Language, Analysis, Society, Compassion: AI, Biology, & Types of Intelligence
How can and should we use machine learning and AI technology to better understand life and processes on planet earth?
A Language Question
LLM-GPT’s remarkable similarity to H.Sapiens-humans strongly suggests the Raniwala Firth-Intelligence hypothesis that something about intelligence (if not all of intelligence itself) exists in language-system’s flowing and transmission between people. Generative GPT-LLM AI is better at solving problems than other non-language based AI before it, which also tells us something (if not entirely clear yet) about the nature of language, intelligence, and problems. And GPT-LLM is capable of object-handling which is hugely significant.
But what about animals who are good at solving puzzles but show no, or limited, use of language? (if such a case exists)
As a non-exhaustive starting list:
- language use
- - analytical problem solving
- - social skills and society
- - compassion and empathy
What about areas other than analytical problem solving? Analytical problems may be what we can best define, but what about problem-space more generally? (Or even a problem and non-problems event space?)
Across all existing species, how much is there one cluster of related types of intelligences, or clusters that are close but not just one, or scattered groups of types of intelligences?
How would we experimentally disentangle ‘language’ from a study of biological intelligence, given that language is so embedded in the information tools we use?
Are there problem-solving animals who do not have something akin to language comprehension capacity? Given how little we know about brains, is there any way to distinguish between language-type brain processes and other types for the above areas? While problem solving might be a one-off situation in some cases, the set of four may not be separable from language overall.
A. We can say that language is at least one ‘type’ of intelligence.
B. Social interaction (and the idea that some intelligence is ‘social’ by nature) is likewise not separable from interaction-data. In a case where we are looking for intelligence that is separate from language, how much can social interaction signal data be separated from some category of language? How would linguistic social interactions and non-linguistic social interactions be defined independently? Is all language social in some way? Is there language which is not a Raniwala-Firth flow transmission of intelligence across ~individuals? (Language that is written may be possibly not a transmission, but would it not be more of an asynchronous signal transmission or a medium of transmission question?) (Note, language is also embedded with thought and cognition and learning itself, so some ‘transmission’ is actually more along the line of self-transmission (which socially gets confused with ‘talking after thinking’ language signals).
C. Compassion for others likewise is probably not separable from interaction-signals with others.
And while problem solving is ultimately multi-participant problem solving for our purposes…even the most ‘solitary’ problem solving also needs to be translated into social signals (perhaps another way of approaching the ‘externalization’ topics).
Nevertheless, solitary vs. social problem solving may be an interesting area of study.
Farther Afield
Aside from the all too real and tragic problems of H.sapiens-humans having a miserable time staying connected enough to reality to survive let alone be lucid, with ELIZA-effect type fixations wanting to believe all manor of things the conflict with reality (even in terms of basic project management), the world may indeed be stranger than we can suppose and we should seek out empirical feedback to align our thinking with data rather than form ‘liberal’ or ‘conservative’ aesthetics to try to impose on the world.
Farther off the easier examples, what about:
- normal plant intelligence
- CRV
- the Michael Pollan, Dean Radin, Rupert Sheldrake areas
(spell their names wrong)
LLM’s and Biological Signals
Could LLM-GPT be used to study ‘bird song’?
- cell-signaling?
- secondary chemistry signaling in ecosystems? (chemical linguistics)
At first glance it may be a classic case of how making a map-replica of something may not tell you anything more about the original, but perhaps there are some tricks of methods that could somehow in some ways aid analysis.
An Ocean of Signals and Information
Cryptography and Nonhuman signals
What Do Dinosaurs Talk About?
Whether or not it fits some H.Sapiens-human standard for natural-language, there is something going on with bird vocalization and it would be fascinating if some kind of analysis were possible. E.g. given the way LLM’s work, even if we could feed enough bird-talk input it to create a bird-talk passive-reflective gpt-LLM, that alone (like the exact-map problem) would not in an of itself tell us anything we did not already know. However, if this could be combined with more information about bird behavior and social behavior, instead of a sea of encoded signals, we may be able to fit some world-event data to it. Biological information from the birds may be (if infeasible) interesting to have: signs of stress, for instance. At the very least, given how old bird species are, I cannot imagine not at least trying to find out what dinosaurs are talking about, even if like the generation of secondary compounds by plants, does not closely fit the social customs of signals generated by H.Sapiens-humans and the canids they likely were largely influenced by.
