Here is how I would state the Question in a bit-longer form:
Isn't a major central question of AI (and OF psychology)(and related to biology principles as well): How can "something" be (in much of its nature) "bottom-up" _AND_ (at the same time) a start of a new "top-down" structure/ability?
My "Answer"/notes on this: You cannot guess all the things "bottom stuff" might be useful for or part of; you cannot suppose to well-guess the nature and limits (or even the full set of possible parameters) of the things at the "top" in your "top-down" thinking: THUS, your models are bad both ways. How about finding that which IS "found" by some of the "bottom stuff", using the bottom stuff, but which is also the start of a new great 'top-down'-type of functioning (e.g. a new way to classify, classifying NEW sets of aspects of the environment/environmental change), using some bottom-up as components, helpers and pieces to be found anew, as used in new ways, in the new developing [begun-to-be-'known'] context? Some such thing could be "seen" (in a sense)(BY the individual organism, itself) before its nature is fully formed or/and before it is really functional (in other words: before it is useful/well-used, and LIKELY even before becoming ANY center of attention, i.e. before clearly conscious **) -- and given how it develops and what is involved in its development (plus how variable and open its development must be), this is the way it WOULD BE. [(Doesn't this provide both the kind of openness and adapted-ness required (biologically) -- and required in real AI?)]
This is the way I see that which is SOUGHT, and first sought-to-be well-hypothesized (and then found) AS indicated (just indirectly, which is now the only way it can be) by its special-typical "products"/results/consequences described in "A Human Ethogram ...". Yet the "products", etc., should be guiding one toward hypotheses of new actual direct-observables (being seen, very likely requiring the use of new technology, eye-tracking and computer-assisted analysis ***), though also requiring imagination and one well-learned, with good knowledge of earlier similarly-qualitative new learnings and developments in ontogeny. In "A Human Ethogram ..." the NEW behavioral "ingredients" in these key learnings changes are simply perceptual "shifts" (or perceptual/attentional "shifts") -- which would suffice.
SEE: Article A Human Ethogram: Its Scientific Acceptability and Importanc...
AND https://www.researchgate.net/project/Developing-a-Usable-Empirically-Based-Outline-of-Human-Behavior-for-FULL-Artificial-Intelligence-and-for-Psychology
** FOOTNOTE: It would be this way if you need all the "pieces" (AND perhaps put together anew) before the top-down thing is literally notable.
*** Because they would OTHERWISE not be seen (ARE not seen) by us or the organism (clearly, consciously).
P.S. This is an empirical foundation for cognitive-development studies and theories (some foundation is presently missing from the psychology of cognitive development -- HERE IT IS !)
Another P.S. It may be too damned bad, but surely may be true that: some more discoveries in behavioral (biological) science ARE NEEDED before true AI can be done (it is not like that field has not have time to try with the knowledge we have or the models they generate from their own minds).
P.S. What, better than what's described above "to build an integrated system that is self-consistent" (" it is crucial to build the system around a general and flexible core") (Wang 2004 and Pei Wang, Ben Goertzel 2012])? And, how else are you going to get "substructure upon substructure of causal information connections, which in turn form architectures within architectures within architectures, nested at numerous levels of granularity, each having a complex relationship with the others both within and between layers of granularity"? Only "thing" other than via something like I propose, would be : BY MIRACLE. It is really time we "get real" ! (And as Thorisson has said: "the fundamental principles of an AGI must be addressed holistically" -- this is another characteristic my perspective on the human [developing] conforms with. )
And, by the way, if psychologists cannot or do not do the foundational work you need: It is not necessarily terribly or especially difficult to find and work with someone who can. To get an idea of what's needed, see:
https://www.researchgate.net/post/Are_Embodiment_Theories_Enactment_Theories_1_PIECE_OF_RESEARCH_away_from_becoming_trivialized_we_all_no_longer_doomed_to_poor_evidence?
P.P.S. This also approach/research I recommend (quite interestingly) is a way to "see beyond ourselves" (what we can usually see) -- isn't that what we would like to do? Even if psychologists cannot see this, AI people may see that doing such things could get a long way fast.
I agree with the following quote byThorisson (but do not find using the word "complicated" helpful [(let's be optimistic !)]): "architecture must have built-in ways to compare its own status between days, [etc].... might involve pattern matching of large parts of the realtime mind, that is, the part of the mind that controls the creature from moment to moment at different points in time. For a large, heterogeneous architecture such architecture-scale pattern matching can get quite complicated. But it is unlikely that we will ever build highly intelligent artificial systems without it." Look for congruity or consistency with what I say must be "looked-for" and discovered: can you see a similar role "in the system" -- OR AT LEAST A KEY PART IF IT (for real, really). That context in which the new emerges and the new becomes established may surely be complex, though [(YET not a "complex" you imagined OR imagine (as a whole anyway), but a complex you will have evidence of from substantial longitudinal developmental research (if that's done))].
My perspective is also in-line with Thorisson's proposal for good AI architectures : "same small set of basic principles can be used throughout to construct every function of a cognitive system".
Dear Kristinn R. Thórisson and his followers:
Want to see the SPECIFICS: See my Comment under:
Chapter A New Constructivist AI: From Manual Methods to Self-Constru...
WAIT : Let me quote what is said in that Comment:
I can tell you that getting an answer to how to achieve 4 out of your 5 or 6 new characteristics needed by your proposed constructivist approach (and by AGI) is NOT difficult. It just requires that you are able to believe that developmental psychology still needs some foundational work (foundation work which finally provides SPECIFICATION of Piaget's EQUILIBRATION type 2 **, which he only said is "due to maturation"). And:
Once you can acknowledge this (above), can you accept an answer that puts behavioral patterns (new to be discovered, with new technology) in essentially a biological-type perspective? [(Of course you can !!)]
HERE, Professor Thorisson, ARE YOUR "SEEDS" !! -- ALL HERE IN THE MAIN QUESTION ABOVE (to which this is an "Answer") AND THE LINKED-TO PAPER.)
** FOOTNOTE: Also see the following Q and A, to avoid going a DISTINCTLY INCORRECT DIRECTION:
https://www.researchgate.net/post/Do_you_know_why_Piaget_himself_would_NOT_approve_of_Sensorimotor_Contingency_Theories?
and see my other Questions (with Answers) about "Embodiment" 'Theory' -- something basically in the PLACE OF filling out Piaget's view; and its peer-reviewed in published work as hopelessly ill-founded and offering little (or nothing) (yet this is where some cognitive psychologists WOULD direct you; AND, I have already seen the idea of "embodied" thought in some major artificial intelligence essays, so this advice I am giving you comes none too soon).
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence