The goal of this presentation is to provide you with a basic understanding of AI and to prepare you to think about how your organization might apply it.
2. 226/23/17 26/23/17 2
The goal of this presentation is to provide you with a basic understanding of AI
and to prepare you to think about how your organization might apply it
3. 33
Adela
VILLANUEVA
W.W.W:
www.adelavillanueva.com
.linkedin.com/in/adelavillanueva/
A_d_e_l_a
About the author:
Trilingual Consultant with 12 y. of experience working with F500-Corporations, Public
sector, Academia and Start-ups around the Globe developing innovation capabilities &
ecosystems, creating growth strategies, building new businesses, products & services
and managing digital transformations.
I am also the Founder & CEO of The GIN (www.the-gin.com) a global collective of
innovators. [ If you are an innovator, please feel free to register]
This presentation is inspired by the “Cognitive technologies and Applications for
Business” course from Deloitte University, that I enrolled in 2016.
Enjoy!
6. 66
Cognitive
Technologies
or
Artificial
Intelligence (AI): Computer systems that can perform discrete
tasks* that normally require human intelligence.
* Visual perception, speech recognition, decision making under uncertainty, learning and translation under languages etc.
9. 99
And they do it through:
- Rules based systems [ If…. then]
- Taxonomies [ classification]
- Bayer nets [ diagnosis: from symptoms to cause & predictions: from cause to symptoms]
10. 1010
AI agents
are supposed to:
- operate autonomously
- perceive their environment
- persist over time
- adapt to change
- create and pursue goals
13. 1313
Computer perception and physical actions
Computer
Vision
// What:
The ability of computers to identify objects, scenes, and activities in unconstrained (that is, naturalistic) visual
environments
// How it works:
Break each image into its components parts:
• Break image into pixels or dots of colors
• Layer it with low level features
• Then go to high analysis like lines and areas and even objects
// The challenge:
We don’t know “how we see” so it is difficult to program a computer to do so. Objects look different
depending on position , angle, light…
// The benefits:
This technology can bring a lot to the multimedia process ( e.g.: reproduce physical features to virtual world)
// Examples & applications:
Taggin photos on Facebook, medical imaginering, autonomous driving, robotics, surveillance, gesture
detection, parking management, etc.
Computer
Vision
14. 1414
Computer perception and physical actions
Computer
Vision
// What:
The ability of computers to work with text the way humans do—for instance, extracting meaning from text or
even generating text that is readable, stylistically natural, and grammatically correct
// How it works:
Splitting by tasks
• Begin by segmenting the document into sections
• Then Paragraphs
• Then sentences
• Then breaking into words
• From words to meanings
// The challenge:
Machines sometimes struggle to understand the context within which the conversation is taking place
// The benefits:
Ability to process quickly very large amount of text and provide real-time/ fast feedback
// Examples & applications:
Google translator, chat-bot customer feedback, summarizing documents, extracting information, etc.
Natural Language
Processing
15. 1515
Computer perception and physical actions
Computer
Vision
// What:
The ability to enable agents to interact with their physical environment, in which multiple cognitive
technologies are combined. We can find 3 categories:
• Manipulators ( or robots arms, anchored to their work place)
• Mobile Robots ( ground our air vehicles, like drones)
• Mobile Manipulators ( combine both as humanoids)
// How it works:
High-performance sensors, actuators, effectors & hardware
// The challenge:
Uncertainty (machines need to observe all its environment and sometimes they don’t understand it
accurately)
// The benefits:
The new generation of autonomous robots can sense and respond to their environment, plan for their
actions, and in some cases interact and work alongside people
// Examples & applications:
Manufacturing, transportation, healthcare ( surgery robots), robots for hazardous environements, personal
services ( vacumm cleaners), entertainment ( personal robots), human augmentation ( exoskeleton), etc.
Robotics
16. 1616
Computer perception and physical actions
Computer
Vision
// What:
The ability to automatically and accurately transcribe human speech (words & even the tone or emotion )
// How it works:
Breaking each action:
• Analyze the recording of speech [ acoustic]
• Recognize speech- understanding phonemes
• Match them with words
• Then word sequence recognition
// The challenge:
• Handling with different accents and background noise
• Homophone distinction
• Working quickly at speed of human speech
// The benefits:
It is 80% of accuracy today. It will allow us to access to huge information through spoken, recorded human
speech as well as to handle big amounts of data . It is hands-free which delivers a great UX experience
// Examples & applications:
Siri, hands-free writing ( medical dictation), mobile devices control, computer system control ( car),
surveillance commands, etc.
Speech
Recognition
17. 1717
Computer task and analysis actions
Computer
Vision
// What:
The ability to automatically devise a sequence of actions to meet goals and observe constraints
// How it works:
Automatically devising a plan to achieve goals given:
• A description of the initial stage
• The desired goal
• The list of actions that are possible
// The challenge:
Managing the complexity of options at a computation time
// The benefits:
Ability to re-plan [ in real time] and see other potentials solutions
// Examples & applications:
Google maps, Routing / delivery optimization, ressources planning, etc.
Planning
& Scheduling
18. 1818
Computer task and analysis actions
Computer
Vision
// What:
The ability of computer systems to improve their performance by exposure to data without the need to
follow explicitly programmed instructions. There are three main types of machine learning: supervised
learning, unsupervised learning and reinforcement learning
// How it works:
The machine analyzes data and automatically builds models from that data. The machine can feed on
data (from internet or from their own data base) and adapt itself to make more precise predictions and act
accordingly. 2 techniques for machine learning: Artificial Neural-networks & Support vectors-machines
// The challenge:
Acquiring and labeling training data ( can be costly and take longtime to create a large training data set)
// The benefits:
Allow to anticipate all possibilities (which humans can’t] . ML can discover patterns, make predictions, better
on time with exposure to data. Humans may not know how to program a solution, so we’re going to expect
machine-learning systems to learn in the wild or learn more autonomously with less feedback from the users.
( to run more independently)
// Examples & applications:
IBM Watson ML, Google Ads, Personal assistants, recommendation engines, etc.
Machine
Leaning
19. 1919
Computer task and analysis actions
Computer
Vision
// What:
The ability to use databases of knowledge and rules to automate the process of making inferences about
information ( also known as the ‘expert system’)
// How it works:
Through “if-then" statements that uses a set of assertions, to which rules on how to act upon those assertions
are created
// The challenge:
Acquiring the requisite knowledge from experts. These systems need very specific rules containing the
necessary textbook and judgmental knowledge about their domains, which sometimes it is not 100%
complete and accurate (inferences are done only under the provided knowledge); producing sometimes
inaccurate output for inadequate knowledge base
// The benefits:
Rule-based systems are generally implemented as single-thread programs. The advantage to this type of
approach, is that if the system is designed well, then the expert's knowledge can be maintained fairly easily,
just by altering whichever rules need to be altered
// Examples & applications:
Mortgages, credit card authorization, fraud detection, e-commerce, demographic personalization, etc.
Ruled based
systems
20. 2020
Computer task and analysis actions
Computer
Vision
// What:
The ability to automate complex decisions and trade-offs about limited resources, in order to find the
optimal solution out of all possible solutions
// How it works:
Traditional methods of optimization use differential calculus or brute force search to guide the exploration of
the solution space (a space consisting of all possible solutions)
// The challenge:
They can mislead us into believing that we have optimized a problem, when in reality we have only reached
one part of it
// The benefits:
Possibility to maximize desirable traits
// Examples & applications:
A/B testing , Route optimization for smart waste management, Company/ product profit recommendation,
etc.
Optimization
30. 3030
PRODUCT AI to deliver a value to the end consumer/ customer.
// benefits:
// examples:
/ User convenience (e.g.: rapidity of execution)
/ Simplicity of the task [ e.g.: knowing preferences)
/ Confidence (e.g.; doing complex business decisions]
/ Emotional (creating emotional bonding [e.g.: speech recognition/ voice control)
/ Netflix. Uses machine learning for recommendations.
/General Motors use computer vision in new models of cars
( eye tracking recognition to identify fatigue )
31. 3131
PROCESS AI to automate tasks or business process internal to an organization.
// benefits:
// examples:
/ Reduce time of execution
/ Increase predictability and reliability
/ Reduce cost ( skilled workers can focus on important tasks)
/ Hong Kong’s subway: predictive maintenance. They use algorithms that
create optimal engineering schedules; saving them 2 days of planning
work per day.
/ Cincinnati's ‘Children Hospital Center’: they developed a natural
language processing system to read free- forms clinical and handwriting
notes. Reducing nurses’ work flow by 22%.
32. 3232
INSIGHT AI to analyze large amounts of data to discern paths or make
predictions
// benefits:
// examples:
/ Better and faster decisions that can improve operating and strategic performance
/ Make all kinds of analysis possible
/ Intel: uses machine learning to improve sales effectiveness ( buying
patterns, promotions, etc.)
/ Aetna: used machine learning to produce models that predicted
patients’ risk of developing metabolic syndrome. It recommended
which kind of medical intervention was the most suitable.
36. 3636
Look at your business, processes, products, customers, markets
to determine where the use of those technologies may be:
VIABLE
When cognitive technologies
actually work
VALUABLE
When it is worth* applying
cognitive technologies
VITAL
When cognitive technologies
may be absolutely required
* Just because something can be done doesn't mean that we must to do it