notion image
    Follow
    🔥🔥🔥🔥🔥
    Reply
    Love this!
    Reply
    Doesn’t this seem obvious, I mean, who doesn’t check their work a few times over before providing a final answer?
    Reply
    Im curious to see what might happen when we place these small parameter domain specific models into ecosystems of other domain specific small parameter models in which they all can communicate and solve problems recursively collaboratively. I could see a similar architecture bringing up the efficiency of large general models! This is truly exciting stuff.
    Reply
    Sounds kinda like DeepSeek was a few months ago. Very interesting. 7M is phenomenal.
    Reply
    Yeah, but not on the same range of subjects, right. I.e. can the small model answer the numbers as the large models? The small ones can't have the same amount of training data.
    Reply
    Bigger isn't necessarily better. I think this approach is quite interesting and appreciate the efficiency especially with the resources that can be saved.
    Reply
    View all 1 replies
    Thanks for the video 👏👏
    Reply
    View all 1 replies
    It’s not peer reviewed and it’s published by an profit-seeking company. It’s motives are not necessarily to advance the field as much as it is to gain notoriety and profit.
    Reply
    Add lazy evaluation and 🔥
    Reply
    Is this the sign that the AI bubble's about to burst? Instead of pouring money into infrastructure, similar results can be achieved with simpler models.
    Reply
    View all 5 replies
    Small Language Models are future just like everyone will have there "own ai" in there phone
    Reply
    View all 3 replies
    This could open the doors for edge AI integrations, bringing AI to edge would make it more accessible and reduce reliance on cloud and massive data centers, we could see so many everyday tech coming with integrated ai
    Reply
    View all 2 replies
    I mean. I’ve been saying this for months being patently pending in this tech.
    Reply
    Please note : this model is domain specific, not intended for general use cases. Although it could make future LLMs way more efficient.
    Reply
    View all 4 replies
    Log in to like or comment.
    More posts from parthknowsai
    Researchers at Google recently published a new tool to help revolutionize how textbooks work in education #ai #education #learning #tech #learn #viral
    Thinking vs computing part 2 - researchers at Oxford university conducted thought experiments to show AI does not think, instead it just remixes ideas from its training materials. #ai #tech #learn #science #education #reel #viralreels
    Scientists and researchers at Florida state recently conducted a study which found that AI & LLM tools are changing human vocabulary #ai #tech #education #science #research #learn #learn #fyp
    California places strict rules against cloning actors faces without their permission. Learn more. #ai #tech #education #science #research #learn #fyp
    English
    © 2025 Instagram from Meta
    Hobby note - Sam van Noord | Sam van Noord