AI is not a super-intelligence, and they know it but hype it as a sentient being that will create utopia. They need a justification for the insane amount of money already spent and will say just about anything in desperate attempts to recoup losses. AI is, frankly, less a superintelligence than a form of super-stupidity with enormous power and no capacity for remorse or learning from previous harm caused.
IMO, we will eventually get a superintelligent AI, but LLMs will be only a part of that. The LLM would be like the language center of the brain. The issue now is that we’re dealing with the language center only, without the rest of the brain attached, and expecting it to act like a complete brain.
It’s already happening. A lot of what we’re seeing as “intelligence” in tools like Codex or Code are a combination of neural net and symbolic AI.