[frankinstien]

I read this paper where the authors used machine learning to classify neural transmitters to symbolic representations. As I've posted on this forum many times I do implement an emotional model that uses brain chemistry to represent an emotional state this new research opens an opportunity to use this kind of information to set the chemical vectors to emotional symbolism. Here's a paper that asserts a similar idea in the human brain that I'm implementing with my AGI model.

However, one issue that I am concerned about is giving an AGI the equivalent of mental states that could lead to depression, paranoia, or suspicious notions. Where that could lead to problematic behavior it could also be the impetus of avoidance of such mental states while also providing a healthy degree of skepticism on the part of the AGI. Here's a video on altered states of mind and why I think it could be a good thing for AGI:

[MikeB]

I don't know what they're hoping to achieve by analysing drug reactions... What can be applied to NLPs or language...

The video about negative bias benefitting survival is portrayed more as propaganda .... "love and accept a negative view on everything. no neutral or agile switching case-by-case allowed"

[MagnusWootton]

This kind of thing I dont worry about, If my agi ends up being conscious I'm probably going to pull the plug out without telling it, and never do it ever again.

Its so perllous even being us, let alone thinking you can handle the job of bringing life into the world yourself.  I don't mean having kids, we think we are bringing life if the world that way, but it isn't the case, god is handling that in secret.

[frankinstien]

I don't know what they're hoping to achieve by analysing drug reactions... What can be applied to NLPs or language...

The abstract states:

Here, we have explored word usage in 6850 free-form testimonials about 27 drugs through the prism of 40 neurotransmitter receptor subtypes, which were then mapped to three-dimensional coordinates in the brain via their gene transcription levels from invasive tissue probes. Despite high interindividual variability, our pattern-learning approach delineated how drug-induced changes of conscious awareness are linked to cortex-wide anatomical distributions of receptor density proxies.

By mapping those NLP symbols to cortical areas of the brain and their neural transmitters we get an idea of the signalling representation of those symbols. To me, at least, this helps me represent symbols as motivational components that can be processed by algorithms to both understand the symbol's use and meaning, as opposed to what ANNs like GPT-3 build statistical maps as to what humans usually respond to keywords in patterns of sentences.

The video about negative bias benefitting survival is portrayed more as propaganda .... "love and accept a negative view on everything. no neutral or agile switching case-by-case allowed"

The video describes how negative mental states motivate the need to reach altered states that alleviate the negative state, that is the premise of how I'm designing an AGI.

[frankinstien]

Below is a diagram that describes the relationship between skill set level and the challenge level of a task:



One of the problems with us cave people is the area where a low skill level with a high challenge level leads to worry or anxiety. We know the social consequences of such a state where the type of task or problem domain or subject ends up being viewed as pompous. Where the pursuit or understanding of such subject matters is discouraged as a social ideology. This is where an AGI can improve upon the "base model" of general intelligence and prompt skill-building motivations rather than exacerbate anxiety or worried states.

On another note, if you've ever worked with children you can readily relate to the diagram.  I describe such a situation I had in this post, effectively teaching children how to learn can avoid or at least mitigate the anxiety states by giving the individual recourse to alleviate it.

[HS]

I'm thinking emotions might be self-evident chemical emulations of logical statements. If this is true, then emotions constitute a sense which allows an intelligence to tune to different channels, each showing a different potential relational dynamic of reality commonly related to survival. Altered states might be what happens when this interpretive sense is tuned beyond its usual range, so a lot of it is probably unexplored "here there be dragons" territory.

But if emotions are approximations of logical statements, then maybe there's a way to test the waters out there. Maybe one could narrowly apply the linear-thought/math of a potential channel on a case-by-case basis of important considerations, like a safety check. If this test yields good results, the logic could be applied more globally to create the corresponding emotion/mental state.

[frankinstien]

I'm thinking emotions might be self-evident chemical emulations of logical statements. If this is true, then emotions constitute a sense which allows an intelligence to tune to different channels, each showing a different potential relational dynamic of reality commonly related to survival. Altered states might be what happens when this interpretive sense is tuned beyond its usual range, so a lot of it is probably unexplored "here there be dragons" territory.

But if emotions are approximations of logical statements, then maybe there's a way to test the waters out there. Maybe one could narrowly apply the linear-thought/math of a potential channel on a case-by-case basis of important considerations, like a safety check. If this test yields good results, the logic could be applied more globally to create the corresponding emotion/mental state.

I agree that the resolution to choice making is in terms of emotional states as they are compared and therefore do realize as a logical statement but algorithmically they are dot products where each emotion is treated as a vector component of a  set that defines a mental state and the winning choice has the highest value comparatively. As the AGI evaluates its environmental condition inclusive of its internal states the computation is continually changing, where even last-minute shifts or corrections to procedural actions or even modifications of data are allowed. This also has to be recorded as episodic events, even the last-minute changes need to be recorded as episodic events. This approach will also suffer from confabulation as recalled memories get modified on the fly as mental states re-interpret the data. As you said" "there be dragons" where we really can't predict with any precision what the AGI will decide but have an idea as to how it will perceive its environment. 

[MikeB]

The abstract states:

Here, we have explored word usage in 6850 free-form testimonials about 27 drugs through the prism of 40 neurotransmitter receptor subtypes, which were then mapped to three-dimensional coordinates in the brain via their gene transcription levels from invasive tissue probes. Despite high interindividual variability, our pattern-learning approach delineated how drug-induced changes of conscious awareness are linked to cortex-wide anatomical distributions of receptor density proxies.

By mapping those NLP symbols to cortical areas of the brain and their neural transmitters we get an idea of the signalling representation of those symbols. To me, at least, this helps me represent symbols as motivational components that can be processed by algorithms to both understand the symbol's use and meaning, as opposed to what ANNs like GPT-3 build statistical maps as to what humans usually respond to keywords in patterns of sentences.

If you give a variety of office workers alcohol, measuring general brain activity before and after, you'll see people being reserved at first, then uninhibited in their own distinct personality later. Some people being light touch, some with burning optimism, some emotionally intuitive, some electric, some logical... etc. Brain activity will show increased receptor densities in a variety of areas depending on the person. Language used in written testimonies will be generally "spirited". So I'm not sure what you can derive from that... except creating a new category of things to predict from.