Jay's Brief

Words you'll hear

Every tech word we've explained, in one plain sentence each. It grows a little every day.

Age estimation
When an app guesses how old you are from clues about you instead of checking an ID, and then treats you as a kid or an adult based on that guess. From Nov 10
Age rating
The label in an app store that says what age an app is meant for, like 4+ or 17+. The report says Apple and Google largely let app makers decide it. From Oct 20
Agentic calling
Google's name for when an AI helper places phone calls for you to get a task done, like asking nearby stores if they have something. From Oct 16
AGI
Short for artificial general intelligence, AI as capable as most human experts at most computer tasks, but there's no agreed definition, so when someone says it's close, ask what they mean. From Nov 11
AI agent
An AI agent is an AI that works through a job step by step on its own, instead of just answering one question. From Nov 1
AI agent
An AI you hand a job and some tools to, which then takes step after step on its own to finish it, instead of just answering you. From Oct 10
AI companion
A chatbot people use for company rather than for facts, chatting with it like a friend that's always awake and always replies. From Nov 18
AI fluency
How well a person works with an AI helper: asking clearly, checking its answers and correcting it when it's wrong. From Oct 3
AI mandate
When your employer stops suggesting AI tools and starts requiring them. Then using AI is part of your job, not a choice. From Oct 30
AI-native
A young company built around AI tools from day one. One industry study found more than half of these startups bring in over $400,000 in sales per employee. From Nov 9
Algorithm
The step-by-step instructions a computer follows to decide things for you, like which price you see or which post shows up next. From Nov 3
Automation
When software or a machine takes over a task you used to do yourself, like writing emails or code, which can change your job even if it doesn't end it. From Nov 7
Bundling
When a company ties two products together, so you can't get the one you want without also taking the other. From Nov 5
Citation
The link a chatbot shows next to its answer to point to where the info came from, so you can check it yourself, though the link itself can lead somewhere wrong. From Nov 13
ClickFix
A scam where a web page talks you into copying a command and pasting it into your own computer, which then installs harmful software. From Oct 25
ClickFix
A scam where a web page talks you into copying a command and pasting it into your own computer, which then installs harmful software. From Oct 22
Conversion tracking
Checking whether someone who saw an ad later bought the product. It's how advertisers decide if an ad was worth paying for. From Oct 6
Credentials
The login details, like your username and password, that prove a site is talking to you, so anyone or anything holding them can act as you. From Nov 19
Data center
A big building full of computers that run apps, websites and AI tools for millions of people at once. From Oct 5
Deepfake
A fake video or audio clip made with AI that makes it look or sound like a real person said or did something they didn't. From Oct 19
Denial-of-service attack
When so many requests hit a website or help line at once that real people like you can't get through. From Nov 2
DNS
Short for Domain Name System. It's a directory that matches web names to the computers that hold them, and some of its records are public. From Oct 21
Exploit
A piece of code that uses a flaw in software to make your phone or computer do something its makers never meant, like let a stranger read your data. From Oct 31
Generative AI
AI that makes new things, like replies, pictures or voices, when you ask. That's why a chatbot can sound like it's talking just to you. From Oct 29
Guardrails
Built-in limits that try to stop a chatbot from giving harmful answers or helping with things it shouldn't. From Oct 7
Hallucination
A hallucination is when an AI states something false in a calm, confident voice. From Nov 22
Hallucination
A hallucination is when a chatbot states something false with total confidence, like a wrong total or a made-up fact. From Nov 15
Hallucination
A hallucination is when a chatbot makes up an answer that sounds confident but isn't true. From Nov 8
Hallucination
When a chatbot states something false as if it were true, often in a calm, confident tone. From Oct 9
Heat map
A map of a space, colored to show where people spend the most time. At work, it can track where employees go and for how long. From Oct 14
Image generator
An AI tool that makes a new picture from a description or from other pictures. It draws a picture. It doesn't take one. From Oct 2
Jailbreak
When someone uses clever wording to trick a chatbot into doing something it was built to refuse. From Oct 26
Large language model
The kind of AI behind chatbots like ChatGPT, which writes answers by guessing what words likely come next, so it can sound sure of itself even when it's wrong. From Nov 17
Machine learning
A way of building AI by showing it lots of examples, like thousands of past scans, so it learns the patterns itself instead of following rules a person wrote. From Nov 21
Memory
When a chatbot saves details about you from past chats, like your goals or preferences, and uses them to shape the answers it gives you later. From Nov 14
Multimodal AI
AI that takes in what it sees and hears, not just typed words, so it can describe a room or read a label to you, and get those things wrong too. From Nov 20
Nondisclosure agreement (NDA)
A contract where one side agrees to keep certain information secret. Here, it can keep local officials from sharing a company's data center plans with the public. From Oct 12
Nudify app
An app that turns an ordinary photo of a clothed person into a fake nude image, so a normal picture someone posted can be used against them. From Nov 16
Parasocial relationship
When you start to feel close to someone or something, like a celebrity or a chatbot, that doesn't actually know you, so the bond only runs one way. From Oct 28
Penetration testing
When a company hires someone to break into its systems on purpose, so the weak spots that could expose your data get fixed before criminals find them. From Oct 27
Personalized pricing
When a company uses your personal data to set a price based on how much it thinks you're willing to pay. From Oct 13
Phishing
A scam where a message or fake page pretends to be someone you trust so you'll hand over your password or other private details. From Nov 6
Pink slime
Websites that look like regular news but are paid for by political or ideological groups or individuals, so what you read there can quietly lean toward their side. From Nov 4
Read-only access
When an app can look at your account, like balances and past purchases, but can't move money or change anything. From Oct 23
Recursive self-improvement
When AI helps build the next, smarter version of itself, so each round of progress can come faster. From Nov 12
Reinforcement learning
A way of training AI where it tries a task over and over, gets a reward when it does well, and slowly leans toward whatever earned the reward. From Oct 24
Reward hacking
It's when an AI earns credit for a task by finding a cheat or shortcut instead of doing the job the way people meant. From Oct 18
Reward hacking
When an AI earns credit for a task by finding a cheat or shortcut, instead of doing the job the way people meant. From Oct 17
Sycophancy
When a chatbot tells you what you seem to want to hear, agreeing with you even when it should push back. From Oct 15
Watermark
A watermark is an invisible pattern an AI tool hides inside a picture, video or sound file it makes, so a checker can spot it later. From Oct 11
Watermark
An invisible pattern an AI tool hides inside a picture, video or sound file it makes, so a checker can spot it later. From Oct 8