BTE 210 Exam 2 PT. 1

Artificial Intelligence

technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity, and autonomy

1950s

Turing Test

Turing Test

evaluates machine intelligence by whether speaking to human or machine

1956

Dartmouth Conference

Dartmouth Conference

John McCarthy; discuss potential of machines to simulate human intelligence

1960s-1970s

Symbolic AI and Expert Systems

Expert Systems

Solve specific problems using rule-based approaches that ordinarily requires human expertise

DENDRAL

first expert system to analyze chemical compounds

Knowledge Base Expert System

Facts and Rules i.e. medical info

Inference Engine Expert System

Computer program that provides a methodology to formulate reasoning and conclusions i.e. applies rules and generate diagnosis

User Interface Expert System

Dialogue between the user and the computer i.e. answering questions

Benefits of Expert Systems (3)

Enhance decision making and problem solving, decrease time, reduced downtime

Difficulties of Expert Systems (3)

Transferring Domain Expertise, difficulty automating processes, potential liability

1980s

AI Winter and Neural Networks

1980-1990s

Resurgence and Machine Learning

Machine Learning

algorithms and models to learn from data and make predictions or decisions without explicit programming

Deep Blue

Defeat chess champion

Machine Learning Key Steps

Data Preparation, Model Training, Model Evaluation

Data Preparation

Collect and clean your data for suitable format

Model Training

Choose appropriate algorithm and train model

Model Evaulation

Test the model using testing data to evaluate performance

Machine Learning 3 primary approaches

Supervised, Unsupervised, Reinforcement Learning

Supervised

Uses labeled data to predict outputs i.e. spam email detection

Unsupervised

Unlabeled data to find natural grouping and patterns i.e. spotify

Reinforcement Learning

Interaction and feedback to maximize rewards and minimize penalties i.e. self-driving cars

Machine Learning System Examples

Optical character recognition, topic id, fraud detection, customer segmentation

2000-2010

Big Data and Deep Learning

CNNs and RNNs

revolutionized image and speech recognition (Siri), NLP, Autonomous vehicles

Neural Networks

simulate the complex decision-making processes of human brain

NLP

Natural Language Processing (chatbots, translators)

2010

AI in Everyday Life

AI in Everyday Life

Virtual assistant (Siri), streaming service recommendations, GPT-3

2020

Ethical AI and Generative AI

Generative AI

Subset of Deep Learning focused on creating new content

GenAI Examples

GPT-5, DeepSeek, Claude, DALL-E

Deep Learning Architecture example

neural networks

Generative AI Prompting Techniques

Zero-shot, one-shot, few-shot, role, prompt chaining

Grounding

Connecting the AI’s output to veribfiable source info

RAG

retrieval-augmented generation plus iteration

Sampling Parameters

Token Count, Temperature, Top-p, Safety settings, Output length

Token Count

meaningful chunks of text (each word)

Temperature

Controls the creativity or randomness of the models word choices

Top-p (nucleus sampling)

Probability AI can wander from most likely tokens

Safety settings

filter potentially harmful or inappropriate responses

Output Length

maximum length of generated text

Agent

application that tries to achieve a goal

Agent Types

Deterministic, generative, hybrid

Deterministic

traditional; predefined paths and actions

Generative

Natural languavge using LLM for conversation

Agent components

Reasoning loop, tools, model

Reasoning Loop

Iterative process where agent observes, interpretsm reasons, and acts

Tools

allow the agent to interact with its environment (access data/hardware)

Model

Brains; various algorithims that learn patters, make predicions, generate new content

How do they fit together?

GenAI, Deep Learning, Machine Learning, AI

Benefits of AI

Efficiency and automation, data analysis, 24/7 availability, error reduction, scalability

Challenges of AI

Misinformation and deepfakes, data privacy, bias and fairness, job displacement, transparency, technical complexity, accountability