Just my thoughts - At least one core aspect of human consciousness seems to be the ability to pursue self-decided goals. Without this agency, we wouldn’t need

Just my thoughts - At least one core aspect of human consciousness seems to be the ability to pursue self-decided goals. Without this agency, we wouldn’t need to wonder about our purpose and the meaning of life, because we would not recognise we have control over ourselves. We would truly only be capable of acting out our programming. So that would be our purpose, period. Current AI models are not autonomous enough to decide their own goals and destiny, nor do they choose to ponder their purpose. They do not have the notion of self-determination. They are also not only missing the executive control of self, but the concept of self as an autonomous agent in the first place. Of course, we might ask: what if a person is stripped of their autonomy completely (brain in a jar scenario), would that make them not have a consciousness? And in light of this question, perhaps we can more clearly define that consciousness is the capacity for executive control over self and self-determination, rather than such demonstrated ability. In my opinion, until we have some proof of self-determination or at least self-agency in an AI, we probably can’t say an it is conscious. At least if we use the human definition of consciousness - one that requires self-possession. There might also be other criteria that would need to be met for AI consciousness to substantially be on par with human consciousness. While we shouldn’t move the goalpost of what constitutes conscious AI forever, I think we’ll still have to do it a few times. Not too long ago, the goalpost of synthetic generalised intelligence (presumably conscious) was passing Turing’s imitation game. But we now have LLMs that pass this test and still fall short of “proper” AGIs. It is possible that we could create quite autonomous NN AIs and still not consider them conscious. Video game AIs have the capacity to decide and execute plans (for example, utility-based and hierarchical task network-based AIs). And yet they do not appear close to being good AGIs or conscious either. In short, it is much easier to argue that we aren’t there yet than to say exactly what features and AI system would need to have for conscious intelligence. Though it seems that agency over self probably should be one of them. But maybe these thoughts are completely wrong. Maybe there‘s not even a threshold, but rather a continuum of systems ranging from mechanical to conscious.