“You can’t improve what you don’t measure.”- Peter Drucker
PROLOGUE
Douglas Adams in the Hitchhiker’s Guide to the Galaxy states that the answer to life, the universe and everything is 42, but what really is the question? We talk about Big Data and all the analytics that it entails to try to intelligently solve problems, but we always seem to encounter edge cases where the brittleness of AI causes things to go badly wrong.
In The Book of Why, Judea Pearl posits that “You are smarter than your data. Data don’t understand cause and effect, humans do. To build truly intelligent machines, teach them cause and effect, but the AI field got mired in probabilistic associations. Causality is not correlation.” Just knowledge isn’t good enough, the wisdom to act upon it with judiciousness in the context of jurisprudence in appropriate jurisdiction is more important. We also do not need to be deluged with data (volume) to draw our inferences since speed (velocity), vitality, veracity, value and validity (aka the five Vs) are essential for a good deduction and actionability.
In this post we summarize our seminal article on the post-hoc fallacy and supplement our prior posts on Everything Connected to emphasize that data is not information is not knowledge is not intelligence is wisdom. We will also stress on why developing contextual awareness is needed so that AI embedded machines can know how to act appropriately leading to what we call Existential Intelligence (EI).
We will talk about Question Focused Datasets (QFDs) based on our Tau Codex Orchestrator – a multi-modal AI engine that will prevent large biases from corrupting our predictions, so that actionable intelligence has the correct basis for doing.
TABLE OF CONTENTS
- The Background
- AI Progression
- Complexity and Phenomenology
- The Reality
- Analyzing Causation
- The Context – Sense & Sensibility
- The Swerve
- Our Perspective on Orchestration
- Multimodal Inference
- The TAU Codex Transformer
- Summary
The world (and this universe) is a Big Data challenge. We need to move beyond data and information to knowledge, intelligence and wisdom so that we do not necessarily have to use large amounts of data, but question focused sets of information that are wisely selected to get to the heart of the matter. There is need to move from today’s brittle AI based on trained and reinforced networks to contextual awareness where intelligence would evolve based on environment, interactions, rationality and pragmatism – philosophical tenets that technology still needs to make sense of where rapid meta-analysis of various valid inputs will inform decision making. Such multi-modal orchestrations will be needed for the correct actionability of intelligence. Our formulations to define the 5th order of Cybernetics will be the basis for developing this construct.
AI has been remarkably successful and outperforms human experts in certain tasks, even in complex domains such as medicine. Humans on the other hand are experts at multi-modal thinking and can embed new inputs almost instantly into a conceptual knowledge space shaped by experience. In many fields the aim is to build systems capable of explaining themselves, engaging in interactive what-if questions. Such questions, called counterfactuals, are becoming important in the rising field of explainable AI (xAI) as was discussed in our prior article.
AI is as useful as it is ubiquitous, but they are only as intelligent, rational, thoughtful, and unbiased as their creators. Safe, ethical, and effective AI will be vital for our future. This is an important issue in AI because systems are increasingly making decisions that have an impact on people’s lives, and we need to understand why they make the decisions they do. We need to understand their ramifications too, in terms of its progression from Analytical, Actionable, Emotional to Existential on how it would interact with humans in terms of cooperation, collaboration, and even competition (coopetition?).
In our thesis on Explainable AI (XAI), we began to address these issues for making it responsible, trusted, and ethical. XAI should have the capacity to evaluate future impact and cascading events that the informed decisions recommended by it would have on individuals, industries, and civilizations as well as on environment. Such Analytics will precede any actionable and sustainable solutions. More details on this in our subsequent article on Existential Intelligence (To Be or Not to Be).
In a post on managing complexity (Of Knowns and Unknowns), we had looked at how knowns and unknowns form the basis for articulation of the duality of intelligence – the thinking and doing – and how a series of multi-modal schemas are needed to make sense of data and information to progress to a meaningful solution. Also, on an article on the post hoc fallacy – After this, therefore because of this – we had discussed phenomenology and causal relations. The premise of that article was:
The models that AI builds must encode the fundamental patterns of experience and they must be rich and deep so that the AI can access and reason over the causes of their inputs. There should be an interactional relationship of harmonized reciprocity between the seeing and the doing, the observation and the observer. For the purpose of the development of artificial intelligence, we must understand how human intelligence can be replicated. Hence our progression into what we call Existential Intelligence.
We need Strong Artificial Intelligence so it can help us understand the nature of the universe to satiate our curiosity, devise cures for diseases to ease our suffering, and expand to other star systems to ensure our survival. To do all this, AI must be able to learn representations of the environment in the form of models that enable it to recognize entities, infer missing information, and predict events. And because the universe is of almost infinite complexity, AI must be able to compose these models dynamically to generate combinatorial representations to match that complexity. Models are more than recipes of what to do, they are encoded pieces of the world and provide scaffolding over which to optimize behavior.
Current AI systems crash when faced with novel data. They are bad at transferring learning to new tasks. They fail in deployment when the world changes unexpectedly. And they are susceptible to adversarial attacks.
The reality is that today’s AI cannot do the things that are rational and pragmatic. We don’t know how to map the brain, which means we have no way of building a dataset to train AI how to interpret it.
Case and Point: My Tesla’s Full Self Driving capability has issues negotiating left turns and comes to a stop in the middle of the intersection when the signal turns red even if there is enough room to cross the intersection (Rainman anyone?).
As Ian Goodfellow at Apple, Yann LeCun at Facebook, and Jeff Dean at Google can all tell you: if you could solve self-driving cars, the human brain, or AGI with money, it would have already been solved. When DeepMind was founded, its purpose was not to win chess or Go games. Its purpose was to create an Artificial General Intelligence (AGI).
It is the same with GPT-3 and just about any other multi-modal AI system being developed today. We’ve seen the development of multi-billion-dollar neural networks that push the very limits of compute and hardware. And, even with all of that, we could still be decades or more away from self-driving cars or algorithms that can interpret human neuronal activity.
Data is digital but the world is analog
Today’s AI has a mapping problem. Any “map” automatically suffers from severe data loss. On a map, you just see a tiny representation of the immense reality. Maps are useful for directions, but if you’re trying to count the number of trees on your property or determine exactly how many wolverines are hiding in a nearby thicket, they’re pretty useless. When we train a deep learning system to “understand” something, we have to feed it data. And when it comes to massively complex tasks such as driving a car or interpreting brain waves, it’s simply impossible to have all of the data. We just sort of map out a tiny-scale approximation of the problem and hope we can scale the algorithms to task. In short, we mistake cognitive grasp of a map for comprehensive knowledge of the territory. It requires entirely different skill sets to find a way in any territory that has no prepared road maps. Existential Intelligence will need to have the cognitive flexibility of the explorers.
A useful way to think about causation is to realize that all causation is “organic.” The concept of “organic” implies that the nature of the elements, the nature of the relationships among the elements, and the possibilities for reaction in the elements, are all in a flux or a constant state of change. In other words, every cybernetic system tends to maintain a dynamic balance and a readiness to be fluctuating appropriately in response to the feedback loops to stay on target. Such a system will be considered to be in a flux.
According to DeepMind – “Causal models have a unique combination of capabilities could enable a deeper understanding of complex systems and allow us to better align decision systems with society’s values.” Applying the principle of causality and effectively utilizing causal AI for enterprises promises to unlock progress on some of the biggest open problems in AI.
THE CONTEXT – SENSE AND SENSIBILITY
In every ancient philosophy, there is a dictum related to five base elements that need five senses for perception. This was the precursor to experimental and evidence based scientific methods that we follow today.
De rerum natura (On the Nature of Things) is a first-century BC didactic poem by the Roman poet and philosopher Lucretius with the goal of explaining Epicurean philosophy to a Roman audience. The poem explores Epicurean physics through poetic language and metaphors, namely:
- the principles of atomism,
- the nature of the mind and soul,
- explanations of sensation and thought,
- the development of the world and its phenomena, and
- explains a variety of celestial and terrestrial phenomena.
The universe described in the poem operates according to these physical principles, guided by fortuna (“chance”), and not the divine intervention of the traditional Roman deities.
Epicurus was an ardent Empiricist; believing that the senses are the only reliable sources of information about the world. He taught that knowledge is learned through experiences rather than innate and that the acceptance of the fundamental truth of the things a person perceives is essential to a person’s moral and spiritual health. He rejected the Platonic idea of “Reason” as a reliable source of knowledge about the world apart from the senses and was bitterly opposed to the Pyrrhonists and Academic Skeptics, who not only questioned the ability of the senses to provide accurate knowledge about the world, but also whether it is even possible to know anything about the world at all.
Clinamen (derived from clīnāre, to incline) is the Latin name Lucretius gave to the unpredictable swerve of atoms, in order to defend the atomistic doctrine of Epicurus. In modern English it has come more generally to mean an inclination or a bias. In his perspective: “When atoms move straight down through the void by their own weight, they deflect a bit in space at a quite uncertain time and in uncertain places, just enough that you could say that their motion has changed. But if they were not in the habit of swerving, they would all fall straight down through the depths of the void, like drops of rain, and no collision would occur, nor would any blow be produced among the atoms. In that case, nature would never have produced anything.” This was perhaps a precursor to Heisenberg’s Uncertainty Principle, the underpinnings of today’s Quantum Mechanics. Lucretius believed that it was these meanderings (swerving) that created the different states and properties of matter, and the needs for observers (like us) to develop special senses to perceive them.
We believe that both the Epicurean philosophy of sensory perception, and the Euclidean/Archimedean/Platonic view of reason and inference (sensibility) are needed in equal measure to prevent brittleness in future AI and ML constructs.
OUR PERSPECTIVE ON ORCHESTRATION
Over the last decade, the application and performance of Deep Learning has progressed at an astonishing rate. However, the current state of the field is that the neural network architectures are highly specialized to specific domains of application. An important question remains unanswered: Will a convergence between these domains facilitate a unified model capable of performing well across multiple domains? In keeping with the philosophy of Epicurus (sense) and Euclid (reason), we propose an orchestrator that combines sense with sensibility – multi-modal perceptions with an inference engine that we call the Tau Codex Transformer that we will describe in this section.
As we move to the realm of AI, there would be multiple Large Language Model personas combined with science and math toolkits that would be needed to be invoked to provide for innovation compositions.
Innovation unlike invention is a process that needs the collective imagination of creators, the knowledge of subject matter experts and the foresight of visionaries. For businesses, it is also the aspect of commercial viability be it as a replacement to an existing product or solution, or as a disruptor. It is for this reason that large enterprises are wary to undertake drastic changes in their business processes and go-to-market strategies – pivoting a large ship has unforeseen consequences – and it up to the nimble startups to make it or fail.
But what happens when you introduce AI to the equation? At Numorpho Cybernetic Systems (NUMO), we recognize the potential of AI in catalyzing innovation. We are in the process of constituting an Innovation toolkit driven by enhanced generative constructs that would provide for streamlining the workflow, enable collaborations (both within and with partners) and provide for seamless interoperable data integration. Called Manthan (Sanskrit: Churning), it is part of our process automation platform – Mantra M5 – that would provide a multi-modal basis for transformations thus providing a versatile foundation for ideating and iterating on future solutions with a clear foundation and intent.
As we delve into the realm of AI, the need for multiple Large Language Model personas becomes increasingly apparent to foster innovation and creativity in compositions. These personas could be invoked to provide diverse perspectives, creative ideas, and domain-specific expertise, enriching the innovation process.
The TAU is a multi-modal neural network that we are architecting. It draws from the success of vision, language, and audio networks to compose solutions for problems spanning multiple domains, including industrial automation, sustainability, and infrastructure development. It will coordinate the convergence of multiple different sensory perceptions and appropriate data based on different modalities into a single composition.
TAU will consist of
- encoder,
- aggregator and
- decoder
that interact with a computational cybernetic fabric composed of a multi-modal database. It will utilize Microsoft platform and components – namely the Azure Cloud provisioning, Dynamics 365, and the Power Platform, and OpenAI and the Nuance Communication stack to synthesize the solution.
It will also include Engineering and other domain specific plug-ins utilizing the Nvidia Omniverse platform for computer aided simulation, additive manufacturing, GPS and mapping, geo-location, and findability to fulfill different types of needs that would arise in the process of solutioning. AR/VR/XR will be included to provide for the future of the Multiverse, and the progression IoT via Digital Threads and Digital Twins to visualize and interact with the virtual world in conjunction with our ideation curriculum (Manthan), the Digital Twine reference architecture and the integrations protocol (Tendril Connector). Emerging technologies like Distributed Ledger, Genetic Programming and Quantum Computing will also be included in a themed progression in the future.
A high-level interaction architecture diagram of the TAU is illustrated below:
The total volume of data is vast and increasing exponentially. However, many datasets are correlated, expressing similar facts about the world. Businesses increasingly place a premium on identifying the right data to answer their questions and achieve their goals. To do so they need technology that can penetrate beneath correlations to identify distinctive marginal value in data – hence our foray into what we call Question Focused Datasets.
There is also a growing recognition that the explainability, safety and fairness of AI systems are critical in business settings. The wide-ranging incoming legislation mandates that businesses should provide explainability reports, apply stress tests and ensure humans remain “in the loop”. This perfectly matches with the needs for evolving Industry and Services 5.0 to have a human-centric, sustainable, and resilient solutioning process that is driven by empathy and human goodness to move the focus from shareholders to stakeholders. Empathy entails mentalization. Mentalization is the empathic construction of a mental model appreciating the state of discomfort in the consumer who is seeking solutions to minimize the discomfort. Appropriate solutions and their applications have the potential to mitigate the human discomfort as well as catapult the solutions into future, actionable and sustainable systems.
Orchestrating Intelligence for us at NUMO, is not just about aggregating intelligence information feeds but also about creating knowledge structures for business decision-makers to visualize and formulate queries interactively and contextually. The technology is designed to address problems arising from the new digital business environment. We have integrated with a set of technologies that help organizations interactively explore and analyze the data in the question-focused datasets. This enables business decision-makers to generate new insights and drive decision-making.
Our partnerships and using tools like Microsoft’s Power Platform and Nuance Communications, OpenAI’s GPT-3 based NLP engines and Nvidia’s Omniverse platform will be the basis of this orchestration that will be unparalleled in its ability to deliver actionable intelligence to our solutions.
In this AI-driven era, the possibilities for innovation are boundless, and with the right tools and approaches, businesses (both large and small) can navigate the evolving landscape to harness its potential for growth and success. We at NUMO are firmly committed to this endeavor. We aim to be a catalyst for change, working closely with partners, collaborators, and visionaries to shape the future of industries.
NITIN UCHIL Founder, CEO & Technical Evangelist
Summarization by ChatGPT
The whitepaper titled “Orchestrating Intelligence: PROLOGUE” discusses several key concepts related to artificial intelligence, causality, and the development of advanced AI systems. Here is a summary of the main points in the prologue:
- The Question of Meaning: The paper begins by referencing Douglas Adams’ question in “The Hitchhiker’s Guide to the Galaxy” about the answer to life being 42 and the importance of finding the right questions in the context of Big Data analytics.
- Causality vs. Correlation: The paper highlights Judea Pearl’s assertion that true intelligence requires an understanding of causality rather than merely identifying correlations in data. It emphasizes that knowledge alone is insufficient; wisdom is essential for informed decision-making.
- The Five Vs of Data: The paper introduces the concept of the “five Vs” (volume, velocity, vitality, veracity, and value) as essential considerations for deriving meaningful insights from data.
- Existential Intelligence (EI): The paper emphasizes the need for AI systems to develop contextual awareness to act appropriately, leading to what it terms “Existential Intelligence (EI).”
- Question-Focused Datasets (QFDs): The paper discusses the importance of using Question-Focused Datasets (QFDs) and mentions the Tau Codex Orchestrator, a multi-modal AI engine designed to prevent biases and ensure the correctness of actionable intelligence.
- The Background: The authors argue that the world is a Big Data challenge and that we should move beyond data and information to knowledge, intelligence, and wisdom. They stress the importance of context-aware AI systems.
- AI Progression: The paper acknowledges the success of AI but notes that AI systems can still be brittle and need to progress toward becoming rational, ethical, and effective. It introduces the concept of “Existential Intelligence.”
- Complexity and Phenomenology: It discusses the duality of intelligence (thinking and doing) and the need for multi-modal schemas to make sense of data and information.
- Analyzing Causation: The paper advocates for a deeper understanding of causation in AI, stating that causal models can lead to a better understanding of complex systems.
- The Context – Sense & Sensibility: The authors mention ancient philosophy’s emphasis on five base elements and five senses for perception, emphasizing sensory perception and reason as complementary.
- The Swerve: The paper references “clinamen,” the unpredictable swerve of atoms, and suggests that both sensory perception and reason are necessary to prevent brittleness in AI.
- Our Perspective on Orchestration: The authors propose an orchestrator that combines sense with sensibility, integrating multi-modal perceptions with an inference engine called the Tau Codex Transformer.
- Multimodal Inference: The paper emphasizes the need for multiple Large Language Model personas to foster innovation and creativity in AI compositions.
- The TAU Codex Transformer: It introduces the TAU, a multi-modal neural network that combines various sensory perceptions and data into a single composition.
- Summary: The paper concludes by discussing the importance of identifying the right data, the need for explainable, safe, and ethical AI, and the role of empathy and human-centric solutions in the AI-driven era. It mentions partnerships and technologies to deliver actionable intelligence.
In summary, the prologue sets the stage for a discussion on orchestrating intelligence, emphasizing the importance of causality, multi-modal perception, and the need for AI systems to go beyond data and information to achieve wisdom and existential intelligence.