News has arrived recently that they are releasing a new AI model to be trained and that model will be an open sourced one.
They call the new model NVLM-D-72B, and yes it's very confusing for name if you don't know much about AI tech but anyways this model shows a strong step forward in the world of open-source AI.
I believe we need a lot of these open source AI models so that anyone and everyone can have access to the tech and contribute more to the entire AI industry. It's basically like how Google's operating system Android is open sourced and thanks to that, we've got more android phone companies than Apple IOS companies. I guess that's what's going to happen with Nvidia's model.
With 72 billion parameters this new model performs well in text tasks and also understands images and math. I heard that's what GPT 5 is going to get better at, generating better text to image results and image to text which is basically description of whatever image you decide to share with Chatgpt.
Article generation, math and image generation is a very clear challenge to models like GPT-4 which at this moment still remain closed to the public.
I think by NVIDIA making its model open-source, they'll encourage more research and development in AI with lots of people getting their hands on the codes and building better AI tools and services. I made mention in my blog last week that 2025 will see AI domination and I think this is how it'll happen.
Developers and researchers will have complete access to the model to learn from it, but of course with limits on commercial use. We'll see more breakthroughs in areas like coding and data analysis.
But while we're getting good with AI, the negative side of it is still becoming a big issue. Some believe AI could become too powerful and even predicting that it might harm humanity in the future just like the Terminator movie.
The good thing is that NVIDIA’s approach allows research but not commercial misuse and that can manage the risk.