Page 7 - Machine Learning for Subsurface Characterization

Preface

In Gartner’s list of top 10 strategic technology trends for 2019, “artificial intel-
ligence (AI)–driven development” is placed at the very top. In MIT Technology
Review’s 10 breakthrough technologies for 2019, curated by Bill Gates,
AI-driven automation and AI assistants are mentioned as revolutionary
innovations. Terms like artificial intelligence (AI), machine learning (ML),
deep learning (DL), and big data have been used interchangeably—they are
related but not the same.
Artificial intelligence (AI) is a branch of computer science focused on devel-
oping algorithms inspired by certain aspects of natural intelligence to perform
tasks requiring human intelligence, such as visual perception, speech recogni-
tion, problem-solving, and reasoning [1]. AI is the grand goal, and machine
learning and deep learning are some of the many techniques to achieve AI.
Machine learning (ML) is a subset of methods to achieve AI, wherein the
focus is to develop algorithms that learns from large datasets, also referred
as the “big data.” ML algorithms build a data-driven model based on input data
(combination of features and targets) with an objective of using the data-driven
model to make predictions, recommendations, decisions, and various other
intelligence-related tasks, without needing any explicitly programmed instruc-
tions/rules. Deep learning (DL) is a subset of machine learning methods that
processes “big data” to construct numerous layers of abstraction that builds
functional mapping of the features/attributes to the targets. The feature-target
mappings learned by the DL algorithms can be used to make predictions, rec-
ommendations, decisions, and various other intelligence-related tasks. Machine
learning (including deep learning methods) builds data-driven models that
improve over time as the model is fed more and more data. Big data is an impor-
tant component of machine learning and deep learning. Big data is defined as
extremely large datasets that cannot be analyzed, searched, or interpreted using
traditional data mining/processing methods.
In engineering and most other domains, when people say “AI,” they really
mean machine learning (includes deep learning) [2]. ML works by recognizing
patterns using complex mathematical/statistical techniques and algorithms and
many a times brute-force computing. Deep learning is subtly different from
simple/traditional machine learning. DL methods do not require the manual step
of extracting/engineering features to accomplish the learning task. DL instead
requires us to feed large amounts of data to get reliable results. In addition, DL

xix