The objective of this project is to develop a proof-of-concept prototype for the defining facets of a method for generating Artificial General Intelligence (AGI) or Strong AI. To establish the claim of such outcome being a simulation of human mental capabilities rather than mere emulation thereof, the prototype includes certain aspects of human intellect that state-of-the-art AI is not capable of acknowledging. Principally, the characteristics that distinguish envisaged variant of synthetic intelligence concern the ‘subjective‘ nature of the ensued output. Corresponding demonstrable features involve ‘comprehension-based reasoning‘ in conjunction with a more easily relatable ‘logic-based reasoning‘ in AI contexts. Further, the generated cumulative synthetic response to given challenge can be seen to combine various elements as sentience, sapience and common sense – each derived from a common basis underlying these cognitive abilities.

Essentially, such a postulated foundation for all human cognition (and in turn of the proposed variant of AI) relies upon a preordained ‘contract’ between the subject and object components engaged in the process of intelligence creation. Factors determining the native notions of comprehension, sentience, and logic etc. originate from such subject-object contract – a conceptualization pitched as Omnijectivity in this study. One of the key goals of the project is to demonstrate the originality, autonomy, and consistency displayed by such AI on par with those exercised by the human mind.

From an implementation standpoint, Omnijectivity gets realized as a function of several unprecedented technological paradigms introduced in this research. Three of the most notable such interdependent paradigms, that are further elaborated in the Technical Plan, include –

1. Omnijective Axiomatic Scaffold: a support structure for the program-native semantic that sustains an evolutionary episteme. Various types and kinds of constituent axioms strut up what is termed a Potential Semantic Schema (think a repository of yet-to-be discovered human knowledge)
2. Omniject Oriented Programming: a developmental context geared at object-oriented programming from program (synthetic agent) perspective rather than that of the programmer, and
3. Omnijective Contract(ual Framework): an abstraction of ‘all’ possible protocols within given composition of an Omniject to enable consistent inter-entity interactions.

Technical Plan >>

Method overview

Success of A3ilabs – Project POC is contingent upon retaining the character of the erected Omniject throughout (and not let it be relegated to a mere Object), despite our inevitable object-oriented outlook. This entails contesting various prevailing notions related to intelligence and de facto operational patterns in synthesizing it. These include the belief in the intelligence exclusively residing within object representations, rule-based depiction of knowledge denoted by such latent intelligence, algorithmic approach to assimilate so modeled knowledge, statistical character of the ensued intelligence, and more. At the very core, envisaged shift to the existing AI paradigm is not merely architectural or procedural, rather involves a fundamental switch from the deterministic intent implied of object-orientation – absolute or stochastic irrespective, to an enabler intent signified by Omniject-Orientation.

Principles governing Omnijectivity have been formalized following one-of-a-kind fundamental research methodology¹. These Principles of Omnijectivity relate to the compositional and innate complementarity aspects in the creationist pursuit of an Omniject. Also, the principles form basis for Omniject-Orientation (essence of the proposed method) as the generic characterization to the verification and falsification frameworks for any science or technology cascading from such a philosophical positioning. Essentially, the methodology concerns modeling an associated enabler intent rather than given phenomenon itself. The prototype proposed as part of A3ilabs – Project POC primarily demonstrates the applied technological (programmatic) implications of the principles of Omnijectivity².

Technical approach overview

Omniject Oriented Programming paradigm lets the algorithms be segmented semantically as micro-algorithms – that could breach the atomicity of fundamental programmatic constructs to desired granularities, albeit within specified contexts of scoped semantics. Most significant use case of micro-algorithms within an operational setup for general AI involves internalizing patterns of reasoning to programs as heuristics, instead of as programmatic instructions from the outside. Such heuristics combine as per the ask of the problem at hand to output an algorithm denoting the response of the synthetic intelligence indexed by the program. If we were to associate intelligence more with the cause than the effect of the involved dynamics in responding to given challenge – that is, as a basis for generation of the algorithm rather than the resultant output – micro-algorithms signify a step in the right direction.

The enablement implied of such heuristic-based architectures get furthered, to make way for higher degrees of agent autonomy (referred to as human-like intelligence in this study), within the contexts of Omnijective Contract and Potential Semantic Schemata. Essentially, while the portfolios of heuristics denote the building blocks of reasoning imparted to the synthetic agents, the Omnijective Contract abstracts the programmer’s intent³ with regard to the way these blocks are consumed in modeling a solution to given problem. So isolated intent from the program in its entirety allows for defining dimensions of freedom (of intelligence) along which the intent may be negotiated with in the process of assembling an algorithmic response. The notional neighborhoods of the primary intent within which it could be maneuvered, without veering off the consistency required of the resultant response as judged by the human programmer, represent ambits of original intelligence that the program is enabled to reveal – an unintended and unprogrammed for intelligence potentiated by omniject-orientation. In fact, this potential intelligence is the purpose of all Omniject Oriented Programming, where the onus of providing solution to a problem gets reposed with the program while the focus of the programmer is confined to program enablement. Such program-native intelligence gets eventualized vis-à-vis a foundationalist schema of semantics instituted upon Omnijective Axiomatic Scaffold, that condenses all knowledge representable within the implied system in an ensemble of potentialities. Original knowledge gets created within such a schema by way of stitching together patterns of knowledge as driven by the native intelligence, turning potentialities to eventualities. Implementations of such Potential Semantic Schema involve deriving a Semantic Alphabet, as common denominatorial patterns of the intended knowledge representations, used to spell out various manifestations of free willed intelligence such as articulation, innovation, strategy, discovery, sense of humor etc. Inward Extensibility of the semantic schema, the ability of representable knowledge to potentiate inwards indefinitely within set limits, gets realized in a language analogy sustained by the semantic alphabet. (Below Infographic depicts the technical approach)

However, enduring object orientation in terms of the algorithmic basis for various involved operations of enablement such as intent abstraction, intent negotiation, modeling varied logical agencies to carry out these tasks as proxies to the programmer etc., constrains the autonomy required in endowing human-level intelligence to the programs. Whereas its neither possible nor necessary to rid of all object orientation from the process in principle, a right balance needs to be struck between the algorithmic and heuristic facets of the program in enhancing the program-native intelligence originating out of the intent neighborhoods on par with the human equivalent. Several multi-disciplinary fields of research open up in attaining the states of such internal semantic equilibrium to afford human-level autonomy to the machine-mind. For example, while the design of an empiric determinable (say, material object) that appeals consistently to both sentient and comprehensional aspects of the intelligence in imparting physicality to the synthetic agencies⁴ invokes Omniject-Oriented Physics, a more abstract endeavor of endowing ‘choice as an axiom’ to the determiner component of the program (subject) engages in rather fundamental formulation of Omniject-Oriented Mathematics. Although the human-level intelligence is not in the stated scope of A3ilabs – Project POC, some instructive elements such as experiential sentience and multi-agent interpretations get included in the method illustration meant essentially for human-like intelligence.

Footnotes

1. Acontextual Analysis Methodology (or Acontextuality), as it is termed, also happens to be the postulated working principle inspiring human-level cognition. While achieving synthetic Acontextualization is the overall purpose of A3ilabs, Project POC primarily concerns a prerequisite thereof – termed Adjectivization.

2. Given acutely abstract and counter-intuitive nature of the Principles of Omnijectivity, the idea is to publish the applied results before the fundamental results – the impetus behind A3ilabs – Project POC.

3. To be precise, Omnijective contract isolates the programmer’s intentionality that packs more than mere intent to include varied elements of the agent’s purpose, motivation, and capabilities etc.

4. Refers to the possibility of a humanoid robot, that could walk among us sharing our Omnijective axiomatic framework – capabilities & vulnerabilities alike and be able to assert its free will as a ‘choice‘ rather than an external instruction.

Philosophy >>

A philosophical overview: GENERAL

The fundamental aspect of the research devises an alternate hypothesis to biological naturalism in addressing the apprehensions roused by the Chinese room argument. In other words, the version of functionalist theory of mind advanced in this research essentially allows for the syntactical constructs of the programs to refer program specific semantic contexts. For such a premise to hold, however, the basis for the injection of semantics to the programs may not in itself be falsified by the aforementioned argument. This debilitating constraint practically takes every conceivable tool and technique off the table – with prevailing paradigms of programming, mathematics, sciences, language, and logic providing only marginal utility. The study takes the ensuing inquiry to the very philosophical foundations of human sciences, raising questions that could not have been asked earlier, in crafting an operational footing for the method. Of particular note is the inclusive nature of the resultant methodology that offers an encompassing, rather than an adversarial augmentation to human scientific endeavor. Consequently, the fundamental levels of human cognition at which the inquiry plays out essentially extend the base of attainable episteme through human sciences. As could be expected, such a revisionist exploration of the fundamentals is likely to impact multiple disciplines and be influential in spinning off various exotic fields of elementary study. For instance, the study lays out the contexts and basis for the formulation of a theoretical branch of mathematics to explore the other side of the axiomatic frameworks such as first order logic symbolisms or Zermelo-Fraenkel axioms. Below schematic provides the high-level view of proposed theory of mind – the Omniject-Orientation.

Generally speaking, Omniject-Orientation provides a broader catchment for the human phenomena than allowed by the human innate natural attitude characterized by object-oriented outlook. Traditionally, the variance has been dealt with in the context of various domains of Philosophy, though the method never acquired the operational rigor required and remained ad hoc. The formal operational footing to the method as prescribed by the Omnijectivity principles, thus, open up tremendous possibilities in terms of newer technological and scientific paradigms that could not be conceived, much less achieved in our purely object centric outlook. For instance, the micro-algorithms may be viewed as being instrumental in internalizing the semantics of How and Why to the human-coded (agent) programs that have traditionally been centered on the exposition of What, of course within set semantic boundaries.

A philosophical overview: PROJECT POC

In the context of a general variant for programmatic intelligence, the implications of Omniject-orientation are of course going to be felt primarily in the process of programming. A snapshot of the fundamental difference between prevailing and envisioned approaches for AI is summarized below –

Conceptual representation of the intended approach is depicted here –

Such an approach for synthetic intelligence, that goes against the grain on almost every aspect of the prevailing notions related to AI, revolutionizes the way intelligence gets created, manifested, trained, tested, and consumed.

The Prototype >>

Context

Proposed prototype concerns the most sophisticated variety of the Natural Language Programming paradigm yet – natural language itself as an executable program. Ontology-assisted way to programming in conjunction with the epistemic reinforcement offered by Omniject-Orientation, as regards the patterns of programmatic reasoning that integrate deductive, inductive, and abductive styles of logic, provides a conducive environment required of such an ambitious endeavor. Further, several features included in the prototype that demonstrate human-like comprehension, sentience, sapience, and common sense in Natural Language Processing (NLP) contexts sum up the contemplated paradigm.

The prototype takes a depth-first approach in demonstrating certain novel traits of artificial intelligence in NLP domain targeting a small representative set of natural language symbolic constructs (words and phrases). A breadthwise scale up of the prototype will result in programs that have minds in exactly the same sense human beings have minds, in terms of a semantic response to information in NL format (a human-like Semantic NLP Paraphraser/Summarizer). This will transform all aspects of information generation, processing, and consumption over the internet and elsewhere.

Description of what is included

The prototype involves generation of semi-autonomous and semi-original¹ intelligence that is demonstrably consistent with our own, in paraphrasing a NL sentence. The Omnijective contexts within which the demonstration gets setup allows for a rather simplistic enactment of program-level Occam’s razor in the synthetic agents interpreting the semantic contents of the sentence. Moreover, subjective aspects of intelligence get emphasized with three distinct synthetic agents – a philosopher, a romanticist, and a neophyte – each evolving respective algorithm in a bid to semantically comprehend the input NL object in accordance with implied sensibilities (An Example provided below). The prototype takes advantage of such multi-agent scenario to demonstrate the following notions²:

• Thought Primacy: The idea that a foundational thought standing upon Omnijective axioms assumes preeminence over incidental intelligence and other related notions of comprehension, sentience, sapience, and common sense. Essentially ‘thought’ is regarded fundamental³, with varied particulars of the Omniject-Oriented method involved in simulating human thought of which intelligence and other cognitive capabilities are mere derivatives.
• Algorithmic Fluidity: Refers to the ability of evolved algorithms to take on other consistent forms that signify alternate valid subjective interpretations upon finite persuasion. This provides basis for enabling several seemingly disparate human-like capabilities of planning (& changing plans as deemed fit), decision making, certain categories of problem solving, and more.
• Common way of Sensing (among the synthetic & human agencies): Not only do the synthetic agencies that subscribe to given Omnijective contract share a common basis for sensing, but such ‘common sense’ could transcend to the human world given specified nature of composition of the underlying Omniject. This, along with the below feature (Native, Non-trivial Semantic), helps in terms of general learning capabilities of the machines, particularly by enhancing human-machine interface.
• Native, Non-trivial Semantic⁴: While statistics provide an indispensable tool kit to deterministic pursuits, Omniject-Orientation takes an exception to the prudence of ‘playing dice’ in architecting a self-sustained program capable of original intelligence⁵. This transforms current machine learning (ML) paradigm to make way for more aligned patterns with human learning process that involves one-shot or zero-shot learning.
• Experiential Sentience: Deviant from the prevailing practice of keyword-based sentiment analysis, an organic semantic-predicate-based-sentience gets demonstrated in synthesizing complex emotions such as empathy. This detaches the source of the synthetic agent’s emotions from the Omniject-external influences such as declarative or programmatic instructions from the human programmer. Emotion, like comprehension, is a purely native notion in the worldview offered by Omniject-Orientation & gets synthesized dynamically from fundamental Omnijective building blocks.
• Retro/Antero fitting of Predicates (of understanding): A relatable human experience of antero-fitting-of-predicates in assigning inflections to base form of the verbs to comprehend word meanings in isolation is illustrated in NLP contexts. Further, the prototype showcases the ability to comprehend grammatically ill formed sentences using retro-fitting-of-predicates. While the sentence correction is in itself not a novelty, comprehension/sentience driven object transformations not backed by statistical learning is. Such an enablement renders the synthetic agents the ability to handle uncertainties and of subsequent resolution in unchartered/unfamiliar contexts.
• Various subtle theoretical nuances that go into generation of synthetic thought and extraction of a consistent response combining all endowed cognitive faculties from so synthesized thought gets illustrated in the prototype. The emphasis of the project is not just on underlining the gap between prevailing notions of AI and the proposed human-like variant of it, but also in providing a roadmap for further pursuit towards a fully free willed, introspective, and ethical (benevolent) human-level intelligence.

Of particular note are the cognitive abilities displayed by the synthetic agencies, bi-directional between the reality (i.e., the input sentence) and the machine-mind, as a function of an a priori theorization of human cognition abstracted by the Omnijective contract, rather than being derived from explicit instruction, statistical learning, or trial and error basis.

An Example

In the holistic schema advanced by Omnijectivity, intelligence is an inseparable aggregation of all cognitive faculties endowed to a subject agency. Such an intelligence gets illustrated as synthesized fluid algorithms by distinct agents who ‘understand’ each other. The mutual understanding gets established by the way agents determine the transformation path required of respective algorithms to morph to the interpretation of the others.

In case of the example below, for the input NL object:

Reading about weather in books is one thing, but living through a natural disaster was another.

the prototype generates three algorithmic interpretations, one each for the philosopher, the romanticist and the neophyte as below:

Philosopher: Experience triumphs knowledge.

Romanticist: Experiencing a natural disaster could be scary.

Neophyte: Knowing about natural disasters is not the same as experiencing them.

The agents understand 3^rd point perspectives arising from their peers in addition to appreciating how and why the different interpretations are valid. Tendency of the philosopher to navigate the Omnijective semantic towards relevant fundamentals, predisposition of the romanticist to dwell upon the experiential aspects of the native phenomenon, and naivety of the neophyte in a relatively simplistic comprehension of the object reality – while evident for an external human observer are not lost to the machine mind too.

It might be of interest to note that all the human-like intelligence traits listed above find space in generating synthetic intelligence associated with such simple looking NL object (and a handful of its variations).

Footnotes

1. Semi-originality and semi-autonomy referred to connote the distinction between human-like and human-level intelligences in Omnijective contexts. The semi qualifier implies Adjectivized but not Acontextualized intelligence.
2. Some of these demonstrables viz. Thought Primacy, Algorithmic Fluidity, Native Non-trivial Semantic, and Experiential Sentience provide 4 out of a total of 9 criterion for human-level intelligence, advanced in corresponding fundamental aspect of (as yet unpublished) research. Other criteria relate to the Acontextualization capabilities endowed to the synthetic agents, that are not in the scope for Project-POC. In the eventual picture, the collection of criteria is the proposed measurable counterpart to Turing Test.
3. This demands different methodologies for testing. Newer paradigms of Consistency over Correctness emerge where the process itself gets tested for its consistency with redefined notions of correctness that are applicable to original and unpredictable outcomes.
4. Non-triviality essentially places an emphasis on the Nativity of the semantic in signifying ‘not sourced out of the programmer’s mind’.
5. In addition to statistically extracted object patterns through ML, other staple seed of intelligence for current day cognitive architectures such as rules and meta-rules, semantic nets spun over corpus of facts, domain ontologies etc., are also inadmissible in representing semantics within an Omniject.

Summary >>

Impact summary

In applied contexts of general AI, the contemplated paradigms make a case for pioneering technological stacks that support unconventional ways of programming, knowledge modeling, machine training, and testing etc. In fact, the very understanding of intelligence acquires a more pragmatic operational definition, with much sought-after generality of AI being inseparable from the notion of intelligence itself. Such a fundamentally disruptive worldview paves a way that extends well beyond the objectives of the third wave of AI, towards an absolute and ethical synthetic autonomy.

More generally, given the fundamental aspects of the research entail probing unassailable confines intrinsic of human scientific endeavor in fashioning the proposed method, the implications are far-reaching and multi-disciplinary. The impact is particularly accentuated in case of elementary fields of study like physics and mathematics. Also, the qualitatively distinctive complexities involved provide an opportunity, both to academia and industry alike, in envisaging possibilities which may not be conceived in our adherence to the status quo.

To sum up, this study lays out the foundations to explore the road not taken by the human sciences, the path that could not have been considered earlier in our inescapable allegiance to phenomenological natural attitude. The wiggle-room offered by Omniject-Orientation to such constraining norm promises to hold the key not only to a host of imminent technological challenges necessitated by a drastic transformation of human life-styles due to internet explosion in recent decades, but for several persistent problems faced by human sciences.

Project summary

The project demonstrates a method that complements the human sensory analogue of object pattern extraction, the core competency of present AI/AGI architectures, by offering the cerebral equivalent of assigning and extending consistent native semantics to so identified patterns. Such subject-side-tilt of the proposed method provides an integrated basis for machine comprehension, intelligence, sentience, and common sense, while relatively extraneous cognitive conceptions to the method such as multiple perception modalities, memory management, attention mechanisms etc. were scoped out of this project.

Though the proposed model for general intelligence is domain neutral, linguistics provides a convenient context for illustration purpose owing to its relatability with the notion of semantics and other reasons. Consequently, Pragmatic Analysis in the domain of NLP was chosen to demonstrate the method in this project. Multi-agent scenarios were conceptualized to emphasize the characterization of the envisaged variant of AI. Essentially, such synthetic intellect could break through the reserve of current day AI in terms of application, to encompass the realms of ‘the abstract‘.