Raw LLM Responses

We have developed ML by mimicking what neurons and their connections do, when we virtualized enough neurons and fed an insane amount of data into this net suddenly these models are able to solve pretty complex problems, find creative solutions and reason about certain topics. This is called emergence, it's what our bodies and brains effectively did as well, a lot of simple things in a system suddenly, for some reason not super clear to us, complex behaviours emerge from the system and it is able to do more than its parts can individually. ML is built by mimicking what we learned in nature, we are actually not entirely sure why it works so well, but it does. I would argue these systems are absolutely heading towards sentience. Recently people have been experimenting with the "agent pattern" where multiple MLs get a different "job" for a task and validate each others work according to their given job. Not very different from how each part of the brain has a specific purpose in daily life and together they make you. I understand however why you're hesitant to call this "self-awareness", because it's not doing exactly what living things are doing. These models don't learn by themselves, or think. But instead they are a snapshot of intelligence. When these models were trained that's the moment they were learning and thinking, and we're just talking with the result. From a business perspective it's not interesting for an LLM to keep learning, to think by itself in the background, because we lose control over the conclusions it may draw and people with ill-intent may teach it the wrong things. It's not impossible however, and given that, I feel it's at least fair to start calling these model intelligent.