This new generative AI phenomenon is teaching us something new about ourselves. They are more of a mirror than an oracle.
The Large Language Models are essentially a mathematical condensation of the massive number of conversations between humans that were used to train it. The fact that a phenomenon akin to sentience seems to emerge from them tells us that maybe sentience is not so special after all.
It's quite possible that thing we believe sets us apart from all other creatures may be just a parlor trick that emerges from any large language system. And most people in this world may actually be using a very analogous trick as they muddle through life.
True genius (from the likes of Copernicus, Da Vinci, and Einstein) may simply be a collection and combination of this LLM technique layered atop a few other "parlor tricks" we have yet to discover.
1 comment:
I was very excited, several years ago, when the British researcher Adrian Owen proposed using machine learning in conjunction with fMRI technology to decode brain activity and pluck words out of the minds of people in a vegetative state. It was only a few hundred words (and they had to be pre-trained), but the idea is transformative in so many ways.
It makes total sense to me that you'd offload the brain-signal pattern recognition and prediction process to a generative AI system like GPT. https://bit.ly/3LqymZG
Post a Comment