Google's new AI generates video games

PLUS: Mistral's massive new model

Welcome, AI enthusiasts.

Google just rolled out a new AI model called ‘Genie‘, and it could be a game-changer...

The AI can build video games from text or image prompts, opening the possibility for anyone to design their own playable worlds in the near future. Let’s get into it…

In today’s AI rundown:

  • Google’s ‘Genie’ enables prompt-to-video games

  • Mistral reveals Large, Le Chat, and a new partnership

  • Convert videos to GIFs using ChatGPT

  • Meta’s MobileLLM AI for on-device use

  • 6 new AI tools & 4 new AI jobs

  • More AI & tech news

Read time: 4 minutes



Image source: Google

The Rundown: Researchers at Google DeepMind just unveiled Genie, an AI system that can generate interactive video games from just a text or image prompt — without any prior training on game mechanics.

The details:

  • The model was trained on over 200,000 hours of 2D platform game videos, learning to understand gameplay autonomously.

  • Users can control game characters and objects on a frame-by-frame basis despite Genie not being trained with action labels.

  • The foundation model can also determine who the main character is and enable a user to control that character in the generated world.

  • Genie is a research-only model for now and can only create games at 1 FPS — limiting its current viability despite future promise.

Why it matters: Anyone will be able to create their own entirely imagined virtual worlds soon, and that’s a wild sentence to say out loud. The model’s ability to learn and develop world models is also a massive step towards general AI agents.

See Genie’s wild game generations in action here.


The Rundown: OctoAI makes it simple to evaluate, optimize, and scale the newest AI models at the best price/performance.

Why flexibility matters:

  • LLMs are getting better, faster, and cheaper

  • New models are launched nearly every week

  • Quickly productionizing models is ‘do or die’ to compete

OctoAI’s singular API will help you swap cost-optimized endpoints into your app at scale.


Image source: Mistral

The Rundown: French AI startup Mistral just released its highly anticipated Mistral Large model while also announcing a new chatbot and major partnership with Microsoft.

The details:

  • Mistral Large boasts top-tier reasoning abilities and multilingual fluency, claiming to rank second only to GPT-4 on key benchmarks.

  • The new model has a 32k context window, and its API usage pricing comes in cheaper than GPT-4T at $8 / million tokens.

  • The company also launched a beta version of its ChatGPT competitor ‘Le Chat’.

  • Microsoft announced a new partnership with Mistral, bringing the startups' models to Azure users while providing infrastructure and scaling resources.

Why it matters: Mistral AI is now competing with the world’s top models. Large’s capabilities mark a serious milestone — and with massive funding and new strategic partnerships (like Microsoft), the startup is rapidly ascending the ranks as a dark horse AI leader.


Image source: Chase Lean (@chaseleantj on X)

The Rundown: ChatGPT can now easily convert a .mp4 video file to a GIF by reading, extracting, and converting frames using Code Interpreter.

Simple step-by-step:

  1. Go to ChatGPT and create a new chat.

  2. Upload your .mp4 video and prompt ChatGPT to convert it to a GIF using the prompt in the image above.

  3. Download the GIF file provided by ChatGPT.

Bonus tip: You can ask ChatGPT to slow down or speed up the GIF based on the output. For example, try prompting ChatGPT to “slow down the gif by 30%”.


The Rundown: Learn 10+ AI tools and practical workflows to help you increase your productivity and grow your business with GrowthSchool’s free 3-hour workshop.

You’ll learn how to:

  • Accelerate your business with AI insights

  • Automate daily tasks to boost creativity & productivity

  • Integrate AI for Data Analysis and upgrade your presentations

Register to join 600,000+ people who’ve already completed the workshop. It’s free for the first 100 people!


Image source: Meta

The Rundown: Meta researchers just proposed MobileLLM, a family of optimized smaller models for on-device applications, which set new SOTA benchmarks for accuracy among similar models.

The details:

  • MobileLLM emphasized model architecture over data and parameter quantity, contrary to prevailing beliefs.

  • The 125M parameter MobileLLM is 2.7% more accurate than the previous best small models.

  • The models show particularly strong performance on chat and API calling, vital for mobile app functionality.

Why it matters: While it’s clear that mobile devices are about to have an AI explosion, incorporating high-performing models is a science still being worked out. Meta’s impressive results highlight a new potential architecture for optimized on-device AI models.


  • 🎯 Dart - Project management with AI roadmap planning, report generation, automatic task execution, and more*

  • 💡 Saner AI: An AI management system to help you capture, find, and develop ideas

  • 🤖 DocGPT- GPT for Google Docs AI, Sheets, Gmail, Slides, Forms

  • 📹 GoEnhance AI: AI-powered animation and image enhancement for video content

  • 🤳🏻 Cartoonifier: Transform your photos into whimsical cartoons

  • 🎬 Videotok: Create AI-Powered TikTok, Reels, and Shorts Videos with Ease

  • 👨🏻‍🔬 Anthropic- Research Engineer, Alignment Science

  • 🔍 Open AI- Account Associate - EDU accounts

  • 👩‍💻 Synthesia- Senior/Staff/Principal Frontend Engineer

  • 🛡️ Shield AI- Senior Staff Mechanical Engineer

*Sponsored listing


Nvidia CEO Jensen Huang said he believes humans will no longer need to know how to code in the future, with AI replacing programming languages via human prompts.

Deustche Telekom revealed its app-less, AI-powered ‘T-Phone’ at the Mobile World Congress 2024, also announcing a partnership with Perplexity AI for real-time info.

Google executive Brian Rakowski revealed that he expects the company’s Gemini AI model to be integrated into Android phones in 2025.

Avolution AI researchers released an open-source AI video diffusion model called AnimateLCM, allowing for text and image-to-video generations.

Google plans to bring image generation back to Gemini in ‘the next few weeks’ after controversy surrounding the model’s overly diverse tuning and issues with generating people.

Figure CEO Brett Adcock shared new footage of the startup’s Figure 01 humanoid robot completing real-world tasks like staking and moving crates, operating completely autonomously without human intervention.

Google DeepMind CEO Demis Hassabis said he predicts AI-designed prescription drugs will be in clinical trials within the next few years.



Get your product in front of over 500k+ AI enthusiasts

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.


How would you rate today's newsletter?

Vote below to help us improve the newsletter for you.

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

Join the conversation

or to participate.