CAG vs RAG: Which One to Use?

CAG vs RAG: Which One to Use?

If you're using ChatGPT or other AI models, you've probably noticed they sometimes give incorrect information or hallucinate.

RAG helps solve this by searching through external documents, but this new approach takes a completely different approach - and it might just be what you need!

Good morning everyone! As always, this is Louis-Fran?ois, co-founder and CTO at Towards AI, and today, we'll dive deep into something really exciting: Cache-Augmented Generation, or CAG.

In the early days of LLMs, context windows, which is what we send them as text, were small, often capped at just 4,000 tokens (or 3,000 words), making it impossible to load all relevant context.

This limitation gave rise to approaches like Retrieval-Augmented Generation (RAG) in 2023, which dynamically fetches the necessary context.

As LLMs evolved to support much larger context windows—up to 100k or even millions of tokens—new approaches like caching, or CAG, began to emerge, offering a true alternative to RAG...

Learn more in the video (or written article here):


And that's it for this iteration! I'm incredibly grateful that?the What's AI newsletter?is now read by over 20,000 incredible human beings. Click here to share this iteration with a friend if you learned something new!


Looking for more cool AI stuff? ??

Want to share a product, event or course with my AI community? Reply directly to this email, or visit my Passionfroot profile to see my offers.


Thank you for reading, and I wish you a fantastic week! Be sure to have?enough sleep and physical activities next week!


Louis-Fran?ois Bouchard

Imtiaz M.

Expertise in full IT service life cycle, focusing testing, delivery and analysis to Improve infrastructure and application observability, performance, availability and reliability.

4 周

CAG make more sense needing less GPU/TPUs, response is faster with less memory depending upon rule requirements , embedding, fine tuning and how much context aware response is needed, surly faster but more effectively and efficient for specialized area/ jobs

回复
nick trendov

I help teams navigate and negotiate change. Applying real-time alerts to align products, influencers and customers is my forte. ???? ??

4 周

?? KNOWLEDGE, just like TRUST, is a simple vendor MYTH ?? ???????? ?????? ???? ??

回复

要查看或添加评论,请登录

Louis-Fran?ois Bouchard的更多文章

  • Want to start programming in the AI era? This is for you...

    Want to start programming in the AI era? This is for you...

    Good morning! If you’ve been wanting to break into AI development but feel like your coding foundation isn’t quite…

  • Using AI for Writing

    Using AI for Writing

    Good morning! We’ve (Towards AI) been using AI to research, plan, help us with drafts, and refine our lessons for our…

    4 条评论
  • How LLMs Are Changing Every Job

    How LLMs Are Changing Every Job

    Good morning! Today, I’m sharing our third video out of 6 we made for our “8-hour Generative AI Primer” course. In this…

  • LLM Developers: The future of software development

    LLM Developers: The future of software development

    Software engineers vs. ML engineers vs.

    1 条评论
  • Real Agents vs. Workflows

    Real Agents vs. Workflows

    What most people call agents aren’t agents. I’ve never really liked the term “agent”, until I saw this recent article…

    1 条评论
  • Why LLMs Are the Future of Work

    Why LLMs Are the Future of Work

    Good morning! Today, we start the new series of videos for our most recent Towards AI course: 8-hour Generative AI…

    1 条评论
  • Introducing Our 8-Hour Generative AI Primer

    Introducing Our 8-Hour Generative AI Primer

    Once again, I’m super excited to share some news from the Towards AI team—we’ve just launched a brand-new 8-hour…

    11 条评论
  • Best Practices for Building and Deploying Scalable APIs in 2025

    Best Practices for Building and Deploying Scalable APIs in 2025

    Good morning! When we talk about building powerful machine learning solutions, like large language models or…

    1 条评论
  • Lessons from Nvidia Minitron

    Lessons from Nvidia Minitron

    We’d all love to build a model from scratch like Llama, but how realistic is that? The computing, architecture, and…

    3 条评论
  • When NOT to Use Large Language Models

    When NOT to Use Large Language Models

    Good morning! Are you feeling the pressure to adopt the latest AI trends but aren’t sure if it’s the right move for…

    3 条评论

社区洞察

其他会员也浏览了