What is Artificial Intelligence?

An
Informed Definition
Last updated on December 21, 2018, published by Daniel Faggella
Daniel Faggella is the founder and CEO at Emerj. Called upon by the United Nations, World Bank,
INTERPOL, and many global enterprises, Daniel is a sought-after expert on the competitive strategy
implications of AI for business and government leaders.
Share to:
LinkedIn Twitter Facebook Email

What is artificial intelligence (AI)? We take the term for granted, but how might
we phrase a formal definition? And are the technologies that we have today
really reflective of all that this term implies?

Traditionally a branch of computer science, AI as a holistic concept has pulled

from many areas of academic arenas, from philosophy to physics. Many are
aware of the recognized origins of the term – famed computer scientist John
McCarthy in 1956 at the “Dartmouth summer research project on Artificial
Intelligence.” Since that time and as research and technology in this area has
evolved, definitions of AI have shifted across a wide spectrum, and
academics, businesspeople, and laypersons have a range of definitions
(some better informed and reasoned than others – though again, the utility of
such a term can depend on background and objectives).

One of the reasons AI is so difficult to define is because we still don’t have a

set definition or one solid concept for intelligence in general. Intelligence is
often dependent on context. A traditionalist might define intelligence as level
of reasoning power, and this seems one of the reasons why a popular
determiner of AI has often been games – man and machine try to ‘outthink’
the other, or in the case of the machine at least match the human, so that it
becomes difficult to tell where man begins and machine ends.

But, in the end, mastering a game (like Go) is very different from sealing a
successful business deal in the real world, then driving home to have a meal
with your family and reading and reflecting on a bit of Plutarch before bed. In
any case, researchers Shane Legg and Marcus Hutter have made the case
that intelligence includes the following features:
 Intelligence is a property of some entity or agent that interacts with some form of
environment

 Intelligence is generally indicative of an entity’s ability to succeed (by a given set

of criteria) at a particular task or achieving a stated goal

 When speaking of an “authentic” intelligence, there is an emphasis on learning,

adaptation, and flexibility within a wide range of environments and scenarios

Emerj’s Definition of Artificial Intelligence:

*NOTE: Artificial intelligence (AI) can be separated into two branches of
entities – that of ‘smart’ computers or systems (such as today’s deep
learning), and a still unrealized ‘artificial general intelligence’ or AGI. We
include this as a preface for helping to distinguish between the two in our
present state of technological development. Our definition attempts to define
an entity rather than a field of study, and also utilizes broad or somewhat open
terminology to allow room for evolution and growth of the AI field as we know
it. Our attempt at an informed, “living” definition of AI is below:

“Artificial intelligence is an entity (or collective set of cooperative entities), able to

receive inputs from the environment, interpret and learn from such inputs, and
exhibit related and flexible behaviors and actions that help the entity achieve a
particular goal or objective over a period of time.”

* How We Arrived at Our Definition:

As with any concept, artificial intelligence may have a slightly different
definition, depending on whom you ask. We combed the Internet to find five
practical definitions from reputable sources:

1. “It is the science and engineering of making intelligent machines, especially

intelligent computer programs. It is related to the similar task of using
computers to understand human intelligence, but AI does not have to confine
itself to methods that are biologically observable.” – Stanford

2. “Artificial Intelligence is the study of man-made computational devices and

systems which can be made to act in a manner which we would be inclined to
call intelligent.” – The University of Louisiana at Lafayette
3. “Defining artificial intelligence isn’t just difficult; it’s impossible, not the least
because we don’t really understand human intelligence. Paradoxically,
advances in AI will help more to define what human intelligence isn’t than
what artificial intelligence is.” – OReilly

4. “The ability of a machine communicating using natural language over a

teletype to fool a person into believing it was a human. “AGI” or “artificial
general intelligence” extends this idea to require machines to do everything
that humans can do, such as understand images, navigate a robot, recognize
and respond appropriately to facial expressions, distinguish music genres,
and so on.” Matt Mahoney, PhD, Data Compression Expert

5. “The scientific understanding of the mechanisms underlying thought and

intelligent behavior and their embodiment in machines.” AITopics.org

We sent these definitions to experts whom we’ve interviewed and/or included

in one of our past research consensuses, and asked them to respond with
their favorite definition or to provide their own. Our introductory definition is
meant to reflect the varied responses. readers should note, Machine learning
and artificial intelligence share the same definition in the minds of many,
however, there are some distinct differences readers should recognize as
well. Below are some of their responses:

Dr. Andras Kornai, Budapest Institute of Technology

I like them all, except for #3, in that no exact definition is needed to work on a
problem. We still don’t fully understand gravity, and can measure it only to one
part in ten thousand (most other physical phenomena we can now measure to
one part in a billion or even better), yet it would be silly to say “it’s impossible
to define gravity”.

Some clarity may be required to distinguish “good old-fashioned” AI from

modern AGI. GOFAI was centered on conceptual modeling by
symbol manipulation, e.g. the planning required to win a chess game, and
took the Turing Test as its central goal. AGI demands more,
reaching comparable levels in all forms of human intelligence are seen
as goals. This development is well emphasized in #4.

#1, #2, and #5 are right in emphasizing that A(G)I is primarily about creating
algorithms that show intelligent behavior, and is not to be confused with
cognitive science or brain modeling which aim at explaining how a particular
hardware, the human brain, gets there. We may, or may not, be able to steal
ideas from nature, this remains to be seen.

Dr. Ashok Goel, Georgia Institute of Technology

Artificial Intelligence is the science of building artificial minds by understanding

how natural minds work and understanding how natural minds work by
building artificial minds.

Dr. Pei Wang, Temple University

This is a complicated problem. I have a paper on it: What Do You Mean by

“AI”?, in which “intelligence” is defined as “adaptation with insufficient
knowledge and resources.”

(In response to definitions 1 and 2) – Defining AI using “intelligent” or

“intelligence” is a circular definition. The statement is agreeable, but does not
provide clear guidance to the research.

(In response to definition 3) – Our understanding of human intelligence is a

matter of degree. This opinion is encouraging blind trial-and-error, which is not
good advice for any scientific research.

(In response to definition 4) – Too anthropocentric. AGI can be easily

distinguishable from human beings, while still being considered as highly
intelligent.

(In response to definition 5) – Better than the others, though still uses
“intelligent”.

Dr. Vincent Müller, Anatolia College

Definition 1 is okay. Of course it leaves the minor question, ‘what ‘intelligence’

means’, open. I think it’s important to see that AI is about making, and that it is
distinct from cognitive science – even though traditionally this was seen
otherwise. That’s not a definition, however.
(Dr. Müller provides a link to his related work: New developments in the
philosophy of AI)

Dr. Dan Roth, University of Illinois at Urbana-Champaign

Any definition of Artificial Intelligence will have to be vague enough due to our
inability to define Human Intelligence. But I would say that this is the scientific
field that attempts to understand the foundations of intelligent behavior from a
computational perspective. It focuses on developing theories and systems
pertaining to intelligent behavior, at the heart of which is the idea that learning,
abstraction and inference have a central role in intelligence.

Dr. Stuart Russell and Dr. Peter Norvig

We also found and chose to include a more recent and commonly-accepted

textbook definition to build on our perspective. In “Artificial Intelligence: A
Modern Approach”, Stuart Russell and Peter Norvig defined AI as “the
designing and building of intelligent agents that receive percepts from the
environment and take actions that affect that environment.” This definition by
its nature unites many different splintered fields – speech
recognition, machine vision, learning approaches, etc. – and filters them
through a machine that is then able to achieve a given goal.

AI Semantics:
Strong AI versus Weak AI

Strong AI – Also known as deep AI or what some might deep AGI; the idea
that a computer can be made or raised to intelligence levels that match
human beings’.

Weak AI – Otherwise known as narrow AI; the idea that computers can be
endowed with features that mirror or mimic thought or thinking processes,
making them useful tools for figuring out how our own mind works. Narrow AI
systems also enhance or augments human “intelligence” by delivering
calculations, patterns and analyses more efficiently than can be done by a
human brain.
 The field of artificial life branches out further from traditional AI to include the
study and mimicry of various biological forms and organisms that exhibit a
range of “intelligent” behaviors.

AI Solutions:
One way to categorize AI solutions for commercial and scientific needs is by
level of complexity of the application: simple, complex, or very complex
(though these are, clearly, also open to interpretation). This is an idea
borrowed from the Schloer Consulting Group:

Simple – Solutions and platforms for narrow commercial needs, such as

eCommerce, network integration or resource management.

 Examples: Customer Relationship Management (CRM) software, Content

Management System (CMS) software, automated agent technology

Complex – Involves the management and analysis of specific functions of a

system (domain of predictive analytics); could include optimization of work
systems, predictions of events or scenarios based on historical data; security
monitoring and management; etc.

 Examples: Financial services, risk management, intelligent traffic management

in telecommunication and energy

Very Complex – Working through the entire information collection, analysis,

and management processes; the system needs to know where to look for
data, how to collect, and how to analyze, and then propose suggested
solutions for near and mid-term futures.

 Examples: Global climate analysis, military simulations, coordination and control

of multi-agent systems

AI Continuum:
In similar fashion to types of AI solutions organized by capability, there exists
a continuum of AI in regards to level of autonomy:
 Assisted Intelligence – Involves the taking over of monotonous, mundane tasks
that machines can do more efficiently.

 Example: Robotic Shelf Picking from IAM Robotics

Augmented Intelligence – A step up in a more authentic collaboration of

“intelligence”, in which machines and humans learn from the other and in turn
refine parallel processes.

 Example: Editor from NyTimes

Autonomous Intelligence – System that can both adapt over time (learn on its
own) and take over whole processes within a particular system or entity.

 Example: NASA’s Mars Curiosity Rover

Approaches to Achieving AI:

It seems important that what any definition does not do (as noted by Legg and
Hutter in their paper A Formal Definition of Intelligence for Artificial Systems),
is limit or restrict the inner workings of AI and/or the approaches used to
creating an AI entity.

The methods taken toward achieving a “true AI” or AGI are wide and varied,
but some are closer in line achieving an adaptive, flexible and autonomous
intelligence that is more characteristic of human beings (and likely
intelligences that do/will exist beyond our own).

Approaches that have evolved and continue to receive wide recognition in the
media include (though are not isolated in approach or limited to) the following:

 Artificial neural networks

 Reinforcement learning

 Self-supervised learning

 Multi-agent learning
 Machine learning

Limitations in Defining AI:

No matter what we do, we can’t (at present) escape our biological and social
trapping as human beings, which means that any definition put forth and/or
any test that we conceive to test for a “true” artificial intelligence is at risk
from anthropocentrism and subjectivity.

What we can do is cultivate an awareness of our biases and opinions and

strive to seek out a broader or more “universal” notion of intelligence that
encompasses a range therein; achieving the creation of an artificial
intelligence in our likeness may be one of the ultimate challenges of our times,
but there are likely thousands or even millions types of artificial intelligences
which we could aim to (or may not be capable) of conceiving and creating.

While it’s science’s aim to discover truth and knowledge at the heart of every
system and process in the universe, we should recognize that we may not be
able to understand the workings of an artificial intelligence that we one day
create – in fact, that’s already the case with deep learning and neural networks,
similar to our current complete lack in understanding how the human brain works. At
some point, we may not be able to keep up with an AI’s processing powers
and ways of literally seeing and conceiving of reality as we know it.

NOTE: It’s beyond the scope of this article to give a cohesive, historical
overview of AI, or of today’s landscape. Instead, our intent is to provide a
jumping off point for understanding and further exploring the history, workings,
and consequences of AI in today’s increasingly automated and
augmented landscape.

References Cited:
http://www.schloerconsulting.com/understanding-artificial-intelligence

http://www.recode.net/sponsored/11895802/what-artificial-intelligence-really-
means-to-business

http://www.vetta.org/documents/universal_intelligence_abstract_ai50.pdf
 http://aima.cs.berkeley.edu

http://www.nature.com/news/can-we-open-the-black-box-of-ai-1.20731

What is the difference between data mining, statistics, machine learning and
AI?

http://www.coneural.org/reports/Coneural-03-01.pdf

Related Interviews and Articles on Emerj:

If you’re interesting in getting a lay-of-the-land perspective on the implications
and applications of artificial intelligence, you might enjoy this curated selection
of some of our more popular AI overview topics, listed below:

 Popular Interview: Dr. Yoshua Bengio – The Influence of Neural Networks in Our

Everyday Lives

 Popular Interview: Dr. Nando de Freitas – Deep Learning is Like Building with
Lego

 Infographic: Artificial Intelligence Risk – What Researchers Think is Worth Worrying

About

 Consensus Article: What is Machine Learning?

Image credit: Nomura Connects

What is Machine Learning?

Last updated on November 21, 2019, published by Daniel Faggella
 Daniel Faggella is the founder and CEO at Emerj. Called upon by the United Nations, World Bank,
INTERPOL, and many global enterprises, Daniel is a sought-after expert on the competitive strategy
implications of AI for business and government leaders.
Share to:
LinkedIn Twitter Facebook Email

Typing “what is machine learning?” into a Google search opens up a

pandora’s box of forums, academic research, and here-say – and the purpose
of this article is to simplify the definition and understanding of machine
learning thanks to the direct help from our panel of machine learning
researchers.

In addition to an informed, working definition of machine learning (ML), we aim

to provide a succinct overview of the fundamentals of machine learning, the
challenges and limitations of getting machine to ‘think’, some of the issues
being tackled today in deep learning (the ‘frontier’ of machine learning), and
key takeaways for developing machine learning applications.

This article will be broken up into the following sections:

 What is machine learning?

 How we arrived at our definition (IE: the perspective of expert researchers)

 Machine learning basic concepts

 Visual representation of ML models

 How we get machines to learn

 An overview of the challenges and limitations of ML

 Brief introduction to deep learning

 Works cited

 Related ML interviews on Emerj

We put together this resource to help with whatever your area of curiosity
about machine learning – so scroll along to your section of interest, or feel
free to read the article in order, starting with our machine learning definition
below:

What is Machine Learning?

* “Machine Learning is the science of getting computers to learn and act like
humans do, and improve their learning over time in autonomous fashion, by
feeding them data and information in the form of observations and real-world
interactions.”
The above definition encapsulates the ideal objective or ultimate aim of
machine learning, as expressed by many researchers in the field. The
purpose of this article is to provide a business-minded reader with expert
perspective on how machine learning is defined, and how it works. Machine
learning and artificial intelligence share the same definition in the minds of
many however, there are some distinct differences readers should recognize
as well. References and related researcher interviews are included at the end
of this article for further digging.

* How We Arrived at Our Definition:

(Our aggregate machine learning definition can be found at the beginning of
this article)

As with any concept, machine learning may have a slightly different definition,
depending on whom you ask. We combed the Internet to find five
practical definitions from reputable sources:

1. “Machine Learning at its most basic is the practice of using algorithms to

parse data, learn from it, and then make a determination or prediction
about something in the world.” – Nvidia
2. “Machine learning is the science of getting computers to act without
being explicitly programmed.” – Stanford
3. “Machine learning is based on algorithms that can learn from data
without relying on rules-based programming.”- McKinsey & Co.
 4. “Machine learning algorithms can figure out how to perform important
tasks by generalizing from examples.” – University of Washington
5. “The field of Machine Learning seeks to answer the question “How can
we build computer systems that automatically improve with experience,
and what are the fundamental laws that govern all learning processes?”
– Carnegie Mellon University
We sent these definitions to experts whom we’ve interviewed and/or included
in one of our past research consensuses, and asked them to respond with
their favorite definition or to provide their own. Our introductory definition is
meant to reflect the varied responses. Below are some of their responses:

Dr. Yoshua Bengio, Université de Montréal:

ML should not be defined by negatives (thus ruling 2 and 3). Here is my

definition:

Machine learning research is part of research on artificial intelligence, seeking

to provide knowledge to computers through data, observations and interacting
with the world. That acquired knowledge allows computers to correctly
generalize to new settings.

Dr. Danko Nikolic, CSC and Max-Planck Institute:

(edit of number 2 above): “Machine learning is the science of getting

computers to act without being explicitly programmed, but instead letting them
learn a few tricks on their own.”

Dr. Roman Yampolskiy, University of Louisville:

Machine Learning is the science of getting computers to learn as well as

humans do or better.

Dr. Emily Fox, University of Washington:

My favorite definition is #5.

Machine Learning Basic Concepts

 There are many different types of machine learning algorithms, with hundreds
published each day, and they’re typically grouped by either learning style (i.e.
supervised learning, unsupervised learning, semi-supervised learning) or
by similarity in form or function (i.e. classification, regression, decision tree,
clustering, deep learning, etc.). Regardless of learning style or function, all
combinations of machine learning algorithms consist of the following:

 Representation (a set of classifiers or the language that a computer understands)

 Evaluation (aka objective/scoring function)

 Optimization (search method; often the highest-scoring classifier, for example;

there are both off-the-shelf and custom optimization methods used)

Image credit: Dr. Pedro Domingo, University of Washington

The fundamental goal of machine learning algorithms is to generalize beyond

the training samples i.e. successfully interpret data that it has never ‘seen’
before.

Visual Representations of Machine Learning Models

Concepts and bullet points can only take one so far in understanding. When
people ask “What is machine learning?”, they often want to see what it is and
what it does. Below are some visual representations of machine learning
models, with accompanying links for further information. Even more resources
can be found at the bottom of this article.

Decision tree model

Gaussian mixture model

 Dropout neural
network
 Merging chrominance and luminance using Convolutional Neural Networks

How We Get Machines to Learn

There are different approaches to getting machines to learn, from using basic
decision trees to clustering to layers of artificial neural networks (the latter of
which has given way to deep learning), depending on what task you’re trying
to accomplish and the type and amount of data that you have available. This
dynamic sees itself played out in applications as varying as medical
diagnostics or self-driving cars.

While emphasis is often placed on choosing the best learning algorithm,

researchers have found that some of the most interesting questions arise out
of none of the available machine learning algorithms performing to par. Most
of the time this is a problem with training data, but this also occurs
when working with machine learning in new domains.
Research done when working on real applications often drives progress in the
field, and reasons are twofold: 1. Tendency to discover boundaries and
limitations of existing methods 2. Researchers and developers working with
domain experts and leveraging time and expertise to improve system
performance.

Sometimes this also occurs by “accident.” We might consider model

ensembles, or combinations of many learning algorithms to improve accuracy,
to be one example. Teams competing for the 2009 Netflix Price found that
they got their best results when combining their learners with other team’s
learners, resulting in an improved recommendation algorithm (read Netflix’s
blog for more on why they didn’t end up using this ensemble).

One important point (based on interviews and conversations with experts in

the field), in terms of application within business and elsewhere, is that
machine learning is not just, or even about, automation, an often
misunderstood concept. If you think this way, you’re bound to miss the
valuable insights that machines can provide and the resulting opportunities
(rethinking an entire business model, for example, as has been in industries
like manufacturing and agriculture).

Machines that learn are useful to humans because, with all of their processing
power, they’re able to more quickly highlight or find patterns in big (or other)
data that would have otherwise been missed by human beings. Machine
learning is a tool that can be used to enhance humans’ abilities to solve
problems and make informed inferences on a wide range of problems, from
helping diagnose diseases to coming up with solutions for global climate
change.

Challenges and Limitations

“Machine learning can’t get something from nothing…what it does is get more
from less.” – Dr. Pedro Domingo, University of Washington

The two biggest, historical (and ongoing) problems in machine learning have
involved overfitting (in which the model exhibits bias towards the training data
and does not generalize to new data, and/or variance i.e. learns random
things when trained on new data) and dimensionality (algorithms with more
features work in higher/multiple dimensions, making understanding the data
more difficult). Having access to a large enough data set has in some cases
also been a primary problem.

One of the most common mistakes among machine learning beginners is

testing training data successfully and having the illusion of success; Domingo
(and others) emphasize the importance of keeping some of the data set
separate when testing models, and only using that reserved data to test a
chosen model, followed by learning on the whole data set.

When a learning algorithm (i.e. learner) is not working, often the quicker path
to success is to feed the machine more data, the availability of which is by
now well-known as a primary driver of progress in machine and deep learning
algorithms in recent years; however, this can lead to issues with scalability, in
which we have more data but time to learn that data remains an issue.

In terms of purpose, machine learning is not an end or a solution in and of

itself. Furthermore, attempting to use it as a blanket solution i.e. “BLANK” is
not a useful exercise; instead, coming to the table with a problem or objective
is often best driven by a more specific question – “BLANK”.

Deep Learning and Modern Developments in Neural Networks

Deep learning involves the study and design of machine algorithms
for learning good representation of data at multiple levels of abstraction (ways
of arranging computer systems). Recent publicity of deep learning
through DeepMind, Facebook, and other institutions has highlighted it as the
“next frontier” of machine learning.

The International Conference on Machine Learning (ICML) is widely regarded

as one of the most important in the world. This year’s took place in June in
New York City, and it brought together researchers from all over the world
who are working on addressing the current challenges in deep learning:

1. Unsupervised learning in small data sets

2. Simulation-based learning and transferability to the real world
Deep-learning systems have made great gains over the past decade in
domains like bject detection and recognition, text-to-speech, information
retrieval and others. Research is now focused on developing data-efficient
machine learning i.e. deep learning systems that can learn more efficiently,
 with the same performance in less time and with less data, in cutting-edge
domains like personalized healthcare, robot reinforcement learning, sentiment
analysis, and others.

Key Takeaways in Applying Machine Learning

Below is a selection of best-practices and concepts of applying machine
learning that we’ve collated from our interviews for out podcast series, and
from select sources cited at the end of this article. We hope that some of
these principles will clarify how ML is used, and how to avoid some of the
common pitfalls that companies and researchers might be vulnerable to in
starting off on an ML-related project.

 Arguably the most important factor in successful machine learning projects is

the features used to describe the data (which are domain-specific), and having
adequate data to train your models in the first place

 Most of the time when algorithms don’t perform well, it’s due a to a problem with
the training data (i.e. insufficient amounts/skewed data; noisy data; or
insufficient features describing the data for making decisions

 “Simplicity does not imply accuracy” – there is (according to Domingo) no given

connection between number of parameters of a model and tendency to overfit

 Obtaining experimental data (as opposed to observational data, over which we

have no control) should be done if possible (for example, data gleaned from
sending different variations of an email to a random audience sampling)

 Whether or not we label data causal or correlative, the more important point is
to predict the effects of our actions

 Always set aside a portion of your training data set for cross validation; you want
your chosen classifier or learning algorithm to perform well on fresh data

Works Cited
1 – http://homes.cs.washington.edu/~pedrod/papers/cacm12.pd

2 – http://videolectures.net/deeplearning2016_precup_machine_learning/
 3 – http://www.aaai.org/ojs/index.php/aimagazine/article/view/2367/2272

4 – https://research.facebook.com/blog/facebook-researchers-focus-on-the-
most-challenging-machine-learning-questions-at-icml-2016/

5 – https://sites.google.com/site/dataefficientml/

6 – http://www.cl.uni-heidelberg.de/courses/ws14/deepl/BengioETAL12.pdf

Related Machine Learning Interviews on Emerj

One of the best ways to learn about artificial intelligence concepts is to learn
from the research and applications of the smartest minds in the field. Below is
a brief list of some of our interviews with machine learning researchers, many
of which may be of interest for readers who want to explore these topics
further:

 The Science of Machine Learning with Dr. Yoshua Bengio (one of the world’s
foremost ML experts)

 UPENN’s Dr. Lyle Ungar on Using Machine Learning to See Patterns and
Meaning on Social Media

 Silicon Valley AI Consultant Lorien Pratt on the Business Use Cases of Machine
Learning