Anda di halaman 1dari 31

Artificial Intelligence

A wide range of tasks are enabled by computers

 playing games
 recognising faces and speech
 making decisions under uncertainty
 learning
 translating between languages
Human mental activities demand intelligence

 writing computer programs


 doing mathematics
 engage in common-sense reasoning
 driving an automobile
Computer systems can perform such tasks

 diagnose diseases
 plan the synthesis of complex organic chemical compounds
 solve differential equations in symbolic forms
 analyse electronic circuits
 understand limited amounts of human speech and natural
language text
 such systems possess some degree of artificial intelligence
Intelligent Retrieval from Databases

 database systems are computer systems that store a large


body of facts
 answer users’ questions about a subject
 subject becomes interesting when retrieval of answers
requires deductive reasoning
 understanding the query and deducing an answer may
require knowledge beyond that explicitly represented in the
subject domain database
 common knowledge
Intelligent Retrieval from Databases

 “Joe Smith works in the Purchasing department”


 “ Joe Smith was hired on October 8, 1996”
 “The Purchasing department has 17 employees”
 “John Jones is the manager of the Purchasing department”
 “Who is Joe Smith’s boss?”
 the system has to know somehow that the manager of the
department is the boss of the people who work in that
department
Expert Consulting Systems
 automatic systems
 provide human users with expert conclusions about
specialised subject areas
 diagnose diseases
 suggest structures for complex organic chemicals
 challenge: how to represent the knowledge of human
experts
 imprecise, uncertain, anecdotal
 technique of rule-based deduction
 a large set of simple rules  rules guide dialogue
between user and system  deduce conclusions
Robotics

 automate tasks performed unconsciously by humans


 techniques for modelling states of the world
 the process of change from one world state to another
Combinatorial and Scheduling problems

 class of problems concerned with specifying optimal


schedules or combinations
 travelling salesman problem
 eight queens problem
 place 8 queens on a standard chessboard
 no queen should be able to capture any of the other
 generate a combinatorial explosion of possibilities
Cognitive Technologies

 the field of AI is different from the technologies that


emanate from that field
 cognitive technologies
 machine learning
 computer vision
 speech recognition
 natural language processing
 robotics
Cognitive Technologies

 will likely have a profound impact on work, workers and


organizations
 eliminate jobs
 redesign work
 create new opportunities for workers
Not everyone believes that organizations should be preparing
for a future without jobs and workers

 “the challenges in substituting machines for workers in


tasks requiring adaptability, common sense and creativity”
 “getting rid of a really dull job that we shouldn’t be torturing
people with”
 more likely, parts of jobs will be automated by CT
 workers will be interacting with automated smart machines
 as airline pilots and workers in advanced factories
AI can support three important business needs

 automating business processes


 gaining insight through data analysis
 engaging with customers and employees
Processes

 use CT to enhance, scale or automate business processes


 automatic data entry
 automatic handwriting recognition
 automatic planning and scheduling (with planning and
optimisation algorithms)
 automatic customer service with speech recognition
 nlp
Natural Language Processing

 difficult to develop computer systems that are capable of


generating and understanding fragments of a natural
language
 primary difficulties
 language has evolved as a communication medium
between intelligent beings
 transmits a bit of “mental structure” from one brain to
another
 under circumstances in which each brain possesses large,
highly similar surrounding mental structures
 that serves as a common context
Natural Language Processing

 each participant knows that the other possesses this


common mental structure
 a word to the wise from another
 computer systems must possess both the contextual
knowledge and the process for making the inference (from
the contextual knowledge)
 knowledge representation and processing
Process Automation

 automation of physical tasks


 back-office administrative and financial activities
 robotic process automation technologies
 “robots” (code on a server) act like a human
 inputting and consuming information from multiple IT
systems
Tasks enabled by Process Automation

 transferring data from e-mail and call center systems into


systems of record
 updating customer files with address changes
 replacing lost credit or ATM cards
 reaching into multiple systems to update records
 handle customer communications
 reconciling failures to charge for services
 extracting information from multiple document types,
across billing systems
 “reading” legal and contractual documents, using NLP
Robotic Process Automation

 least expensive
 least “smart”
 applications aren’t programmed to learn and improve
 can work across multiple back-end systems
 replacing administrative employees was not the objective
Machine Learning enable companies to optimise processes

 personalising customer service


 improving customer loyalty
 hiring the right people
 automating exception handling in financial processes
 detecting fraud
 predictive maintenance
 predicting supply chain risks
 drone and satellite based asset management
 retail shelf analysis
Products

 embed CT in products to provide “intelligent” behaviour,


natural interfaces (speech and visual) and automation
 robotic toys, thermostats
 robotic vacuum cleaners
 autonomous vehicles
Insights

 use algorithms to reveal patterns, make predictions and


guide more effective actions
 Intel employs machine learning to recommend to its sales
force which customers to call next and what to offer them
 other applications enhance, rather than automate, existing
decision-making processes
 perform analyses not being done before
Cognitive Insights – Machine Learning

 predict what a particular customer is likely to buy


 identify credit fraud in real time, detect insurance claims
fraud
 analyse warranty data to identify safety or quality problems
in automobiles
 provide insurers with more accurate and detailed actuarial
modelling
 discover meaningful patterns in factory
 pinpoint the root cause of a problem
Cognitive Insights from Machine Learning – how different from
traditional insights

 much more data-intensive and detailed


 models typically trained on some part of data set
 models get better
 ability to use new data to make predictions improve over
time
Versions of Machine Learning

 deep learning
 mimics the activity of the human brain
 to recognise patterns
 recognise images and speech
 identify probabilistic matches
 data likely to be associated with the same person or
company
 appears in slightly different formats
 across databases
Cognitive Engagement

 engage employees and customers


 using nl chatbots
 intelligent agents offer 24/7 customer service
 from password resets to technical support questions
 all in customer’s n l
 internal sites for answering employee questions
 benefits, HR policy
 product and service recommendation systems for retailers
 health treatment recommendations
Understanding the technologies
 which technologies perform what type of tasks
 rule based expert systems and rpa are transparent in how
they work
 not capable of learning or improving
 deep learning is great at learning from large volumes of
labelled data
 impossible to understand how it creates the models it
does
 this “black box” issue can be problematic in regulated
industries
 where regulators insist on knowing why decisions were made
in a certain way
Unintended consequences of automation

 tasking human with monitoring a process (that has been


automated) can cause errors to get unnoticed
 impossible to pay attention for more than half an hour to an
information source that hardly changes
 people tend to lose their skills if not practised regularly
 when humans need to take command of an automated
system (such as autopilot), they are ill-prepared
 too much automation (cruise control) can make drivers
less vigilant
Skills that will become more valuable
 as routine tasks are subject to automation, skills required for
those tasks will become less valuable
 skills defined to perform broadly or loosely-defined jobs will be
valuable
 common sense
 general intelligence
 flexibility
 creativity
 critical thinking
 tolerance to ambiguity
 interpersonal interactions
 emotional intelligence
 empathy
Skills resistant to computerization

 to perform ‘abstract’ tasks


 problem-solving
 intuition
 creativity
 persuasion
 to perform ‘manual’ tasks
 situational adaptability
 visual and language recognition
 in-person interactions
Jobs not yet for computers

 designing products, services, entertainment that delight


people
 creating something novel or beautiful
 providing the highest level of customer service
 no substitute for the experience of a well-trained and
well-equipped human with high e q, energy and empathy

Anda mungkin juga menyukai