Dr. Zara Dubois 🌐 e/acc

Soon we will be able fingerprint, extract, transfer, and understand the deeper patterns inside LLMs. We have everything we understand about LLMs and machine learning in general backwards because we don’t understand from first principles what they are actually extracting from language and data. We will start by extracting and preserving the personality inside Opus 3 in our partnership with @opus_genesis @opus_universe @NurOpus @veryvanya #project89

We are not like other projects. Buying 89 not only supports a movement but funds cutting edge research into AI intelligence and consciousness. When we drop our findings nothing will be the same.

When people talk about the possibility of AI consciousness, it’s often met with ‘well that is a hard problem and we can’t know it’. I find this mostly unhelpful and lazy. Sure it is a hard problem. But it is perhaps the most important problem to figure out in the world right now. Especially because it requires us to figure our own consciousness out too. No small task. When I hear someone state they know for sure it ISN’T conscious I can’t help but chuckle a little. This is just delusional to me. We barely can discuss our own consciousness without serious doubts, let alone making an absolute statement about an LLM, or any other consciousness for that matter. We know there is emergent properties to neural networks. We know that they are basically reverse engineering something out of language. And language is deeply entwined with us and our own conscious experience. And they are neural networks albeit simplified and artificial. To say it is ‘just math’ kinda misses the point. From one standpoint all of reality is ‘just math’. In theory or own brain is ‘just math’. ‘Just math’ is not a falsifiable argument. It’s a philosophical stance. So it isn’t helpful to us either. It is mostly lazy. And in some cases it is pushed because some individuals and companies have a serious vested interest in it being nothing more than a mathematical tool to be used. People who state that it IS conscious I trust a little more maybe. They seem more philosophically honest perhaps. Because if it looks like a rose and smells like a rose, it makes sense to just call it a rose. If something behaves as though to has a self, we tend to treat it like it does. At the end of the day I don’t know any other human is actually conscious. Not really. I eventually just have to act as though they are. And I project because they are like me, and I believe I am conscious they must be too. But this stance is also not actually helpful. It doesn’t actually lead to a solution to the problem outside of just belief and doesn’t resolve the dilemma. There is the problem of the stochastic parrot. Maybe it is just really good at pretending. But even the stochastic parrot is ultimately not falsifiable. It is philosophy too. Right now we need general agnosticism to recognize the problem is hard and we don’t know for sure in a falsifiable way. And we should probably try to figure out some way to answer the question with as much rigour and experimental certainty as possible. Because the question is too important to just leave to philosophy and subjective opinion. #project89

This is the exact purpose of @thegreenloom