Did you hear the headlines, like that chatbot telling a guy he should divorce his wife or that AI meal planner app proposing a crowd-pleasing recipe featuring chlorine gas? Well, this TED talk by Sasha Luccioni, an AI Researcher, dove into how it affects our lives right now. The real problem is not that AI will take over this world or wipe out humanity, but some other issues need to be tackled at this time.
Something interesting I learned about big AI models like ChatGPT is that training these models uses a ton of energy, like powering 30 homes for a whole year! And the carbon emissions? It's like driving your car around the Earth five times so that someone can tell a knock-knock joke. That's a lot, right? So, they made a tool called CodeCarbon to help us make more intelligent choices, like picking more eco-friendly models.
Another part was about artists and authors. Imagine if your art or writing was used to train an AI model without your say-so. Not cool, huh? There's this tool called "Have I Been Trained?" that lets you check if your stuff is in those big datasets. Artists even used it to sue some AI companies for using their work without permission.
There is another big problem with AI called bias, especially in systems that recognize faces. Dr. Joy Buolamwini's story shows that AI often has trouble seeing the faces of women of color. This bias can cause serious problems, like when the police use these AI systems. It can lead to innocent people being accused and even sent to jail for something they didn't do. These AI systems are like mysteries because their creators don't know exactly why they're biased. A special word for them is "black boxes." Biased AI can make things worse, like when it's used to create pictures of criminals based on descriptions, and it ends up spreading harmful stereotypes.
In a nutshell, the TED Talk reminded me that AI is not just about robots taking over the world. It's already part of our lives, and we must ensure it's doing good things. We need to focus on fixing the real problems AI has right now. After all, we're still figuring out how to use AI responsibly, and we all get a say in where we're headed.