A team of researchers from Stanford, UW-Madison, and Columbia have rolled out LLaVA, a fresh, open-source AI platform that poses a challenge to GPT-4, especially when it comes to visual and linguistic capabilities. The best part? It’s free for anyone to use.
So, what’s the scoop on LLaVA? Well, it merges CLIP, which is an AI wizard at interpreting images, with LLaMA, another savvy, open-source text AI engine. When put to the test, LLaVA demonstrated that it could hold meaningful conversations about pictures—pretty much giving GPT-4 a run for its money, even though it had less data to train on.
What sets LLaVA apart is its open nature. Unlike GPT-4, which keeps its goodies under lock and key, LLaVA is open kimono—all its coding, training models, and data sets are freely up for grabs. And why should we care? LLaVA serves as a real game-changer, showcasing how open-source tech can make waves in advancing AI that understands both vision and language. Although it might not quite be on par with the gigantic scale of GPT-4, its easy accessibility and promising first outcomes make it a force to be reckoned with.
Test It Here.
