A lot of what you said there is what I've been perceiving recently. It wasn't until I started to look into history that I realized why the study of consciousness was laughed at by neuroscience up to the 90s. You would think that field would be interested in that, but perhaps it was not a common thought among those who went through the training to get to the point of getting a graduate degree. So, back then it was considered a waste of time.
The same happens in medicine, where I recently learned that doctors didn't administer anesthesia to babies getting surgery until the mid 80s. The thought is that babies nerves weren't developed so they didn't feel pain. Ok let's say one believes that at the time. How does one still operate on the baby who is crying and screaming? Was there no empathy to push for anesthesia, even just to calm the baby?
"Reality isn’t “really” mechanistic any more than objects are “really” there."
I agree that discrete objects as such are an ideal, but my prior on the universe is that it is mechanistic. Of course, the experience of sapience is something that is not amenable to being modelled with "quarks and stuff". Our bodies have materials that have particular properties, alone and in concert, that eventually allow philosophy to happen.
You glossed over LLMs-as-left-hemisphere-minds analogy; I think it's very strong. LLMs await external goals, they choose the level of abstraction to solve a problem, they filter and isolate, they articulate words. If given a poorly specified or impossible goal, they will keep spitting out tokens pointing in what the LLM thinks the general direction of the goal is. This feels very similar to distraction patterns.
A lot of what people see as LLM failures are caused by lack of volition: the thing that in humans comes from the animal and not the tool. "Write me a story about grief" is a dogshit prompt because the LLM has nothing it *wants* to tell you about grief, no heavy sadness it wants to get off its chest. The writing the LLM produces in response to such a prompt is correspondingly dogshit — the sentences all fit and the metaphors are appropriately somber, but nothing meaningful about grief is being communicated.
Perhaps this will make it easier for people to relate to their own minds in the way you write about. If your mind has failed to solve your problem it's either because the problem is too hard or, what's likelier, you gave it the wrong prompt. You asked it for "make me stop experiencing bodily anxiety about the task being procrastinated on" instead of "generate an action plan to get the task done". And from the opposite direction, I relate to LLMs explicitly as extensions of my own left hemisphere as opposed to a separate agent I have a relationship with. It works pretty well.
WOW, Just read https://blog.morphenius.com/p/valentines-logos
A lot of what you said there is what I've been perceiving recently. It wasn't until I started to look into history that I realized why the study of consciousness was laughed at by neuroscience up to the 90s. You would think that field would be interested in that, but perhaps it was not a common thought among those who went through the training to get to the point of getting a graduate degree. So, back then it was considered a waste of time.
The same happens in medicine, where I recently learned that doctors didn't administer anesthesia to babies getting surgery until the mid 80s. The thought is that babies nerves weren't developed so they didn't feel pain. Ok let's say one believes that at the time. How does one still operate on the baby who is crying and screaming? Was there no empathy to push for anesthesia, even just to calm the baby?
It seems that we only just woke up....
"Reality isn’t “really” mechanistic any more than objects are “really” there."
I agree that discrete objects as such are an ideal, but my prior on the universe is that it is mechanistic. Of course, the experience of sapience is something that is not amenable to being modelled with "quarks and stuff". Our bodies have materials that have particular properties, alone and in concert, that eventually allow philosophy to happen.
You glossed over LLMs-as-left-hemisphere-minds analogy; I think it's very strong. LLMs await external goals, they choose the level of abstraction to solve a problem, they filter and isolate, they articulate words. If given a poorly specified or impossible goal, they will keep spitting out tokens pointing in what the LLM thinks the general direction of the goal is. This feels very similar to distraction patterns.
A lot of what people see as LLM failures are caused by lack of volition: the thing that in humans comes from the animal and not the tool. "Write me a story about grief" is a dogshit prompt because the LLM has nothing it *wants* to tell you about grief, no heavy sadness it wants to get off its chest. The writing the LLM produces in response to such a prompt is correspondingly dogshit — the sentences all fit and the metaphors are appropriately somber, but nothing meaningful about grief is being communicated.
Perhaps this will make it easier for people to relate to their own minds in the way you write about. If your mind has failed to solve your problem it's either because the problem is too hard or, what's likelier, you gave it the wrong prompt. You asked it for "make me stop experiencing bodily anxiety about the task being procrastinated on" instead of "generate an action plan to get the task done". And from the opposite direction, I relate to LLMs explicitly as extensions of my own left hemisphere as opposed to a separate agent I have a relationship with. It works pretty well.