3.

6 - Artificial
Intelligence
3.6.A. Types of AI
 Index
● Types of AI
● Types of Machine Learning
● Deep Learning
● Neural Network
● Evaluation of AI
● AI Dilemmas
TURING TEST
The Turing Test is a deceptively simple
method of determining whether a
machine can demonstrate human
intelligence: If a machine can engage
in a conversation with a human without
being detected as a machine, it has
demonstrated human intelligence.
(1950)

AI21 Labs - Test

CAPTCHA (Completely Automated Public Turing Test to tell Computers and Humans Apart)
The reasons for using CAPTCHA by web sites are:
Spam Protection: CAPTCHA helps protect websites from automated bots that might attempt to submit forms, create accounts, or post spam
content. By requiring users to solve a CAPTCHA, websites can ensure that the entity interacting with the site is likely a human.

Bot Mitigation: Bots can be programmed to perform various actions on websites, such as scraping data, launching DDoS attacks, or attempting to
exploit vulnerabilities. CAPTCHA serves as a barrier to these automated processes, preventing malicious bots from carrying out unwanted activities.

Security Measures: In scenarios where security is crucial, such as during login processes or when making sensitive changes to user accounts,
CAPTCHA can add an additional layer of authentication. This helps ensure that only legitimate users, rather than automated scripts, are accessing
sensitive areas of a website.

DDoS Attack Prevention: During Distributed Denial of Service (DDoS) attacks, where a website is flooded with traffic to overwhelm its servers,
CAPTCHA can be used to differentiate between legitimate users and bot traffic. This can help mitigate the impact of DDoS attacks by distinguishing
real users from automated requests.

User Account Protection: CAPTCHA is commonly used to prevent automated scripts from attempting to guess passwords or gain unauthorized
access to user accounts. This helps protect user accounts from brute-force attacks.

Data Scraping Prevention: Websites often use CAPTCHA to prevent automated bots from scraping and extracting data from their pages. This is
especially important for sites that rely on proprietary or sensitive information.
What about ChatGPT?
Did ChatGPT passed the Turing test?
TYPES OF AI

Strong
AI
WEAK or NARROW AI

Weak AI has a limited function or can only perform a specific task.

Many weak artificial intelligence systems are changing the world we live in as they can complete tasks
much quicker than humans and can discover patterns in big data.
WEAK or NARROW AI - Examples
Email Spam Filters: Spam filters used by email service providers to protect users from unwanted emails are examples of
Weak AI. They perform a specific task (spam detection) without possessing general thinking capabilities.

Gaming Artificial Intelligence: Artificial intelligence used in computer games is typically designed to compete against
players within specific games. These AI systems can perform well in the context of the game but lack general thinking
abilities.

Recommendation Algorithms (Netflix, Spotify): Recommendation algorithms on platforms like Netflix or Spotify, which
suggest content based on user preferences, are examples of Weak AI. These systems focus on the specific task of
content recommendation.

Voice Assistants (e.g., Siri or Google Assistant): Voice assistants are examples of weak artificial intelligence designed to
perform specific tasks. They concentrate on limited tasks such as understanding user-spoken commands and performing
specific tasks like setting reminders or providing weather information.

Image Recognition System: An image recognition system is one of the examples of weak artificial intelligence designed
to detect specific objects or patterns. For example, it can be used by security cameras to detect specific movements and
trigger alarm situations.
DOMAIN SPECIFIC
Domain Specific AI AI
In certain domains, artificial intelligence can perform tasks better than humans. This is known
as domain-specific AI. Examples include artificial intelligence that can detect cancer better than
human doctors, and artificial intelligence such AlphaGo, developed by DeepMind to beat the world
champion at the board game Go.
DOMAIN SPECIFIC
Domain Specific AI AI - Examples
Autonomous Driving Systems: Artificial intelligence systems used in autonomous vehicles are specifically designed to
manage driving and traffic. These systems focus on the domain of driving, making them examples of Domain Specific AI.

Patient Monitoring Systems: Patient monitoring systems used in hospital environments continuously monitor vital signs to
assess specific health conditions. These systems are examples of Domain Specific AI tailored to the needs of the
healthcare sector.

Medical Image Analysis: A medical image analysis system can be a specially designed artificial intelligence application.
This system may focus on detecting specific diseases by analyzing X-ray or MRI images. This is an example of Domain
Specific AI developed for the specific needs of the medical field.

Financial Fraud Detection: An artificial intelligence system used in the financial sector can detect signs of fraud by
analyzing customer transactions and financial data. This is an example of Domain Specific AI specifically designed for a
task unique to the financial industry.
DOMAIN SPECIFIC AI VS WEAK AI
The distinction between these concepts generally highlights how general or narrow the focus of an artificial
intelligence system is.

Domain Specific AI focuses on a particular domain (such as AI applications specialized in

medical image analysis, financial fraud detection, or natural language processing)

Weak AI generally refers to systems that concentrate on a limited task or set of tasks.
However, these terms can sometimes be used interchangeably, and their meanings may vary depending on the
context of use.
STRONG AI (full AI or artificial general intelligence (AGI))

Strong artificial intelligence will be when machines get to the point that they can develop consciousness
and make decisions better than humans. Additionally, strong artificial intelligence will be able to
demonstrate self-awareness and emotions, and have full human cognitive abilities.
https://divis.io/en/2019/03/ai-for-laymen-part-2-symbolic-ai-neural-networks-and-deep-learning/
SUPER
Super AIAI
Super AI, which is when machines surpass human intelligence. When artificial intelligence gets to this
point, there are concerns (such as those found in science fiction) that machines will overthrow the
human race. In technology, singularity describes the hypothetical future where AI becomes so
advanced that it is superior to human intelligence. Nanotechnology is perceived as one of the key
technologies that will make this a reality. https://twitter.com/elonmusk/status/1734763060244386074
3.6 - Artificial
Intelligence
Lesson2
 The Expert System
Expert systems in AI are computer programs that perform decision-making activities like human experts.

An expert system is an example of domain-specific AI and is frequently used in medical or car diagnosis,
petroleum engineering, financial advising and identifying unknown items.
The components of an expert system
An expert system has three main components:

The user interface, inference engine and the knowledge base.

● The knowledge base represents facts and rules and is created by a knowledge engineer who collects the
information provided by human experts.
● The inference engine uses programmed rules to interpret and evaluate the facts in the knowledge base. Logic is
applied in the inference engine using if/then statements, Boolean logic and fuzzy logic. To provide a solution or
answer, the inference engine uses forward or backward chaining.
● The user interface is the part of the system that allows a non-expert to ask questions of the expert system and
receive advice.
Key Terms
How is this different to machine learning, neural networks and deep learning?
 Machine Learning
Machine learning or ML is itself a sub-branch of

artificial intelligence. It involves collecting and

processing huge amounts of data, identifying patterns

and links within this data that wouldn’t immediately be

apparent to humans, and the use of those

observations in making decisions and improving the

operations of the system.

TYPES AND USES OF A MACHINE LEARNING
https://teachablemachine.withgoogle.com/
Types and uses of machine learning
● It needs a set of good data (often called training data).

● It uses algorithms (based on a computer model) to look at the data and understand it.

● Data is input into the algorithm to make a prediction.

● The prediction is then evaluated and, if inaccurate, this data is fed back into the

system to update the model/algorithm.

TYPES OF MACHINE LEARNING
● Supervised Learning

● Unsupervised Learning

● Reinforced Learning
Supervised Learning
A classification algorithm will take input data and assign it to a category that is based on the
training data.

A regression algorithm is different in that it is used for predictions and looks at the relationship
between variables. This algorithm is commonly used to make predictions, such as sales or test
scores, based on the analysis of the data set.

Applications that use supervised learning include spam filters, fraud detection and handwriting
character recognition.
Supervised learning examples
Supervised learning is effective for a variety of business purposes, including sales forecasting, inventory
optimization, and fraud detection. Some examples of use cases include:

● Predicting real estate prices

● Classifying whether bank transactions are fraudulent or not
● Finding disease risk factors
● Determining whether loan applicants are low-risk or high-risk
● Predicting the failure of industrial equipment's mechanical parts

https://www.coursera.org/articles/types-of-machine-learning
Unsupervised Learning
Unsupervised learning is used to find underlying patterns in data. It uses data sets that do not have
any labels but instead uses the data’s features. The goal of unsupervised learning is to analyse
data to find hidden patterns that cannot be noticed by a human observer. The algorithms use
cluster analysis, which looks for ways to group the data.

Applications that use unsupervised learning include social networking analysis and market
segmentation.
Unsupervised learning examples
Unsupervised algorithms are widely used to create predictive models. Common applications also include
clustering, which creates a model that groups objects together based on specific properties, and association,
which identifies the rules between the clusters. A few example use cases include:

● Creating customer groups based on purchase behavior

● Grouping inventory according to sales and/or manufacturing metrics
● Pinpointing associations in customer data (for example, customers who buy a specific style of handbag
might be interested in a specific style of shoe)

https://www.coursera.org/articles/types-of-machine-learning
Kahve içip kebap yiyen Ahmet
tahminen bu gece kaçta uyur?
Reinforced Learning
This type of machine learning is based on rewarding positive behaviour. The algorithm is trained
by rewarding the system if it performs correctly and gives a penalty if it performs
incorrectly. The software agent makes decisions to maximize its reward and keep adapting to the
environment. This form of machine learning requires a lot of data, which is why some of the first
applications were based on how readily available the data was.

Applications that use reinforced learning include autonomous vehicles, predictive maintenance,
games and robotics.
Reinforcement learning examples
Practical applications for this type of machine learning are still emerging. Some examples of uses include:

● Teaching cars to park themselves and drive autonomously

● Dynamically controlling traffic lights to reduce traffic jams
● Training robots to learn policies using raw video images as input that they can use to replicate the
actions they see

https://www.coursera.org/articles/types-of-machine-learning
Machine Learning Specialization (by DeepLearning.AI & Stanford University)
Machine Learning Specialization (by DeepLearning.AI & Stanford University)
Machine Learning Specialization (by DeepLearning.AI & Stanford University)
 Test
5 min.
Machine Learning
https://www.coursera.org/articles/types-of-machine-learning