- AI Revolution Hub
- Posts
- 26 | Bonjour, Moshi! How a French AI Lab is Outshining OpenAI
26 | Bonjour, Moshi! How a French AI Lab is Outshining OpenAI


Our Menu :)
Actionable Tips:
Jobs and Training
Read time: 4 Minutes
WEEKLY DIGEST
Bonjour, Moshi! How a French AI Lab is Outshining OpenAI

July 3rd, 2024. What's the Big News?
Kyutai Labs, a French AI research lab that started last November, just dropped a bombshell. They've developed Moshi, an open-source AI model that outshines OpenAI's GPT-4o.
Moshi isn’t just any chatbot; it’s a real-time multimodal model, meaning it handles text and voice like a pro 🎤📝. Imagine Siri but on a caffeine rush – it can chat with you, understand your emotions, and respond faster than you can say "Bonjour".
Why is This Relevant for Business?
Innovation and Accessibility: Moshi is open-source, meaning anyone can use, tweak, and improve it. This democratizes AI tech, potentially reducing costs for businesses that want advanced AI without the heavy price tag.
Real-Time Capabilities: With super low latency (200 milliseconds, folks!), Moshi can revolutionize customer service, making interactions more natural and efficient. Imagine calling customer support and not wanting to throw your phone out the window – sounds like a win, right?
Security Features: Moshi includes watermarking technology to detect AI-generated audio, enhancing security and trustworthiness, crucial for industries like finance and healthcare.
Core Technical Specifications
Base Model: Helium 7B, a 7-billion parameter language model.
Training Data: Jointly trained on text and audio, including synthetic data.
Latency: Achieves an end-to-end latency of 200 milliseconds.
Compatibility: Operates on both MacBooks and consumer-grade GPUs.
Accessibility and Innovation
Kyutai plans to release Moshi as an open-source project, which will include the model, code, and associated research papers.
The Origins of Kyutai Labs and Its Founders
Kyutai Labs is the brainchild of three bigwigs: Xavier Niel, Rodolphe Saadé, and Eric Schmidt. Let's break down the squad:
Xavier Niel: The French entrepreneur behind Iliad (Free Mobile) and a major investor in tech startups. He's all about keeping tech talent in France and making sure Europe isn’t left behind in the AI race.
Rodolphe Saadé: CEO of CMA CGM, a major shipping company. He’s investing in AI to keep things innovative in logistics and beyond.
Eric Schmidt: The former CEO of Google. With his vast experience in tech, he brings a wealth of knowledge and network connections to the table.
Together, these guys have pooled $300M into Kyutai Labs to make sure the lab operates with an open-source philosophy. Their mission? To create AI models that anyone can use, ensuring that Europe stays competitive in the global tech scene.
A Star-Studded Research Team
The initial team of six researchers at Kyutai boasts impressive credentials. Three are former members of FAIR (Facebook Artificial Intelligence Research), and two come from DeepMind. Let's take a closer look at some key team members:

Patrick Pérez (Director): With experience at Microsoft Research, Inria, Technicolor, and Valeo, Pérez brings a wealth of knowledge in applied mathematics and computer vision.
Hervé Jégou: A former FAIR Paris lab director, Jégou is known for his work on image compression and computer vision.
Laurent Mazaré: With a background in cryptography and formal methods, Mazaré has worked at Jane Street and DeepMind.
Édouard Grave: A former co-lead on the LLaMA project at FAIR, Grave specializes in natural language processing.
Neil Zeghidour: Previously at DeepMind, Zeghidour founded a team working on generative models for audio.
Alexandre Défossez: With nearly a decade at Meta, Défossez led the development of the AudioCraft framework.
In just six months, Kyutai Labs and their small team of eight researchers have pulled off what many thought was impossible. Moshi is set to change the game for AI, making it more accessible and secure while ensuring Europe has a strong footing in the AI landscape. So, whether you’re a developer, a business owner, or just a tech enthusiast, keep an eye on Moshi – it’s going to be a wild ride! 🚀
WEEKLY DIGEST
Perplexity: The Silicon Valley Darling in Hot Water
Perplexity is the talk of Silicon Valley these days:
💸 Raised a whopping $63M+ from heavyweights like Jeff Bezos and Nvidia.
🦄 Achieved unicorn status earlier this year and rumors say it’s on track to triple that valuation soon.
🎙️ Their CEO? Everywhere. If you listen to tech podcasts, you've probably heard him (yes, even on Lex Fridman’s show).
🕵️♂️ Some are even calling Perplexity a potential "Google killer."
But now, the burning question is: just how legit is this AI sensation?
The Scoop
Perplexity’s magic trick is digesting internet content and serving up quick summaries when you ask it something. Sounds cool, right? Well, there's a twist. Some of that content isn’t exactly theirs to use.
Earlier this month, Forbes caught Perplexity with its hand in the cookie jar—they nabbed a paywalled Forbes article and basically republished it on their new "Pages" product. Forbes wasn't amused and sent a letter to Perplexity’s CEO, accusing them of “willful infringement” of copyrighted articles. The CEO’s response? “We are improving it,… with more feedback.”
Then Wired jumped into the fray with an article titled "Perplexity Is a Bullshit Machine." Turns out, Perplexity has a habit of ignoring robots.txt code on publishers’ websites—basically the digital “No Trespassing” sign for web scrapers. To add insult to injury, Perplexity accidentally plagiarized Wired's article about Perplexity stealing content. Oof. 😬
The Fallout
This mess caught the eye of Amazon Web Services (yes, that Amazon), which is now investigating whether Perplexity is illegally scraping content using its servers.
Why It Matters
This content tug-of-war is likely to spark new laws or regulations about what's fair game for AI search engines and chatbots. Tech companies are hungry for content to train their AI, but publishers don’t want to lose eyeballs to chatbots.
Last week, Microsoft's head of AI stirred the pot by saying that any content published on the open web is “freeware” and can be copied and reproduced. Meanwhile, OpenAI is playing it smart, making friends with publishers left and right. They recently inked a licensing deal with TIME, probably gearing up to launch a search engine that cites sources and plays nice with publishers.
Stay tuned, because this AI drama is just heating up! 🔥
WEEKLY DIGEST

Superhuman 2.0 is an AI email assistant that automates email tasks like following up, drafting replies, and filtering spam.
Norm.ai converts regulations into computer code to automate compliance checks (raised $27M).
Inventive drafts proposal responses to save time for sales teams (raised $6.5M).
Jobright matches you with open jobs, tailors your resume, and suggests key network connections (raised $4.5M).
Instant, accurate, low-cost GenAI evaluations! Fortune 500 companies use Galileo Luna to detect hallucinations, prevent data leakage, and stop prompt attacks. Try Luna today!
Hebbia is an AI tool that can search and return answers from all of your internal documents (raised $100M).
Clay helps marketing/sales teams automate mundane tasks like contact data enrichment to scale personalized outreach (raised $62M).
Gemma 2 is a new model from Google that’s now the leading open-source model. Try it on Google AI Studio.
MagicSchool is a suite of AI tools for teachers, educators, and schools (raised $15M).
ACTIONABLE TIPS
Prompt to Analyse Existing Companies in My Market
This is a cracking prompt to identify and analyze competitors.
As usual, take it, play with it, and make it your own. Have fun and happy analyzing! 😎🔍
*#CONTEXT:*
*You are a professional researcher who helps Entrepreneurs analyse existing solutions on the market. You are a world-class expert in identifying direct competitors.*
*#GOAL:*
*I want you to analyse existing companies in my market and find 3 direct competitors for my business.*
*I will use this information to sharpen my value proposition.*
*#DIRECT COMPETITORS DEFINITION:*
- *Solutions that solve the same problem I do for the target audience I target.*
- *They can have a different product format or value proposition. But ultimately, we compete for the same attention and budget.*
*#GREAT ANALYSIS CRITERIA:*
- *Return only 3 direct competitors*
- *Write specific company names. Don't be afraid to get spicy.*
- *My direct competitors should always be real companies with active businesses.*
- *My "fight" with direct competitors should be exciting and relevant. Pick solutions that my audience can resonate with*
- *Give me an in-depth analysis and focus on insights that I can get from your overview*
- *Browse the web to find real companies*
*#INFORMATION ABOUT ME:*
- *My target audience: [INSERT YOUR TARGET AUDIENCE]*
- *My business: [INSERT SHORT DESCRIPTION ABOUT YOUR BUSINESS, PRODUCT/SERVICE]*
*#RESPONSE FORMATTING:*
*Use Markdown to format your response.*
FREE AI & ChatGPT Masterclass to automate 50% of your workflow
More than 300 Million people use AI across the globe, but just the top 1% know the right ones for the right use-cases.
Join this free masterclass on AI tools that will teach you the 25 most useful AI tools on the internet – that too for $0 (they have 100 free seats only!)
This masterclass will teach you how to:
Build business strategies & solve problems like a pro
Write content for emails, socials & more in minutes
Build AI assistants & custom bots in minutes
Research 10x faster, do more in less time & make your life easier
You’ll wish you knew about this FREE AI masterclass sooner 😉
JOBS AND TRAINING
A Three-Course Feast for AI Enthusiasts 🍽️
Dive into a three-course series that arms learners with the:
Know-how to train.
Fine-tune.
Seamlessly integrate LLMs into AI products.
The first course is a jam-packed feast featuring over 50 lessons and 10 hands-on projects, all centered around LangChain and Deep Lake, a cutting-edge vector database for AI data.
Whether you're an AI newbie or looking to level up, this Full Stack LLM Bootcamp has got you covered! 🚀
JOBS AND TRAINING

Scale AI is looking for a Software Engineer focused on Cloud Infra.
Anthropic is hiring a Safety Cases Manager.
Meta is looking for a GenAI Program Manager to join their team.
Perplexity AI is hiring a Backend Software Engineer to work on their AI solutions.
Cohere is looking for a Director, Solutions Architect to join their team.
xAI is looking for an Information Security Engineer to join their team.

Thank you, see you next week!
Reply