Ch. 15 - Emerging Technologies: Blockchains, AI and Process Automation

Blockchain

A decentralized, distributed ledger system that records transactions in blocks that are chained together

Smart Contracts

Programmable code (often on Ethereum) that pulls together logic and business rules to execute agreements automatically

Immutability

A core characteristic where transactions, once confirmed on a blockchain, are tamper-proof and cannot be altered or deleted

Consensus

The requirement that all parties on a network are aware of and agree to transactions before they are written to the blockchain

Proof of Work (PoW)

A consensus algorithm where miners solve complex mathematical problems using significant computer power to create new blocks.

Proof of Stake (PoS)

A consensus method where validators lock up cryptocurrency as a deposit to ensure honest behavior, reducing computer costs

Artificial Intelligence (AI)

Intelligence exhibited by machines that perform tasks associated with human intelligence, like logical thinking or speech recognition

Machine Learning (ML)

A type of AI where a computer learns from experience and data patterns rather than specific instructions

Neural Networks

Mathematical models that convert inputs to outputs or predictions, acting as the engine for machine learning

Deep Learning

A form of machine learning using complex, multilayer neural networks (more than two non-output layers)

What are the three main types of blockchain?

Public, Private, Consortium

Types of Blockchain: Public

permissionless, anyone can join

Types of Blockchain: Private

permissioned, internal to an enterprise

Types of Blockchain: Consortium

permissioned, allows several organizations to participate

What makes blockchain faster and cheaper than traditional systems?

It eliminates the middleman, reduces transaction delay from days to minutes, and lowers service fees

Bitcoin

a cryptocurrency for peer-to-peer transactions

Ethereum

supports "Smart Contracts" and is much more programmable for business rules

What are the four families of AI algorithms?

Two-Class Classification, Anomaly Detection, Regression, Clustering

Questions AI algorithms answer: Two-class classification

Is it A or B?

Questions AI algorithms answer: Anomaly Detection

Is it different?

Questions AI algorithms answer: Regression

How much?

Questions AI algorithms answer: Clustering

How are these alike?

What is the "Virtuous Cycle" of Machine Learning?

Models learn from data => make predictions => generate new data from those products => use that new data to improve future predictions

Supervised Learning

uses labeled input-output pairs to predict known values

Unsupervised Learning

uses unstructured data to find patterns or outliers without predefined labels

How does blockchain impact audit and assurance?

It makes "Continuous Audit" possible through real-time information retrieval, but requires auditors to learn skills like auditing smart contracts

What are two specific AI applications in accounting?

Natural Language Processing and Robotic Process Automation

Natural Language Processing (NLP)

A technology used to understand text, extract semantic meaning, and discern sentiment from human language

Example: An AI tool that reads thousands of legal contracts to flag specific clauses for an auditor

Robotic Process Automation (RPA)

A software tool designed to perform high-volume, repetitive accounting tasks that were previously done by humans'

Example: A "bot" that automatically logs into a portal, downloads an invoice, and enters the data into QuickBooks every morning.