Thanks for the excellent video. I strongly believe SLMs have a lot of potential, especially with personal and small devices becoming more and more compute and storage capable. I believe they have the ability to transform the lives of those who may not have access to services that use and charge for cloud resources. We live in exciting times where each of us has the power to change the world for the better using the advancements in AI with all of the freely available knowledge out there. I have been around since the mainframe days and I am more excited than ever about SLMs democratizing access to sophisticated computing for all!
While the promise of small language models is intriguing, the video oversimplifies the real technical challenges. Having worked on fraud detection systems, I know that many specialized tasks still require the deep contextual understanding of larger models. The Spotify playlist analogy is clever, but downloading songs is vastly different from creating reliable specialized AI models that can perform complex tasks offline. SLMs definitely have their place, but suggesting they'll revolutionize fields requiring sophisticated analysis feels premature in 2025.
This matches exactly what i have been thinking for sometime now. Thanks dear. Looking forward to more of such videos.
Thanks for the excellent video. I strongly believe SLMs have a lot of potential, especially with personal and small devices becoming more and more compute and storage capable. I believe they have the ability to transform the lives of those who may not have access to services that use and charge for cloud resources. We live in exciting times where each of us has the power to change the world for the better using the advancements in AI with all of the freely available knowledge out there. I have been around since the mainframe days and I am more excited than ever about SLMs democratizing access to sophisticated computing for all!
While the promise of small language models is intriguing, the video oversimplifies the real technical challenges. Having worked on fraud detection systems, I know that many specialized tasks still require the deep contextual understanding of larger models. The Spotify playlist analogy is clever, but downloading songs is vastly different from creating reliable specialized AI models that can perform complex tasks offline. SLMs definitely have their place, but suggesting they'll revolutionize fields requiring sophisticated analysis feels premature in 2025.