WHY WE NEED SMART AI WORLD
We live in a stupidly globalized world where 8 capitalists own the wealth of 3,2 bn people, where fewer than 10% of the world's public companies account for more than 80% of all profits.
This is the key obstacle to new revolutionary ideas, innovative startups, global sustainable developments, human prosperity and progress.
EIS Encyclopedic Intelligent Systems LTD is engaged in creating AI World:
WA Big Confusion from Big Internet Companies: Confusing ML Systems with AI Systems
Artificial intelligence is divided as “narrow AI”, designed to perform specific tasks within a domain, and “general AI”, which can learn and perform tasks anywhere.
Machine learning (ML) as the development of new statistics-based algorithms and models in computer science is referred to “narrow AI”.
ML is to empirically discover models/algorithms/knowledge/insights/rules through learning from historical relationships and trends in the data, specifically, Big Data.
As such, ML involves computational statistics, statistical computing and mathematical optimization, whereas AI draws upon many sciences and technologies: computer science, mathematics, psychology, linguistics, philosophy, neuroscience, physical science, engineering, etc.
AI is about creating intelligent systems [that can know, learn, reason, plan, perceive, process natural language, act], involving machine intelligence, artificial consciousness, and intelligent communities.
ML is just automated feature engineering, feature learning or data representation learning, to automatically discover the representations needed for feature detection or classification from raw data, or real-world data as images, video, and sensor data.
Crucial, modern ML systems are not real AIs for many good reasons:
designed to perform specific tasks within a domain (e.g. language translation);
decisions as not preprogrammed by humans;
decisions as emergent properties of the learning algorithms and the biased data set they are trained on;
multiple hidden layers in an artificial neural network;
performing either supervised learning, reinforcement learning, or unsupervised learning;
the lack of unifying theory surrounding an uncountable number of machine learning algorithms and methods;
a lack of discovery component;
requiring mathematically and computationally convenient input;
a lack of transparency and interpretability in decision-making;
safety and security issues;
considerations regarding accountability;
issues of data quality and potential bias.
ML systems could be trained with data that is biased, inaccurate, incomplete or misleading.