AI through the looking glass

I invited ChatGPT again to configure elements of my glossary. See previous post A critical AI dictionary. This time I prompted the AI to reword some of the definitions in the manner of Lewis Carol’s Through the Looking Glass.

Dall-e: A generative graphics product accessed via ChatGPT and other LLMs. DALL·E uses a transformer model that first analyzes the prompt requesting an image by breaking it into tokens. Then, it generates an image by predicting, step by step, the pixels or patches of the image that correspond to the text, using patterns from training on large datasets of text-image pairs. It maps the textual description to visual elements, gradually assembling them into a coherent image.

DALL·E: A curious contraption, much like a painter with a mind full of dreams, who listens carefully to your every word. With a twirl and a whirl, it breaks your request into tiny puzzle pieces—each a token, mind you! Then, with a most peculiar magic, it begins to stitch together a picture, pixel by pixel, as if pulling colors from the very air. With patterns learned from endless picture-books and stories, DALL·E spins your words into a vivid tapestry, turning thoughts into sights with a dash of whimsy and a sprinkle of wonder.

Diffusion Model: A type of machine learning model that generates data, like images or text, by simulating a process that gradually adds and then removes noise to transform a simple pattern into a more complicated one. It’s used in tasks like image generation and reducing noise (e.g. speckles and smudges) in images.

Diffusion Model: Imagine, if you will, a peculiar kind of alchemist who begins with a simple splash of ink on a page. At first, it’s all rather a mess—just noise and nonsense—but watch closely! For as the ink spreads and dances about, it starts to shape itself into something quite extraordinary. By adding a bit of this and taking away a bit of that, the Diffusion Model conjures up images and words from the chaos, like a wizard revealing hidden patterns in a swirling mist.

Embedding: A numerical representation of words or phrases in a continuous vector space, capturing semantic and other relationships for processing by AI models. The size of an embedding typically ranges from a few dozen to several hundred dimensions, with larger embeddings capturing more subtle nuances in word meanings and relationships but also requiring more computational resources.

Embedding: Picture a curious map, not of lands and seas, but of words and ideas! Each word, phrase, or thought is a little dot in a vast and colorful space, connected by invisible threads of meaning. The Embedding is the cartographer of this strange land, drawing the paths and borders that show how close or far our words are from each other. Some dots are clustered together, whispering secrets to one another, while others stand alone, guarding their own mysteries. It’s a map not just of where words are, but of what they mean and how they dance together in the grand waltz of language.

Encoding: The process of converting input data, such as tokens, words or phrases, into a structured format like embeddings, which allows AI models to process and analyse the data. Encoding transforms raw text into numerical vectors that capture syntactic, semantic, or positional information, depending on the AI model’s requirements.

Encoding: Ah, here we have a most peculiar machine, like a translator of thoughts! When you feed it a string of letters, words, or phrases, it busily sets to work, turning them into neat little bundles of numbers. These numbers, mind you, are no ordinary ones—they’re enchanted! They hold the essence of meaning, the shape of sentences, and the rhythm of thoughts, all packed into a form that the great minds of AI can ponder and play with. It’s as if raw words are spun into golden threads, ready to be woven into something marvelous.

Existential Risk: The potential threat to human existence or long-term survival, often discussed in the context of advanced AI systems capable of actions that could be catastrophic for humanity.

Existential Risk: Beware, says the looking-glass, for there lies a shadowy specter at the edge of reason, whispering of dangers too great to ignore. The Existential Risk is a frightful thing indeed—a threat that looms over the very existence of us all. In the world where machines think and act, one must be cautious, for a single misstep could lead to a tumble down a rabbit hole from which there may be no return. It’s the chessboard’s final move, where the game could end in a checkmate of the world itself.

Extinction: The complete disappearance of a species or group, often used metaphorically to describe the potential irreversible impact of AI on human society.

Extinction: The final curtain falls, and with it, the stage is empty, the players gone. Extinction is a word heavy with the weight of endings, a silence where once there was song. It’s the vanishing of a species, the last note in a symphony that will never be played again. In the looking-glass world, it might be the end of a story—an irreversible closing of the book, where the pages turn to dust and the words are lost forever. A most somber thought, indeed, for one must wonder what role AI might play in such a tale, whether as hero or harbinger of the end.

I feel this exercise is helpful to my compilation. Though I probably wouldn’t adopt Looking Glass language in my glossary, it serves as a way of looking at things differently, rendering the familiar strange and in new ways.