This post is part of Lifehacker’s “Living With AI” series: we explore the current state of artificial intelligence, discuss how it can be useful (and how it can’t), and assess where this revolutionary technology is heading. Read more here.
Artificial intelligence (AI) is the latest technology buzzword. Just as the cryptocurrency boom introduced the world to a whole lot of jargon, the AI hype has brought with it a set of terms that are often used but not always explained. If you’re wondering what words like chatbots, LLM or deep learning mean, you’ve come to the right place: we’ve got a glossary of 20 AI-related terms with beginner-friendly explanations.
Artificial Intelligence (AI)
Simply put, artificial intelligence is the intelligence of computers or machines, especially that which mimics human intelligence. Artificial intelligence is a broad term that encompasses many different types of machine intelligence, but the discourse around artificial intelligence currently centers largely around tools that create graphics, content, and summarize or transcribe content. Common examples of these tools include ChatGPT, Midjourney, and Google Bard. Calling these tools “bright” is up for debate, but the term “artificial intelligence” has stuck.
Algorithm
An algorithm is a set of instructions that a program executes to produce a result. Common examples of algorithms are search engines like Google, which display a set of results based on your queries, or social media applications like TikTok and Instagram, which display content based on your interests. Algorithms enable AI tools to create predictive models or create content or art based on input.
Bias
In the context of artificial intelligence, error refers to incorrect results obtained as a result of the algorithm making incorrect assumptions or lacking sufficient data. For example, speech recognition tools may not be able to correctly understand some English accents because they have only been trained with an American accent.
Conversational artificial intelligence
AI tools that you can talk to, such as chatbots or voice assistants, are called conversational AI. Popular examples of conversational AI include chatbots such as ChatGPT and Google Bard, and voice assistants such as Alexa, Google Assistant and Siri.
Data mining
The process of searching through enormous sets of data to find patterns or trends. Some AI tools apply data mining to lend a hand you understand what makes people buy more products in a store or website, or how to optimize your business to meet increased demand during peak hours.
Deep learning
Deep learning attempts to replicate the way the human brain learns by using three or more neural network “layers” to process enormous amounts of data and learn from examples. Each of these layers processes its own view of the data and combines to reach a final conclusion.
Autonomous car software uses deep learning to identify stop signs, lanes and traffic lights through object recognition: this is achieved by showing the AI tool many examples of what a specific object (e.g. a stop sign) looks like, and through repeated training the AI tool will be able to finally able to identify this object with as close to 100% accuracy as possible.
Massive Language Model (LLM)
The Vast Language Model (LLM) is a deep learning algorithm that is trained on a huge dataset to generate, translate, and process text. LLM (like GPT-4 OpenAI) enable AI tools like ChatGPT to understand your queries and generate input from them. The LLM is also powered by AI tools that can identify critical parts of text or videos and summarize them for you.
Generative artificial intelligence
Generative AI can generate artwork, images, text, and other outputs based on input data, which is often powered by LLM. Obvious examples are tools like ChatGPT, but many other compelling examples of generative AI have emerged recently. For example, Adobe Photoshop has a tool called Generative filling that can generate graphics with a few text prompts or turn a vertical image into a widescreen wallpaper.
Hallucination
When artificial intelligence presents fiction as fact, we call it hallucinations. Hallucinations can occur when the AI’s data set is not correct or its training is faulty, so it generates an answer it is certain of based on available knowledge. That said, because AI relies on a sophisticated network of networks, we don’t necessarily understand every example of hallucination. Lifehacker writer Stephen Johnson has great advice on how to easily spot AI hallucinations.
Image recognition
Possibility to identify specific objects in the image. Computer programs can apply image recognition to detect flowers in an image and name them or identify different species of birds in a photo.
Machine learning
When algorithms can improve by learning from experience or data, it is called machine learning. Machine learning is the general practice from which other artificial intelligence terms we discuss are derived: Deep learning is a form of machine learning, and enormous language models are trained using machine learning.
Natural language processing
If a program can understand input written in human languages, it undergoes natural language processing. This is how the calendar app understands what to do when you type, “I have a meeting tomorrow at 8 p.m. at a coffee shop on Fifth Avenue,” or when you ask Siri, “What’s the weather like today?”
Neural networks
The human brain is made up of layers of neurons that constantly process and learn from information. An AI neural network mimics this neuronal structure to learn from data sets. A neural network is a system that enables machine learning and deep learning, and ultimately allows machines to perform sophisticated tasks such as image recognition and text generation.
Optical Character Recognition (OCR)
The process of extracting text from images is done using OCR. The Photos app on iPhone uses OCR to recognize handwritten or typed text and lets you easily copy and paste it.
Brisk engineering
A hint is any series of words you apply to get an answer from a program, such as generative artificial intelligence. All your Google searches are examples of hints. In the context of artificial intelligence, rapid engineering is the art of writing prompts so that chatbots provide the most useful answers. It is also a field that employs people who have artistic ideas for testing artificial intelligence tools and identifying their limitations and weaknesses.
Reinforcement learning from human feedback (RLHF)
RLHF is the process of training artificial intelligence based on human feedback. When the artificial intelligence provides incorrect results, the human shows it what the correct response should be. This allows AI to deliver correct and actionable results much faster than it otherwise would.
Speech recognition
The program’s ability to understand human speech. Speech recognition can be used for conversational AI to understand queries and provide answers, or for speech-to-text tools to understand spoken words and convert them to text.
Sign
When you enter a text query into an AI tool like ChatGPT, it breaks that text into tokens, which are common sequences of characters in the text, which are then processed by the AI program. ChatGPT prices are based on the number of tokens processed: you can calculate this number using the company’s tokenization tool, which also shows how words are divided into tokens. OpenAI says one token is roughly four characters of text.
Training data
A training set or training data is the information that an algorithm or machine learning tool uses to learn and perform its functions. For example, enormous language models can leverage training data by crawling some of the world’s most popular websites to capture text, queries, and human expressions.
The Turing Test
Alan Turing was a British mathematician known as the “father of theoretical computer science and artificial intelligence”. His Turing Test (or “Imitation Game”) aims to test whether a computer’s intelligence is identical to that of a human. A computer is said to have passed the Turing test when a human is tricked into thinking that the machine’s responses were written by a human.
