Chapter 14: Artificial Intelligence IS-300

Asela Thomason

What is AI?

technology that allows machines to stimulate human intelligence such as learning, reasoning, and decision-making

Sebastian Thrun Ted Talk

• Main takeaway: AI learns from data rather than being explicitly programmed, making it highly effective for narrow tasks but not capable of human-like thinking or general intelligence

Sam Harris Ted Talk

main takeaway: AI requires global cooperation and governance

Yoshua Bengio Ted Talk

• main takeaway: uncontrolled AI poses catastrophic risks

Classification by Intelligence

• Narrow AI: task-specific (face recognition, Siri)

• General AI: human-like intelligence to perform any intellectual task (does not yet exist)

Classification by Task

• Prediction: using historical data to forecast future outcomes

• Optimization: best solution

• Anomaly: fraud detection

• Creativity: art, music, writing

Prediction

Using historical data to forecast future outcomes

ex: predicting customer churn, stock price forecasting, weather prediction

Optimization

finding the best possible solution among many options

ex: finding fastest delivery route, optimizing supply chain operations, scheduling airline flights

Anomaly detection

AI identifies unusual or abnormal patterns in data

ex: credit card fraud detection, network security breaches, equipment failure detection

Creativity

AI generates new content such as text, images, music, or designs

ex: ChatGpt writing text, DALLE generating images, AI music composition

Classification of AI machines

• reactive machines

• limited memory machines

• theory of mind machines

• self-awareness machines

Reactive Machines

• simplest form of AI

• reacts only to current input and has no memory

• ex: IBM’s deep blue chess computer

• no learning and no memory

Limited Memory machines

• AI that can use past data or experiences to make decisions

• ex: self-driving cars, recommendation systems

• learns from recent data

Theory of Mind Machines

• AI that can understand human emotions, intentions, or beliefs

• ex: advanced chatbots (theoretical/experimental)

• understands how humans think and feel

Self-Aware Machines

• AI that has consciousness and self-awareness

• hypothetical future AI

Degree of Autonomy

• Fully autonomous AI: operates independently, handling complex multi-step tasks end-to-end without human intervention (Level 5 self-driving car or advanced AI agents setting their own goals)

• Partially autonomous AI: performs specific functions but requires human oversight for critical decisions, handling routine tasks, working with predefined boundaries and often needing human input to manage risk or complex scenarios

OpenAI

mission: build safe and beneficial AI

operates as a non-profit

original founders: Sam Altman & Greg Brockman

products include: ChatGPT3, ChatGPT4, ChatGPT5 & Bring Chat GPT

Prompts and Hallucinations

• prompts guide AI output

• hallucinations occur when AI generates incorrect or false information

AI technologies

• expert systems

• machine learning

• neural networks

• deep learning

Expert Systems

• AI systems that mimic the decision-making ability of a human expert by applying stored knowledge and rules within a specific domain to solve complex problems or support-decision making

• can integrate/manipulate huge amount of data —> sometimes performs better than a single human expert can

• do present problems: transferring domain expertise from human—> system cab be difficult because people cannot always explain what they know

• therefore automating that process may not be possible

• in some context can be potential liability

Machine Learning

• ability to accurately perform new, unseen tasks, built on known properties learned form training or historical data that are labeled

• ex: banking industry, oil/gas industry, life sciences industry, retail industry

Deep learning

• subset of machine learning in which system discover new patterns without being exposed to labeled historical or training data

• ex: speech recognition, image recognition, natural language processing, drug discovery and toxicology, customer relationship management

Neural Networks

• a set of virtual neurons or central processing units (CPUs) that work in parallel in an attempt to simulate the way the human brain works, although in a greatly simplified form

AI in use today

natural language processing, computer vision, intelligent agents, chatbots

Computer Vision

• ability of information systems to identify objects, senses, and activities in images

• ex: medical imaging, facial recognition, shopping, self-driving cars, pinterest, google photo app, microsoft’s traffic prediction project

Intelligent agents

• software programs that imitate humans and perform tasks on command

• the agent uses a limited built-in or learned knowledge base

• accomplish tasks or make decisions on the user’s behalf (routine calls in a call center)

types of intelligent agents

1. Information agents

• Role: search for information and display it to its users

• Ex: buyer agents that help customers find products or services

2. Monitoring and Surveillance Agents (Predictive agents)

• Role: Constantly observe and report on items of interest

• Ex: stock market monitoring tools, network security monitors

3. User Agents (Personal Agents)

• Role: Take action on your behalf

• Ex: email assistants, calendar schedulers, voice assistants like Siri or Alexa