You are on page 1of 19

Week #2

(Lecture 3)
Types of AI systems

Artificial Intelligence &


Intelligent Systems

Slides By Dr.Rami Ibrahim


What is AI?
 AI is the science of making machines intelligent.
 Weak AI: Machines can perform some specific tasks (e.g., Apple
Siri, Amazon Alexa).
 Strong AI: Machines and humans are equal. Machines have the
ability to solve problems and plan for future. (Superintelligence).
 The AI technologies nowadays belong to weak AI but can
improve over time.
The Turing test

• “Turing is widely considered to be the father of


theoretical computer science and artificial
intelligence”

• https://www.youtube.com/watch?v=4VROUIAF2Do&
ab_channel=EyeonTech

• https://medium.com/@jetnew/a-summary-of-alan-
m-turings-computing-machinery-and-intelligence-
fd714d187c0b
The Chinese room problem
”The Chinese room is the scene for an argument originated by
John Searle to show that a symbol-processing machine
(a computer) can never be properly described as having a
mind or understanding or being conscious, no matter how
intelligently it may behave.”

”Running a computer program does not generate


understanding.”

https://www.youtube.com/watch?v=D0MD4sRHj1M&ab_chann
el=BBCStudios
Top-Down AI

It is probably easiest to imagine ‘Top-down AI’ to be


based on a decision tree. For example, a call center
chatbot is based on a defined set of options and,
depending on the user input, it guides the caller through
a tree of options. What we typically refer to as AI these
days.
Top-down approach can be very useful for some tasks
that machines can be programmed to do, like the call
center example above.
Bottom-Up AI

Bottom-up basically defines machine learning.


The system is given data to learn from, and it
uses that information from the past to predict
and make decisions for the future.
for applications such as self-driving cars or
diagnostic systems in health care — would be
defined as ‘Bottom-up AI’ and is based on
machine learning (ML) or deep learning (DL).
AI Systems Types

Top-down Bottom-up

knowledge-based
systems Neural Networks

Natural Language Evolutionary


Processing (NLP) Computing

Fuzzy Logic Swarm Intelligence


Explainable AI (XAI)

 Explainability: Improving the AI system’s explanations


by building an interface between human and AI
systems.
 Interpretability: The extraction of relevant knowledge
that is appropriate for different stakeholders (end user,
ML developer, decision makers).

CAT 85%
CNN
DOG 15%
Convolutional Neural Network (CNN)
Neural Networks Black Box

Input Hidden Output


Layer Layer Layer
Kernels

Conv1
Pool1

Pool2
CAT
Conv2 Pool3
Pool4
Pool5
Conv3
DOG

Conv4

Conv5

Pooling 9
Flatten
Conv. Layer Layer
Explainable AI (XAI)
Providing explanations in AI

10
Source: https://www.darpa.mil/program/explainable-artificial-intelligence
Terminologies
 Machine Learning and Deep
Learning are subsets of AI.
 Machine Learning: Enable the
machine to learn by itself (training)
to make future predictions.
 Deep Learning: The same as
Machine Learning, however it
mimics human brain by applying
neural networks.
AI vs. ML vs. DL
AI vs. ML vs. DL
Machine Learning vs. Deep Learning
 
Deep Learning Machine Learning
Architecture Neural network (Input layer, hidden layers, output Input layer, model, output layer

layer)

Learning Learns incrementally, start from high-level features Domain expert selects features of the model

Features to low-level features

Infrastructure Require massive amount of labeled data and Require less sophisticated infrastructure

sophisticated processors

Training Model Requires long time for training the model (tuning Require less time for training the model since

hyperparameters) parameters are initialized

Performance High accuracy in tasks like face detection, object High accuracy in tasks like classification and

recognition clustering

Interpretability Hard to interpret how the results are produced Easy to interpret how results are produced in
What is Machine Learning
 The science (and art) of programming computers
so they can learn from data.
 The field of study that gives computers the ability
to learn without being explicitly programmed.
Traditional Approach vs. ML
Types of ML
 Supervised Learning: Use of labeled data to train the
algorithm before predicting new data.
 Example: Classification & Regression.
Types of ML
 Unsupervised Learning: Analyze, cluster, detect
patterns of unlabeled data.
 Example: Clustering, Association Rule Mining.
References
Chapter 1 ‘Training Models’ from Hands-On Machine Learning
with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and
Techniques to Build Intelligent Systems’ 2nd Edition

You might also like