MIS Unit 3 Test - Artificial Intelligence

artificial narrow intelligence (ANI)

AI designed for a single, specific purpose; systems operate within pre-defined contexts and cannot perform functions outside their specialization

artificial general intelligence (AGI)

AI with human-level cognitive abilities across various domains; represents a hypothetical form of AI that would possess human-level cognitive abilities across diverse domains

human-like cognition

demonstrate abstract reasoning, creative problem-solving, and the ability to transfer learning between disparate domains

generative AI

a type of ANI; its narrow intelligence is focused in the sphere of content creation

machine learning (ML)

software learning via experience rather than through explicitly programmed instructions; “experience” often looks like data

supervised learning

systems learn from data that includes both inputs and correct answers, enabling them to make predictions on new, unlabeled data

unsupervised learning

systems explore unlabeled data to discover structures, clusters, or patterns without being told what to look for

reinforcement learning

systems learn by performing actions in an environment and receiving rewards or penalties based on the outcome of those actions

neural networks

organize artificial neurons into interconnected layers

input layers

receives initial data such as pixel values from images or word tokens from text

hidden layers

one or more intermediate layers where pattern recognition and feature extraction occur

output layers

produces final results such as classification decisions or generated content

connections and weight

neurons connect to each other through pathways, each carrying a “weight”: a numerical value determining the strength and importance of the signal being transmitted

large language models (LLM)

massive neural networks with billions of parameters, allowing for complex pattern recognition

diffusion models

trained on huge numbers of images

training with noise

with each image shown, a diffusion model practices by taking a clear image and adding a bit of “noise” (static) to it

learning to denoise

then practices cleaning up the static to return to the original clear image → helps to learn the pattern of the image reconstruction

generating from static

the goal is to learn to create an accurate looking image from pure static

hallucinations

when a model generates information that is factually incorrect, nonsensical, or completely fabricated

bias

systemic distortion of information born of limited perspective

static knowledge base

AI models are trained on specific data sets, meaning their knowledge is frozen at a certain point in time

finite short term memory

AI can only hold a certain number or words/token in its active memory

prompt

the cornerstone of communication; the input provided (question, command, instructions) designed to elicit a specific response from AI

garbage in, garbage out

if inputs are flawed, your outputs will be flawed, regardless of the processes in between

role

define the persona or function the AI should adopt to better understand the user’s perspective and generate relevant responses

task

clearly specify the action or goal the AI needs to accomplish, providing precise instructions for the desired output

context

provide relevant background information, data, or prior conversations that help the AI understand the situation and make informed decisions

format

outline the desired structure, length, style, or output type for the AI’s response

pivot tables

fast, flexible summary of large data sets

rows

define how data is organized vertically; each unique value becomes a row in the pivot table, letting you group and organize data by categories like products or regions

columns

split data horizontally, creating additional groupings across the top of your table; creates a grid-like view that helps compare data across different dimensions simultaneously

values

the actual metrics you want to analyze → can be sales amounts, quantities, or inventory levels; values are calculated for each intersection of row and column values

aggregation type

this determines how your values are calculated (sum, average, count, etc.)

SUMIF function

adds values based on a condition; perfect for calculating total sales for a specific season

AVERAGEIF function

calculates the average of values that meet a specific condition; ideal for finding average sales per day during a season

XLOOKUP function

designed to help you find and retrieve data efficiently across spreadsheets

reorder point (ROP)

the inventory level at which you should place a new order to replenish stock

unmanaged costs

data accrues substantial costs related to storage, infrastructure maintenance, and energy consumption

heightened security risk

large datasets become attractive targets for cyberattacks

regulatory compliance burden

increases compliance burdens significantly

data silos/missed information

valuable information if often fragmented, buried, or inaccessible, leading to inefficient decision making and lost insights

IF statements

transforms simple yes/no logic into a sophisticated decision-making system that can handle complex scenarios with multiple possible outcomes

AND operator

requires ALL conditions to be true - like checking multiple boxes on a checklist

OR operator

requires ANY condition to be true - like having backup plans