CT-223

Tópicos de Inteligência
Artificial
Prof. Paulo André Castro
pauloac@ita.br
www.comp.ita.br/~pauloac
ITA (Instituto Tecnológico de Aeronáutica)
Divisão de computação (IEC), Prédio da IEC/IEE, Sala 110
Paulo André Lima de Castro
• Bolsista de Produtividade Desen. Tec. e Extensão Inovadora do CNPq Nível 2.

• Engenheiro de Computação pelo Instituto Tecnológico de Aeronáutica (ITA, 1997), Mestre e

Doutor pela Escola Politécnica da Universidade de São Paulo (Poli/USP 2009).
• Pós-doutorado na City University of New York (CUNY, 2013)

• Atualmente, professor do Instituto Tecnológico de Aeronáutica (ITA) e Chefe do

Departamento de Metodologias de Computação da Divisão de Ciência da Computação do
ITA.

• Participei de diversos projetos de Pesquisa e Desenvolvimento incluindo desenvolvimento

de simuladores, avaliação de segurança da informação em sistemas computacionais e
aplicação de técnicas inteligentes em sistemas distribuídos.

• Realizo pesquisas na área de Inteligência Artificial com ênfase em Sistemas multiagentes,

atuando principalmente nos seguintes temas: agent-based finance, agentes autônomos e
aplicações de técnicas inteligentes especialmente em economia e finanças
Ementa da disciplina
• CT-223 – Tópicos em Inteligência Artificial. Resolução de Problemas
através de Busca. Aprendizado de máquina: paradigmas supervisionado
e não supervisionado. Árvores de decisão, Árvores de regressão e
modelo. Random Forests. Raciocínio probabilístico. Modelos gráficos
probabilísticos. Näive Bayes., Redes Bayesianas. Inferência bayesiana,
clustering . Aprendizado por conjunto (ensemble). Simulações
Multiagentes e Agent Based Modelling . Aplicações de Inteligência
Artificial a economia, finanças e simulação baseada em agentes

• Bibliografia: Witten, I., Frank, E. Data Mining: Practical Machine learning

Tools and Techniques. Elsevier. 2005. RUSSEL, S.; NORVIG, P.
Inteligência Artificial: Uma abordagem moderna. 3a. ed. Rio de Janeiro:
Elsevier Editora, 2009.

• Outras referências: Korb, K. Nicholson, A. Bayesian Artificial Intelligence.

CRC Press.  2011. Prado, M.L. Advances in Financial Machine Learning.
Wiley. 2018.
Forma de Avaliação
• 1º. Bimestre
• 1 Prova [+ Lista de exercícios]
• 2º. Bimestre
• Proposta de Projeto [+ Lista de exercícios]
• Exame
• Projeto: Artigo + Apresentação
Plan
1. Introduction to AI
2. Rational Decisions
3. Probabislistic Reasoning
4. Decision Making with Bayesian Networks
5. Learning Probabilistic Models and Knowledge Engineering
6. Introduction to Practical Machine Learning
7. Evaluation of ML models, Overfitting and Ensemble Methods
8. Dynamic Bayesian Networks
9. Markov Decision Process
10. Reinforcement Learning
11. Applications of Artificial Intelligence
12. Trends in AI
Definitions of Artificial Intelligence
Thinking Humanly
“The exciting new effort to make computers
think . . . machines with minds, in the
full and literal sense.” (Haugeland, 1985)
“[The automation of] activities that we
associate with human thinking, activities
such as decision-making, problem solving,
learning . . .” (Bellman, 1978)
Acting Humanly
The art of creating machines that perform
functions that require intelligence
when performed by people.” (Kurzweil,
1990)
“The study of how to make computers do
things at which, at the moment, people are
better.” (Rich and Knight, 1991)
Thinking Rationally
“The study of mental faculties through the
use of computational models.”
(Charniak and McDermott, 1985)
“The study of the computations that make
it possible to perceive, reason, and act.”
(Winston, 1992)
Acting Rationally
“Computational Intelligence is the study
of the design of intelligent agents.” (Poole
et al., 1998)
“AI . . . is concerned with intelligent behavior
in artifacts.” (Nilsson, 1998)
Definições de Inteligência Artificial
Pensando como seres humanos Pensando Racionalmente
“O novo e interessante esforço para fazer os “O estudo das faculdades mentais pelo uso de
computadores pensarem (…) máquinas com mentes, modelos computacionais” (Charniak e McDermoot,
no sentido total e literal” (Haugeland, 1985) 1985)

“[Automação de] atividades que associamos ao

pensamento humano, atvidades como a tomada de
decisões, a resolução de problemas, o aprendizado..” “O estudo das computações que tornam possível
(Bellman, 1978) perceber, raciocinar e agir” (Winston, 1992)

Agindo como Seres Humanos Agindo Racionalmente

“A arte de criar máquinas que executam funções “Inteligência Computacional é o estudo do projeto
que exigem inteligência quando executadas por de agentes inteligentes” (Poole et al. 1998)
pessoas” (Kurzweill, 1990)

“IA…está relacionada a um desempenho inteligente

“O estudo de como os computadores podem fazer de artefatos” (Nilsson, 1998)
tarefas que hoje são melhor desempenhadas por
pessoas” (Rich and Knight, 1991)
Artificial Intelligence – Is it new ?
• The term “Artificial Intelligence” was used for the first
time in the summer of 1956, in an invitation to a 2-
month workshop organized by John McCarthy, Marvin
Minsky, Claude Shannon and others..…
Artificial Intelligence -birth certificate
• “We propose that a 2 month, 10 man study of artificial intelligence be
carried out during the summer of 1956 at Dartmouth College in
Hanover, New Hampshire. The study is to proceed on the basis of the
conjecture that every aspect of learning or any other feature of
intelligence can in principle be so precisely described that a machine
can be made to simulate it. An attempt will be made to find how to
make machines use language, form abstractions and concepts, solve
kinds of problems now reserved for humans, and improve
themselves.”John McCarthy, Marvin Minsky, Claude Shannon et al.1956

• Perhaps “computational rationality” would have been more precise and

less threatening, but “AI” has stuck....
• McCarthy stated that he resisted the terms “computer” or
“computational” in deference to Norbert Weiner, who was promoting
analog cybernetic devices rather than digital computers
What about …. ?
• Deep learning
• “Deep learning is a subset of a more general field of artificial
intelligence called machine learning” Buduma, N. The fundamentals
of deep learning.
• Machine Learning
• Construção de software “...que pode melhorar seu próprio
comportamento através do estudo diligente de suas próprias
experiências” (Russel, Norvig, 2013)
• Data mining
• Finding patterns in data that provide insight or enable fast
and accurate decision making (Witten,2016) Data Mining:
Practical Machine learning)
• Big data
• Capturing and managing lots of information (computer systems)
• Analyzing these masses of new data (data mining)
Machine learning
• Definitions of “learning” from dictionary:
• To get knowledge of by study, experience, or being taught

• Operational definition:
• Things learn when they change their behavior in a way that
makes them perform better in the future.
• Machine Learning vs Statistcs
• “In truth, you should not look for a dividing line between
machine learning and statistics because there is a
continuum— and a multidimensional one at that—of data
analysis techniques” Witten, Data Mining: Practical Machine
learning
• In fact, it could be also stated to Machine Learning vs Data
mining vs Statistics
A “Reasonable” Graph Representation of
Intersections of Related Areas to AI
Challenges of AI

• Problem Solving
• Solving by Searching
• Constraint Satisfaction problems
• Knowledge, reasoning and planning
• Logical Agents
• Planning and Knowledge Representation
• Uncertain knowledge and reasoning
• Probabilistic reasoning
• Probabilistic reasoning over time
• Learning
• Supervised, (semi-supervised), unsupervised and reinforcement
learning
• Communications,perceiving and acting
• Natural language processing (NLP)
• Sensors
• Robotics
New technologies and the Gartner Hype Cycle
AI: A multidisciplinary field
And What about this discipline?
Problem Solving
• Solving by Searching
• Constraint Satisfaction problems
Knowledge, reasoning and planning
• Logical Agents
• Planning and Knowledge Representation

Learning
• Supervised, (semi-supervised), unsupervised and reinforcement
learning
Communications,perceiving and acting
• Natural language processing (NLP)
• Sensors
• Robotics
Can a Machine Think?,Turing, A. (1950)
Section 1: The imitation game
• How to define machine...and think ?
• Too complex...
• Another approach: The Imitation Game

• Description:
• It is played with three people, a man (A), a woman (B), and an interrogator (C)
who may be of either gender.
• The interrogator stays in a room apart front the other two.
• The object of the game for the interrogator is to determine which of the other
two is the man and which is the woman.
• He knows them by labels X and Y, and at the end of the game he says either
“X is A and Y is B” or “X is B and Y is A.”
• Proposed question: “What will happen when a machine takes the
role of person A in this game?”
• Will the interrogator decide wrongly as often when the game is
played like as he does when the game is played between a man and
a woman?
• These questions replace our original, “Can machine think?”
Can a Machine Think?, Turing, A. (1950)

• Objections:
• The Theological Objection
• The "Heads in the Sand" Objection
• The Mathematical Objection

• The Argument from Consciousness

• And others….
Head in the Sand Objections...or a real
disasters ahead?
• Some Debates on World Economic Forum - Davos 2020
• Humans behind Machines
• How should labour policies adapt to the unique conditions of “digital
assembly lines”?
• How to Implement Responsible AI
• Is it possible to build trust in artificial intelligence?

• A Future Shaped by a Technology Arms Race

Head in the Sand Objections or a real
disasters ahead - ... What about jobs ?
• “Up to 375 million people globally will need to find entirely new
kinds of jobs by 2030 because of automation,...How workers,
governments and businesses tackle these seismeic shits will define
the future of Work”
• American Factory (Oscar winner Documentary), 2019, Netflix

• Dara Khosrowshahi, Uber CEO, speaks with David M. Rubenstein on

June 11, 2019 about Driverless cars
• “..I think within the next five years some driverless vehicles will be out in the
market..starting in a very very limited way”

• “We will get the machines to do the simple stuff and the human do the
difficult stuff..”

• “The two are going to coexist for 10 to 15 years.....”

Head in the sand and The future of AI

• Holywood (Matrix, Terminator..) Hypothesis

• In our humble opinion, it is not going to happen in our lifetimes..or
ever.

• Massive Unemployment Hypothesis

• Well...what do you think?

• Let's just assume that keeping the head in the sand is not
going to solve any problems...
The concept of Agent

• An agent is anything that can be viewed as perceiving its

environment through sensors and acting upon that
environment through actuators. Examples:?
Reflex Agent (Agente reativo)

• A Reflex Agent is based on

condition–action rule (a.k.a
if-then rules)

• Does it learn??

• Does it adapt itself?

• Does it think about things?

• More complex agents are

required...
Model based agent (Agente baseado
em Modelo)

• It keeps track of the

current state of the world,
using an internal model.

• It then chooses an action

in the same way as the
reflex agent
Goal based Agent (Agente baseado
em Objetivos)
• It keeps track of the world
state as well as a set of
goals it is trying to achieve,

• and chooses an action

that will (eventually) lead
to the achievement of its
goals.
Utility-based agent Agent

• It uses a model of the

world, along with a utility
function that measures its
preferences among states
of the world.
• Then it chooses the action
that leads to the best
expected utility,
• where expected utility is
computed by averaging
over all possible outcome
states, weighted by the
probability of the outcome
General Learning Agent

• A learning agent can be divided into four

conceptual components

• Learning element is responsible for making

improvements,
• Performance element is responsible for
selecting external actions
• The performance element is what we have
previously considered to be the entire agent: it
takes in percepts and decides on actions
• Critic gives info to the Learning element on
how the agent is doing, so the Learning
element decides how the Performance
element should be modified to do better in
the future

• Problem Generator is responsible for

suggesting actions that will lead to new and
informative experiences. Why??
Examples of Agents and its key
features
Environment - where the agent perceives and
acts - and its properties
• Fully observable vs. partially observable (Observável x Parcialmente
Observável)

• Deterministic vs. stochastic (Determinístico x Estocástico)

• Episodic vs. sequential (Episódico x Seqüencial)

• Static vs. dynamic (Estático x Dinâmico)

• Discrete vs. continuous (Discreto x Contínuo)

• Single agent vs. multiagent (Agente Único x Multiagente)

Environment - where the agent
perceives and acts
• Fully observable vs. partially observable
• If an agent’s sensors give it access to the complete state of the environment at each point in time, then
we say that the task environment is fully observable. A task environment is effectively fully observable if
the sensors detect all aspects that are relevant to the choice of action; relevance, in turn, depends on
the performance measure.
• Deterministic vs. stochastic
• If the next state of the environment is completely determined by the current state and the action
executed by the agent, then we say the environment is deterministic; otherwise, it is stochastic

• Episodic vs. sequential

• In an episodic task environment, the agent’s experience is divided into atomic episodes.
• In each episode the agent receives a percept and then performs a single action. Crucially, the next
episode does not depend on the actions taken in previous episodes

• Static vs. dynamic

• If the environment can change while an agent is deliberating, then we say the environment is dynamic
for that agent; otherwise, it is static. If the environment itself does not change with the passage of time
but the agent’s performance score does, then we say the environment is semidynamic

• The differences between continuos vs discrete and single agent vs multiagent are pretty
straightforward, but multiagent means that the existence of other agents matter to agent's
performance
What does make the world Uncertainty
(Partially observed or stochastic) ?
1. Ignorance. The limits of our knowledge lead us to be uncertain
about many things. Does our poker opponent have a flush or
is she bluffing?

2. Phyical randomness or indeterminism. Even if we know

everything that we might care about a coin and how we impart
spin to it when we toss it., there will remain an inescapable
degree of uncertainty about it wil land heads or tails
• A strong deterministr person might claim otherwise, that it would be
possible to calculate......but such a view is for the foreseeable future a
mere act of scientistic faith. We are all practical indeterminists

3. Vagueness. Many of predicates we employ appear to be vague.

It is often unclear whether to classify a bird as big or small, a
human as brave or not, a thought as knowledge or opinion
First Exercise (get paper and pen...)

• What about Chess without a clock? changes anything?