Testing AI's Sports Smarts: Text to Video

You know how understanding sports can be tricky, right? Well, imagine trying to teach a computer to grasp all the rules, strategies, and history. That's what scientists have been working on with Natural Language Processing (NLP). They found that even the most advanced AI models struggle with sports, especially when it comes to complex scenarios. So, they tested these models with questions ranging from simple rules to intricate, context-specific reasoning. They didn't just stop at text; they also checked how well these models could understand sports from videos. This is what they call multimodal understanding. Turns out, there are some big challenges here. The AI models need to get a lot smarter to keep up with the dynamic nature of sports. To help improve this, researchers suggested a new benchmark based on existing sports data. They also pointed out common errors to guide future research.