How to build Real AI (RAI): From Deep learning to Deep understanding, from Big data to Smart data
-
24 October 2019 - updated 1 year ago -
0 comments
I share an introduction to the closed corporate report: How to Make Real AI combining World Model- and Data-driven AI: Build General Data Understanding Platform and Feed Smart Data
The cost of building of general deep AI is virtually invaluable: "an investment of $1 billion from Microsoft is to focus on building a platform that OpenAI will use to create new AI technologies and deliver on the promise of artificial general intelligence".
For all who dealing with Data/Analytics/Human Intelligence/Business Intelligence/Machine Learning/Deep Learning/Artificial Intelligence/Superintelligence
http://www.worldxxi.com
Keywords: Data, Big Data, Business Intelligence, Intelligent Data, Machine Learning, Deep learning, Deep understanding, Real Artificial Intelligence, SMART DATA Algorithm
Big Data AI vs. World Model AI
Today, the big-tech-led artificial intelligence is revolving around big data-driven AI, Machine Learning techniques and Deep Learning algorithms.
The reasons are a lot of data aggregated, cheap data storage, fast GPU processors and advancements in statistical learning and computational statistics, neural net algorithms, data analytics and other data-centric techniques.
Deep learning computational models composed of multiple processing layers are to learn representations of data with multiple levels of abstraction. DL discovers patterns and relationships in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep neural convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics.
Model-driven AI cum Data-driven AI
Model-driven AI represents deep understanding of how the world works through causal representations and rules, symbolic models and fundamental laws of nature, as E=mc².
Modeling and simulation (M&S) is the use of models (conceptual, physical, mathematical, or logical representation of system, entity, state, phenomenon, process and relationship) as a basis for computing simulations to solve real-world problems or to develop data for various decision makings.
Conceptual models are largely abstractions of things in the real world. Conceptual modeling is formally describing the world as a whole, its domains, or some aspects of the physical and social world and information world and mental world for deeper understanding, learning, communication and interaction.
In classic AI, conceptual models were used for building knowledge representation and reasoning models, automated reasoning engines, expert systems and knowledge-based systems, known as conceptual graphs, semantic nets, systems architecture, frames, rules, and ontologies or inference engines, theorem provers, and classifiers.
We effectively combine the big data-driven ML/DL with the real world model-driven AI involving computing modeling and simulation to build true AI systems.
Data representation, learning, understanding and reasoning (DRLUR) is combining knowledge representation and reasoning (KR², KR&R) and feature engineering techniques, all to represent data about the world in a form that a computer system can solve any complex problems and special tasks, as processing images, video, speech and audio, diagnosing a medical condition, having a dialog in a natural language, autonomous driving, etc.
As the key DL proponents concluded in 2015 Nature article: “Ultimately, major progress in artificial intelligence will come about through systems that combine representation learning with complex reasoning. Although deep learning and simple reasoning have been used for speech and handwriting recognition for a long time, new paradigms are needed to replace rule-based manipulation of symbolic expressions by operations on large vectors”.
Deep learning; Yann LeCun, Yoshua Bengio & Geoffrey Hinton; doi:10.1038/nature14539; http://pages.cs.wisc.edu/~dyer/cs540/handouts/deep-learning-nature2015.pdf
The AI Paradigm Shift: From Deep learning to Deep understanding, from Big data to Smart data
If Big Data is the driving force of the digital economy, then Smart Data is the engine of the smart economy.
Big data refers to a mass of data fast produced by a high number of diverse sources, as created by people or generated by machines, such as the IoT devices, sensors gathering climate information, satellite imagery, digital pictures and videos, purchase transaction records, GPS signals, etc., covering many sectors, from finance and healthcare to transport and energy.
Generating value at the different stages of the data value chain will be at the centre of the smart AI economy.
https://ec.europa.eu/digital-single-market/en/big-data
To build real AI-based systems, high-quality SMART data is a key factor to secure optimally best cognition, learning, and reasoning, knowledge and understanding, predictions, decision makings, actions and performances.
The idea of Smart Data is making the way to the intelligent information market changing the standard understanding of data and information, as well as data science and technology, data analytics, business intelligence (BI), and AI technology.
BI is "a set of methodologies, processes, architectures, and technologies that transform raw data into meaningful and useful information to enable effective strategic, tactical, and operational insights and decision-making."
SMART BI is "a system of methodologies, processes, architectures, and technologies that transform raw data into smart data to enable intelligent strategic, tactical, and operational insights and decision-making."
Another good reason to transform the Big Data into the Smart Data is predictions. The Big Data predictions are normally failing to be valuable, as Google Flu Trends failed to deliver good predictions, overstating the flu outbreaks, or Twitter failed to predict the results of the 2016 U.S. Presidential Election. Since Big Data very often means dirty data, inaccurate, incomplete or inconsistent data, as in computer systems or databases.
Smart Big Data offers new opportunities for automated decision-making in critical human development areas such as science, technology, education, health care, employment, economic productivity, crime, security, natural disaster, resource management, manufacturing, e-commerce, ICT, media, Internet services, and social networks.
Nobody systematically studies Smart Data.
There are an initiative, the IEEE International Conference on Smart Data (SmartData-2018), which really discussing only the Big Data topics, while recognizing the problem.
“Smart Data aims to filter out the noise and produce the valuable data, which can be effectively used by enterprises and governments for planning, operation, monitoring, control, and intelligent decision making. Although unprecedentedly large amount of sensory data can be collected with the advancement of the Cyber Physical Social (CPS) systems recently, the key is to explore how Big Data can become Smart Data and offer intelligence. Advanced Big Data modeling and analytics are indispensable for discovering the underlying structure from retrieved data in order to acquire Smart Data. The goal ...is …identifying the Computational Intelligence technologies and theories for harvesting Smart Data from Big Data”. http://cse.stfx.ca/~SmartData2018/
Hereby EIS is proposing to the Big Data Market the most original Smart Data innovation when the World Data System is developed as ranging from Categorical Data to Category Data, thus covering and systematizing all the basic kinds of data.
Whether it is geographical information, statistics, weather data, research data, transport data, energy consumption data or health data, there is the need to understand or make sense of ‘big data’ as leading to revolutionary innovations in AI technology as the True, Actual, Deep, General AI systems.