home.social

Search

8 results for “mattjhayes”

  1. In my latest blogpost, we continue our open-source journey with AI, building a lab system to understand how RAG actually works. Is it all it's cracked up to be?
    mattjhayes.com/2025/10/31/putt
    Here's an analogy from the conclusion:
    "Keeping an LLM under control in the face of random and potentially malicious user input is like trying to ride horseback with no reigns,
    steering only by whispering instructions in the horse’s ear. When you’re used to the predictability of driving a car, you may find this unsettling and a bit odd. Results will vary."
    No matter what your views are on AI, it is worth taking the time to understand the
    technology so to have an informed view on it.
    #AI #RAG #OpenSource #LLM #Ollama #ChromaDB #Codeberg

  2. @ibboard
    Nice call-out between what I call deterministic vs statistical computing. I’m on the fence re statistical computing i.e. AI LLMs etc. There have been amazing breakthroughs and LLMs are useful in some scenarios. But horses for courses - the current AI mania is forcing people to deploy LLMs for tasks better suited to deterministic computing (example: automation of repetitive tasks) Hopefully things will calm down over time and there will be industry guidelines on when statistical algorithms should be used, and with what guardrails, and when they shouldn’t (not reliable enough, too expensive, insecure, inscrutable, etc.). The current move towards exposing the true AI usage costs to users will help #AI #LLMs

  3. @jpoesen Glad to know it wasn’t just me. Hopefully it can be fixed soon as it’s really inconvenient when you’re setting up a headless RPi to find it isn’t reachable and then have to plug in KVM and enter parameters again 🥺 #RaspberryPi

  4. Free bike charging station 👍
    Unexpected security error message 👎
    Everyone sees you run Windows: 😭🤮

  5. My local news site published a piece promoting a book with the premise that we have no free will (rnz.co.nz/life/wellbeing/world). This is deeply dangerous, giving a free pass for people to commit bad deeds without repercussions, and also overly simplistic. While our brains are wired to prefer simple ‘a or b’ answers, and this helps sell books, it doesn’t align to the complex messy world we live in. Free will is a component of our existence, with its availability varying between individuals, and over time. Sure, luck and circumstance play large roles, but free will is what defines us. #FreeWill #RNZ #Aotearoa

  6. I stumbled across this visually-impressive interactive explainer of the #LLM Transformer Architecture poloclub.github.io/transformer . Everyone should have a good mental model of LLMs to avoid falling into the trap of thinking they think like a human... #AI #generativeAI #GeorgiaTech