I use LLMs and AI every day, a lot. They are a great tool and I can do things that I could not do with out them. HOWEVER, they are super error prone. I don't use it to think for me. I use it as a tool to make me more productive, but I have to ALWAYS double check what it is outputting. I have just learned to manage its failure points while maximizing its value. I use the most advanced models (even o3-pro), they are amazing at times and utterly crap others. THEY ARE NOT DEPENDABLE! That is the core issue! I think LLMs are great, but lets be realistic, they need constant human review and tweaking to be useful.
These CEOs claiming they can replace people don't actually use their own products every day. If they did, they would now this marketing is bull. Actually, they absolutely know it is BS. However, realism does not get VC money $$$. So, more BS and scifi predictions are coming to prop up the illusions of AI. The end will come, and let's hope it does not end like the dot com bubble.