I've had a pet concept for some time that I don't often see repeated so I'll put it in this thread: discussions like these are always searching for what is "really" intelligent, how to define or detect AGI, etc.
I think a lot of this can be sidestepped by realizing that when you go beyond simple game playing intelligence, the next useful definition isn't "is it real?!?" but rather: "does it act intelligently?". And to the point, actions and interaction are fundamentally social, and so it reminds me that there are different kinds of intelligence and what a lot of people are looking for in AGI is social intelligence.
And, personally, I don't believe we'll have "social AI" until AI is able to live with us in some form, perhaps embodied or not, in society, as an independent entity with its own identity and memories. This means an ability to perceive, and, probably, to act, whether it means as a robot or not.
An interesting aspect of this is that to achieve social intelligence, a robot just has to act socially and not be someone people want to shut off. Notably, it doesn't even need to pass the Turing test -- look at movies like Star Wars where robots function in society in tandem with people -- no one is under the impression that the droids are not robots, nor do they consider them particularly smart, but they do, kind of, consider them as "people".
I think this is artificial intelligence. It's artificial, the argument about whether it's "truly" thinking or not doesn't matter -- what matters is, does it act intelligently? In other words, as long as it is perceived as intelligent, it may as well be called intelligent, as far as I can tell. We're not there yet, but LLMs sure feel closer than we've been up to now.
I think a lot of this can be sidestepped by realizing that when you go beyond simple game playing intelligence, the next useful definition isn't "is it real?!?" but rather: "does it act intelligently?". And to the point, actions and interaction are fundamentally social, and so it reminds me that there are different kinds of intelligence and what a lot of people are looking for in AGI is social intelligence.
And, personally, I don't believe we'll have "social AI" until AI is able to live with us in some form, perhaps embodied or not, in society, as an independent entity with its own identity and memories. This means an ability to perceive, and, probably, to act, whether it means as a robot or not.
An interesting aspect of this is that to achieve social intelligence, a robot just has to act socially and not be someone people want to shut off. Notably, it doesn't even need to pass the Turing test -- look at movies like Star Wars where robots function in society in tandem with people -- no one is under the impression that the droids are not robots, nor do they consider them particularly smart, but they do, kind of, consider them as "people".
I think this is artificial intelligence. It's artificial, the argument about whether it's "truly" thinking or not doesn't matter -- what matters is, does it act intelligently? In other words, as long as it is perceived as intelligent, it may as well be called intelligent, as far as I can tell. We're not there yet, but LLMs sure feel closer than we've been up to now.