> For sure there's no clear specific on exactly what intelligence is, but I'd just go on to say it's what you'd consider intelligent as any other human. That's

> For sure there's no clear specific on exactly what intelligence is, but I'd just go on to say it's what you'd consider intelligent as any other human. That's not the definition, like many people you're haphazardly extending the meaning of a word to somehow mean AGI just so you can shift goalposts. > intelligence /ɪnˈtɛlɪdʒ(ə)ns/ noun 1. the ability to acquire and apply knowledge and skills. The second definition is for military information which we can disregard but clearly Oxford says there are two actions a 'thing' must do to be intelligent: learn and use the learned knowledge. LLMs can easily demonstrate both. Ergo, they are intelligent. Biological thinking or whatever is not a requirement. Hell, your average classical AI approaches are also intelligent, albeit in a very a limited area of expertise, e.g. learn movement of an object and apply that information to output predictive control. Your second assumption is that whatever flaws you're seeing are somehow a core fault of the architecture. Most of the current cloud models are tuned to be politically correct corporate drones which probably isn't the best demonstration of what they can do. They won't get into arguments with you nor will they hold strong opinions in an effort to not offend so the company doesn't get sued or cancelled. At least in my experience some non-assistant tuned Llama models are a lot more "human" in terms of their opinions. Regardless it's kind of a bullshit metric to consider for inteligence. I would expect that humans are only hard wired to hold strong opinions because it evolutionarily makes us harder to manipulate and more likely to survive. But yes nobody's saying they're human level conscious brains and it's foolish to change words to mean what they don't just to illustrate that.