/ Uikit by @BelaBohlender is fully responsive via flexbox, built on the yoga layout engine
Daily Portfolio Summarizer with Langchain, Qdrant, and Mistral AI
Deep Tech
The newest and coolest in the research world that you need to know about:
Qualcomm Released 80 new models compatible with their mobile devices
Unintended Impacts of LLM Alignment on Global Representation
AI Watermarking 101: Tools and Techniques
Symbolic Music Generation with Non-Differentiable Rule Guided Diffusion
OpenHermesPreferences: the largest open dataset for RLHF & DPO
Meta presents MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases
What’s the latest in TensorFlow 2.16?
Closing Thought
WhatsApp reached unicorn status with under 20 people. What will be the smallest AI unicorn?
Work With Us!
The AI Renaissance is coming and we are building the best community of the people making it happen.
Contact us to sponsor your product or brand and reach the exact audience for your needs across our newsletter and podcast network.