End of the road for Structural AI and beginnings of Functional AI
We can classify objects and ideas in different ways, any new way lends an insight as how things are in reality? and how reality can be carved into new objects and ideas? In this effort, I have opted to classify Artificial Intelligence (AI) in two classes namely: structural and functional AIs.
Structural class is prevalent in present-day and a success story, and it is epitomized by Deep Learning, Generative pre-trainers, Large Language Models and their various variants such as GPTs, Bards, CNN, DNN and others. Structural AI at the subliminal level is representation of a behavior by optimizing structure over a set of data points. Repeated iterations of structure on data points results in the best comprise of required behavior. This interplay of data and structural parameters, astoundingly springs before us an almost perfect representation of human needs.
The second class is a novel AI termed as Functional AI, it has data but it is a contextual and relevant behavior building approach to intelligence, and it is inspired from evolution of life, and just like life evolved from simple to complex forms, simultaneously intelligences evolved, and functional AI is envisaged to represent consciousness and awareness. This approach is a leap over structural AI, and it can be seen as a new AI paradigm. To understand functional AI , it is necessary to carves out a set of abstractions that clearly highlight the limitations of the present-day AI approaches and sets the foundations of functional AI. These abstractions of Functional AI are derived from my earlier work [1] “Consciousness: a paradigm towards artificial general intelligence. This work sets the theoretical basis of the proposed Functional AI, which clearly has potential to surpass present state of the art. Below is a list of 18 comparisons of the proposed Functional AI against the present-day Structural AI.
[1] Chaudhry, Adnan Rashid, Consciousness: A Paradigm Towards Artificial General Intelligence (December 8, 2021). Available at SSRN: http://dx.doi.org/10.2139/ssrn.3981058
