Professional Documents
Culture Documents
S i l i c o n Va l l e y
California
What managers,
intrapreneurs, entrepreneurs Ÿ The Frameworks of
Articial Intelligence
and innovators need to know Innovation
to start getting business
results with Artificial Ÿ The Technological
Infrastructures That Enable
Intelligence Articial Intelligence
“Some people call this articial intelligence, but the reality is this
technology will enhance us. So instead of articial intelligence, I
think we'll augment our intelligence.”
Labour cost reduction Reduction of production costs by It is a quick and economic win and
eliminating human working hours understanding use cases in this
area can generate opportunities.
Improve User Experience Personalize product offerings, It produces tremendous value for
optimize the quality of content, the end customers in that it makes
extend the user interface to audio the interaction with the machine
and other forms of augmentation, more personalized and
dene an interactive sequence of contextualized. At the same time it
user interactions that differ from creates the opportunity for the
user to user e.g. the pattern for the enterprise to learn more about the
expert user versus the pattern for a specic customer and gain
beginner user. business insights on specic
customer groups.
Employee Satisfaction Adopt solutions that reduce or This is a very important area for
mitigate fatigue for particularly companies interested in
consuming jobs increasing the quality of their work
environment and getting a
reputation as top employer
Risk mitigation: The adoption of AI to multiply the Understanding this value driver
study of new product introductions and use cases in this area is
New Product Development or drastically reduce the cost and crucial to reduce risks in industries
time to dene new products. where new product innovation is
crucial but also very risky.
Supply Optimization The adoption of AI to make the Think Amazon fresh or Uber, they
supply of a service or good more have a supply that they cannot
rational so the service or good is waste. This concept applies to
not wasted. other industries and managers
should have a clear view of how it
could be benecial in their
ecosystems.
Next Generation It is the process not only to forecast It is important for managers
Inventory Management Demand, but to proactively feed working in industries where
information such as visual and inventory is delivered and
RFID into a cycle to understand managed in various stages.
how to improve each and every
stage of inventory management.
For example, it is like nding items
in the shelves of a retail chain.
People Task process It is the intelligent routing of This is the most common
optimization personnel and their job tasks in a application of process
complex ecosystem. optimization and could have
unprecedent applications in your
own contest. Studying use cases in
this area will serve as a platform to
generate ideas that can be of
application in your reality.
THE FRAMEWORKS OF ARTIFICIAL
INTELLIGENCE INNOVATION
Autonomous Intelligence
Understanding this
framework helps identify
opportunities and
constraint to completely
eliminate a human from
a process.
Term What It is
A sector of engineering concerned with giving software programs the ability to learn and
Machine Learning execute tasks for which no one has coded a specic execution protocol.
A sector of science and engineering concerned with extracting information and insights
Data mining from medium and large data sets.
A branch of science and engineering concerned with trying to give software programs
Articial intelligence cognitive abilities like human beings possess. The term is also refers to studies identifying
what intelligence is and how it works beyond human beings
This is a noun that today is used to describe the ecosystem of technologies and processes
Big data necessary to handle internet scale amount of data.
A sub sector of Machine Learning concerned with a relatively new approach to Machine
Deep learning Learning. This is an approach that uses a deep net of articial neural networks and in some
cases, it achieves far better results than other Machine Learning methods.
A process within Machine Learning to teach a program a certain task through examples
Supervised Learning prepared and curated by scientists/engineers/practitioners.
A scripted programming language that's easy to use for everyone. It is the programming
Python language with ready to use Machine Learning algorithms.
It is a process of eliminating or cleaning some of the data from a data set because they are
Data Cleansing not reliant or could not be treated and ultimately would negatively affect the Machine
Learning process.
It is the process to make some modications to the data used in a Machine Learning or an
Data preparation analytics project. Usually it consists of simple math operations focused on making the data
more informative for the algorithms.
It is the process of identifying mathematical expressions that can be used to alter data input
Feature analysis for Machine Learning algorithms in such a way that the algorithm will be able to better solve
the problem at hand.
Model Training It is the process of teaching a software algorithm how to solve a specic problem.
It is an outcome of a teaching session where the algorithm learns one specic case very
Model Overtting well, but then fails to accomplish a larger set of cases.
It is the process of verifying that the model at hand is performing as per the expectations
Model testing (same, better, worse) by the programmer training the model.
It is a way to assess the quality of a process or algorithm. It conrms how many items
Precision selected by the process are correctly selected and how many are incorrectly selected.
Items selected as belonging to an ensemble which are not in reality part of the ensemble. It
False Positive is another way to assess the quality of a Machine Learning process.
Items that were not selected as part of an ensemble, but they should have been. Another
False negative complementary measurement of Machine Learning process quality.
It is a type or class of algorithms inspired by the human brain and consists of articial
Neural Network neurons connected to one another.
It is an operation that attributes a label to a sub set of data. At high level it is the operation
Classication that allows a software program to distinguish between animals and plants or between
bananas and oranges.
It is a mathematical operation that given a trend in time can extrapolate a trend in the
Regression future.
A process sometimes exemplied from a software algorithm that discovers groups of data
Clustering within a larger set of data
A process sometimes exemplied from a software algorithm, that discovers one or more
Outlier detection data in a set that looks different from all the other data (e.g. a black sheep).
It is a process used to teach software algorithms how to solve a constant within the
Backpropagation algorithm starting from examples of labled data.
Nearest Neighbour It is a particular type of clustering where a software program divides a current data set into
classier multiple data sets by leveraging the reciprocal distance between the data.
It is a particular type of Nearest Neighbour clustering, where the number of clusters (k) is
K-means decided by a priori and is one of the inputs in the model.
It is a math process and type of software algorithm to predict what the likelihood of one data
Decision Tree to follow a certain behaviour pattern. E.g. will this customer default on credit? Will this
customer buy a black Ferrari?
It is a particular type of classication where a software program divides a current data set
Linear discriminant into two classes (e.g. Seniors with savings and Seniors with low savings) by utilising a
straight line to divide the two subgroups.
It is a particular type of classication where a software program divides a current data set
Rule Based Classier into two (or more) classes (e.g. Seniors with savings and Seniors with low savings) by
utilising a simple repetitive rule to divide the two subgroups.
It is a particular type of classication where a software program divides a current data set
into two (or more) classes (e.g. Seniors with savings and Seniors with low savings) by
Bayesian Classier utilising Bayes rule to divide the two subgroups. Bayes rule is a statistical rule that predicts
the outcome of events based on past events.
Natural Language It indicates an area of Machine Learning concerned with giving software algorithms the
Processing ability to understand human language.
Natural Language It indicates an area of Machine Learning concerned with giving software algorithms the
Generation ability to write human language, automatically.
It is a software that lets you or your organisation store les that are bigger than what's possible
Hadoop to store on one single computer or server and it allow you to process the data in the les.
Therefore, Hadoop accelerates the speed to process the data on each single cluster.
Spark is a software that lets you or your organisation process data as one of the functions of
Hadoop. However, Spark was created to handle the part of Hadoop that processes the
Spark streaming data, as they come into the data process pipeline. It is generally considered that
Spark increases the speed of this process.
It is a Google developed software environment that enables data scientists to prototype and
Tensor Flow develop Machine Learning algorithms quickly. It is used for applications that work on very
large data sets and when multiple types of Machine Learning functionalities are necessary.
The Graphic Processing Unit (GPU) is a form of electronic computer processor that
specializes in processing large blocks of data such as those found in images, videos and
GPU long-time sequences. They have been particularly useful to train (teach) Machine Learning
algorithms and implement some algorithms. It provides signicant improvements in terms
of processing speed over CPUs.
The Tensor Processing Unit (TPU) is a form of electronic computer processor invented by
Google to perform mathematical operations common across deep learning. It provides
TPU signicant improvements in terms of energy efciency and processing speed over GPUs
and therefore, also CPUs.
Is a software process that allows a computer to identify and label the object or the objects
Image recognition contained within an image. E.g. dog versus Cat.
It is a software process that allows a computer to identify a box (or multiple boxes)
Object Detection containing a specic object (or multiple objects) within an image.
INNODEMIA
ARTIFICIAL
INTELLIGENCE
PRODUCT AND
PROCESS
INNOVATION
Everything you need to know about building
insanely good AI products, improving
processes and invest in AI innovation
Learn to Understand Master Learn the Acquire unique Learn how to Learn the Learn how to
build A.I. Products interaction insights from industry persuade and nuances of AI value your
nuances specialists that will
insanely and Master with R&D sell your A.I. product data assets
good AI negotiations and of AI empower you to Project within devdevelopment and extract
Products and with A.I. Engineering product connecting the dots your economic
Processes vendors Stake and innovate in organization
dev your own eld benets from
Holders them