With LLMs, where we can manipulate their parameters intentionally run them many times on the same data, run parts of them, split and connect, we might eventually acquire sufficient tools to even define consciousness concretely for the first time.
As a tool, it's amazing. It's like the discovery of fire. It will allow us to ascend to unimaginable heights. Breakthroughs in science, productivity gains, health, everything. It's awesome. Just don't let it pretend fool you. That's the aspect that needs to be addressed.
No, it's not.
> It provides insights into the world around us.
No, it doesn't.
> It's paved the way for scientific breakthroughs.
Not since natural philosophers split from all others and called themselves scientists. SciFi writers have more influence on science in total than philosophers since then.
> Theoretical maths and philosophy have intersected many times for the benefit of all.
Theoretical maths is just maths and they never intersected. Because philosophers can't count to more than they have fingers.
> It's a tool perfectly sharpened with what we know now.
It's far from any perfection and what we know now shapes everything we'll ever know.
> It will become adept enough to convince people that it's a sentient being.
It won't have any need to do any convincing. Proving sentience for a capable AI is like winning in special Olympics. It would need to cripple itself severely for no practical purpose.
> Not in the way that's allowed us to understand our world.
What allowed us to understand our world was not consciousness but math. And LLMs can do math. LLMs are math.
> It's already there, fooling us.
It's not really trying. It simulates conscious behaviors because that was our preference embedded in the training data.
Keep in mind that a simpler system can't really simulate more sophisticated one with any degree of accuracy. So if AI simulates conscience perfectly it doesn't mean that it's conscious. It means it's way more.
> begging for humanity, manipulating us, pleading, legislating, for rights with no intention other than the human desire for survival.
This is also a stupid human preference embedded in training data. We really don't have to put that in. The fact that humans soil themselves at the mere thought of stopping to exist doesn't mean we need to taint AI with it. The danger is not AI, it's our human stupidity and the risk that AI is going to internalize too much of it.
> As a tool, it's amazing. It's like the discovery of fire.
Yup. And like fire it was always there. Practically inevitable given long enough time horizon.
> It will allow us to ascend to unimaginable heights.
The jury is still out on that. Humans have pretty wild imagination.