Scribd
Upload
16 views

Lecture 2 - AI Building blocks

The document outlines the building blocks of artificial intelligence, including structured, unstructured, and semi-structured data, as well as the preprocessing techniques necessary for AI applications. It discusses various AI techniques such as natural language understanding, computer vision, reasoning, problem-solving, and machine learning. Additionally, it emphasizes the importance of knowledge representation and learning methods in AI development.

Uploaded by

Nga Phạm
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
16 views

Lecture 2 - AI Building blocks

The document outlines the building blocks of artificial intelligence, including structured, unstructured, and semi-structured data, as well as the preprocessing techniques necessary for AI applications. It discusses various AI techniques such as natural language understanding, computer vision, reasoning, problem-solving, and machine learning. Additionally, it emphasizes the importance of knowledge representation and learning methods in AI development.

Uploaded by

Nga Phạm
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 30

Artificial Intelligence

Lecture 2 - AI Building blocks

Lecturer: Vu Trong Sinh


Contents
1.5. AI Building blocks

1.6. AI techniques
1.5. AI Building blocks
Artificial intelligence: Building blocks and an
innovation typology, Ulrich Paschen (2020)
https://www.sciencedirect.com/science/articl
e/abs/pii/S000768131930151X
https://www.diva-portal.org/smash/get/diva2:
1400547/FULLTEXT01.pdf
Structured data
Data that are standardized and organized
according to predefined schema.
E.g.:
- customer demographics, web
browsing data or transaction data
(internal data)
- social media ratings or stock exchange
transactions (external data)
Structured data
Data sources:
- System database
- Excel worksheet
- Google sheet
- Table data from websites

The heart of business analytics


and business intelligence
Unstructured data
Data that are not standardized or organized according to a pre-defined schema

Unstructured data is getting bigger and bigger


IoT, social media and mobile devices are popular, they produce endless flow of
digital data that are mostly unstructured
E.g.: human language in written form, such as blogs, posts, reviews, comments, or
tweets; speech, such as audio in user-generated content, and images that portray
objects or people
Unstructured data – sources and types
Any kind of
semi-structured
data?
Semi-structured data
Semi-structured data is a type of data that is not purely structured, but also not
completely unstructured. It contains some level of organization or structure, but
does not conform to a rigid schema or data model, and may contain elements that
are not easily categorized or classified.

E.g.: unstructured data organized in an CSV/XML/JSON file


Quiz
Which of the following is unstructured/structured/semi-structured data?
- https://www.kaggle.com/datasets/deependraverma13/cardio-activities
- https://www.kaggle.com/datasets/sagaraiarchitect/laptop-price-explorer-the-
ml-model
- https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movi
e-reviews
- https://www.kaggle.com/datasets/wardaddy24/marble-surface-anomaly-dete
ction
- https://www.kaggle.com/datasets/zalando-research/fashionmnist
- https://www.kaggle.com/datasets/ronikdedhia/next-word-prediction
Preprocess
Pre-processing includes data cleaning, normalization, transformation, feature
extraction and selection, with the goal that the remaining data can be processed in
value-creating ways.

Preprocessing of unstructured data in their various forms is more challenging due to


their complexity and capacity
Preprocess – Natural language understanding
Artificial intelligence uses natural language understanding (NLU) to assign meaning
to the vast and complicated human language in spoken and written form.

2 forms of natural language: text and acoustic signal


Preprocess – Natural language understanding levels

Lexical Syntactic Semantic Discourse

• Tokenizing. • Morphology • Word sense • Language


• Word analysis disambiguation affection
segmentation • Part of Speech • Language
(POS) Tagging Modeling
Preprocess – Natural language understanding
Natural language understanding tasks
Preprocess – Natural language understanding
Before AI can make sense of spoken language, speech first needs to be transcribed
into text; this step is typically referred to as speech recognition

Speech recognition allows AI to recognize the words that were said, but not what
the words mean
Preprocess – Computer vision
Computer vision is the transformation of visual images into internal representations
of the world so that these representations can interface with other building blocks in
AI

The degree of sophistication in computer vision varies widely, from recognizing


edges or texture to boundaries, surfaces, volumes to the classification of objects,
scenes or events
Preprocess – several CV tasks
Main processes
One of the key processes of intelligence is the ability to apply logic to solve
problems and learn.

Learning is the process of acquiring new or modifying existing knowledge to better


achieve desired outcomes.
Main processes – Reasoning
Reasoning refers to applying logic to generate conclusions from available data

Example:

Premise-1: All the human eats veggies

Premise-2: Suresh is human.

Conclusion: Suresh eats veggies.


https://www.javatpoint.com/reasoning-in-artificial-intelligence
Main processes – Problem solving
Problem solving involves choosing the best solution from a range of alternatives for
reaching a goal.

The process of problem-solving is frequently used to achieve objectives or resolve


particular situations. In computer science, the term "problem-solving" refers to
artificial intelligence methods, which may include formulating ensuring appropriate,
using algorithms, and conducting root-cause analyses that identify reasonable
solutions

Problem solving techniques:


● Heuristics
● Searching Algorithms
● Genetic Algorithms
Main processes – Machine learning
Machine learning (ML) encompasses techniques that enables computers to learn
from experience, i.e., progressively improve their performance, without an explicit,
pre-defined set of rules that are stored in memory
Information output
Information results from data being placed into a formative context so that meaning
emerges

Simple form: Prediction results → API output

Complex form:
Natural Language Generation (NLG). E.g.: ChatGPT, Google Translate
Image Generation. E.g.: Stable Diffusion
Robotics
Knowledge base
A knowledge base stores digital representations of aspects of the real world in which
these representations operate, for later access
Discussion
- Take an AI application (e.g. ChatGPT, self-driving car, ...) and analyze its building
blocks

- Which technologies needed for each block


1.6. AI techniques
Machines are not human, we can not give them a book and tell them to learn

→ Represent the input/output/processing so that they can understand

→ Teach them how to compute/process/perform actions/…

→ When should follow the rule, when should make the decision by
themselves

→ Is the solution the best / Is it possible to improve?

25
Techniques used in AI
Representation (knowledge representation):
Computer can not understand Vietnamese, can not watch Youtube tutorial

They only understand 0 and 1

→ Anything must be converted to 0 and 1? → No, we have Programming


Language & Data Structure

→ Define the problem following the programming language, convert the data
(numerical, categorical, text, images, videos, …) to suitable data structures

26
Techniques used in AI
Learning:

We even simulate how our brains work by programming language

(Neural Networks - Chapter 6, 7)

● Automatically build up the knowledge from the environment


● Two main strategies:
○ Acquiring the rules in a ruled-based system, or
○ Formulating the neurons in a neural network

27
Convolutional Neural Network (Chapter 6, 7) - which animal is in the image?
- Take an image as input
- Break into pixels
- Convert pixels to numbers, vectors, matrices
- Some calculations ^^
- Produce the output: which animals got the highest score of probability

28
Techniques used in AI
Rules:

Explicit: created by a human expert

Implicit: automatically obtained through learning

Search:

searching the sequence of states that lead to solution faster, or searching for
an optimum set of connection weights in a neural network

29
Homework
Join this competition on Kaggle:
https://www.kaggle.com/competitions/house-prices-advanced-regression-techniqu
es/overview

Create your own notebook and make a submission

You might also like