Mistral AI, a burgeoning startup in the AI sector, has set out on a mission to revolutionize generative artificial intelligence (AI) with its first large language model (LLM), Mistral 7B.
The company hopes the new 7-billion-parameter model will become an open-source alternative to current AI solutions.
Mistral 7B And Mistral 7B Instruct Models
While others have set the industry standard with their “black-box” models, Mistral AI believes an open-source, community-driven approach can outpace them.
Drawing comparisons with the open-source movements in web browsers and operating systems, Mistral suggests that community-backed models are the future.
Mistral 7 B’s release comes as the company’s first significant step toward creating specialized models that compete with larger, more established AI solutions.
The raw model weights are distributed with Bittorrent and on Hugging Face. This documentation details the deployment bundle that allows to quickly spin a completion API on any major cloud provider with NVIDIA GPUs.
Mistral AI’s open models aim to offer superior adaptability, enabling customization to specific tasks and user needs.
This approach is touted as advantageous for businesses aiming to keep costs low while maintaining performance.
Additionally, the company believes that open-source models will be critical tools in combating the ethical challenges associated with AI, such as censorship and bias.
As generative models continue to influence society, the ability to audit them for flaws and misuse is becoming increasingly vital.
How To Use Mistral 7B For Free
— Mistral AI (@MistralAI) September 27, 2023
— clem 🤗 (@ClementDelangue) September 27, 2023
In addition, you can chat with the Mistral 7B Instruct model on Perplexity Labs.
Mistral AI Made Headlines With Seed Funding
Mistral AI made headlines this summer when it raised $113 million in seed funding in June, underlining investor confidence in the open-source approach.
Incredible achievement by @InflectionAI – In less than a year, they’ve developed one of the most sophisticated LLMs and launched Pi, the first personal AI product w/ a high EQ. https://t.co/kDWLql8nJG
— Eric Schmidt (@ericschmidt) June 30, 2023
Mistral AI’s team is comprised of data scientists, software engineers, and machine learning engineers plucked from DeepMind, Meta, Hugging Face, and others.
Big news! I'm excited to share that I'll be starting a new chapter in my career at @aimistral. I'm incredibly grateful for the growth and memories I've made during my time at @huggingface. Looking forward to bringing my skills and passion to my new role. 🚀 🚀 🚀
— Saulnier Lucile (@LucileSaulnier) July 18, 2023
Arthur Mensch, Co-founder and CEO of Mistral AI expressed excitement about what the company planned to achieve:
“Our training as AI researchers, combined with our respective professional experiences within the world’s leading technology companies, has convinced us that there is a way forward for an alternative, innovative project that will enable us to responsibly disseminate the most promising technology of our generation as widely as possible.
We are proud to initiate this global project from France, our home country, and to contribute, at our level, to the emergence of a credible new player in generative artificial intelligence from Europe. Over the coming months, we will focus all our energy and passion on honoring the trust placed in us by our investors.”
According to the pitch deck, Mistral’s plans include developing AI models superior to OpenAI’s in 2024.
In that round (Q3 2024), we expect to need to raise 200M, in order to train models exceeding GPT-4 capacities. Strong financing will allow us to train models on larger infrastructures, thereby establishing us as a research leader in AI that will be the go-to provider of the European industry.
Mistral AI hopes to progressively release new models that bridge the performance gap between its open-source solutions and proprietary offerings as part of its ongoing strategy.
France Positioned As Next Leader In AI Development?
In June, French President Emmanuel Macron, a big promoter of French tech startups, was at VivaTech, Paris’s largest European tech trade show.
He wanted to support French startups, help them expand internationally, and attract more investment in AI research and projects in France.
Technology experts have also noted that most developers (11 of 14) of Meta AI’s open-source Llama technology are French, making the latest AI developments unsurprising.
Today we release LLaMA, 4 foundation models ranging from 7B to 65B parameters.
LLaMA-13B outperforms OPT and GPT-3 175B on most benchmarks. LLaMA-65B is competitive with Chinchilla 70B and PaLM 540B.
The weights for all models are open and available at https://t.co/q51f2oPZlE
— Guillaume Lample (@GuillaumeLample) February 24, 2023
The Future Of Open-Source AI
A potentially robust and open-source competitor to existing LLMs like Mistral 7B could offer new opportunities for businesses to utilize AI, with broader customization possibilities and enhanced control over data security.
The move to open-source generative models represents a significant shift in the AI industry, challenging traditional proprietary models on ethical and performance grounds.
Featured Image: The Hornbills Studio/Shutterstock