I’ve seen people dumber than ChatGPT, it definitely isn’t sentient but I can see why someone who talks to a computer that they perceive as intelligent would assume sentience.
Sentient in a nutshell is the ability to feel, be aware and experience subjective reality.
Can an LLM be sad, happy or aware of itself and the world? No, not by a long shot. Will it tell you that it can if you nudge it? Yes.
Actual AI might be possible in the future, but right now all we have is really complex networks that can do essentially basic tasks that just look impressive to us because the are inherently using our own communication format.
If we talk about sentience, LLMs are the equivalent of a petridish of neurons connected to a computer (metaphorically) and only by forming a complex 3d structure like a brain can they really reach sentience.
Yes, you can debug an LLM to a degree and there are papers that show it. Anyone who understands the technology can tell you that it absolutely lacks any facility to experience
I’ve seen people dumber than ChatGPT, it definitely isn’t sentient but I can see why someone who talks to a computer that they perceive as intelligent would assume sentience.
We have ai models that “think” in the background now. I still agree that they’re not sentient, but where’s the line? How is sentience even defined?
Sentient in a nutshell is the ability to feel, be aware and experience subjective reality.
Can an LLM be sad, happy or aware of itself and the world? No, not by a long shot. Will it tell you that it can if you nudge it? Yes.
Actual AI might be possible in the future, but right now all we have is really complex networks that can do essentially basic tasks that just look impressive to us because the are inherently using our own communication format.
If we talk about sentience, LLMs are the equivalent of a petridish of neurons connected to a computer (metaphorically) and only by forming a complex 3d structure like a brain can they really reach sentience.
Can you really prove any of that though?
Yes, you can debug an LLM to a degree and there are papers that show it. Anyone who understands the technology can tell you that it absolutely lacks any facility to experience
Turing made a strategic blunder when formulating the Turing Test by assuming that everyone was as smart as he was.
A famously stupid and common mistake for a lot of smart peopel