10 Things AI Is Still Surprisingly Bad At

AI can write code, solve complex problems, and explain almost anything. So why does artificial intelligence still struggle with surprisingly simple tasks like counting objects, basic math, and knowing when it's wrong? In this video, we explore 10 things AI is still surprisingly bad at, revealing the hidden weaknesses of modern AI systems, including ChatGPT, large language models, and advanced multimodal AI. From AI hallucinations and unreliable answers to long-term planning failures and physical reasoning mistakes, you'll discover why even the most advanced AI models can make errors that humans would never expect. In this video, we cover: 10. Counting Objects Reliably 9. Exact Arithmetic 8. Understanding Physical Space 7. Knowing When It's Wrong 6. Remembering Very Long Tasks 5. Following Complex Instructions Perfectly 4. Distinguishing Truth From Plausible Nonsense 3. Long-Term Planning 2. Reasoning About the Real World 1. Producing Completely Reliable Answers AI technology is improving rapidly, but intelligence and reliability aren't the same thing. Understanding these limitations is essential as artificial intelligence becomes part of our everyday lives. Which AI weakness surprised you the most? Share your thoughts in the comments! Subscribe to This Guy Knows Stuff for more fascinating videos about technology, artificial intelligence, science, and the surprising facts behind how our world works. #ArtificialIntelligence #AILimitations #ChatGPT