random seed
In plain wordsA number that fixes the 'random' choices a computer makes; reuse the same seed and settings, and an image generator produces the same picture again.
A starting value, usually an integer, fed to a pseudorandom number generator; because the generator is a deterministic algorithm, the same seed always yields the same sequence of 'random' numbers, which makes a computation reproducible. In machine learning, seeds control weight initializationweight initialization, n.: Choosing the random starting values of a neural network's internal numbers before training. Good starting values keep signals from fading away or blowing up as they pass through many layers., data shuffling, dropoutdropout, n.: A training trick that randomly switches off a fraction of a network's neurons at each step, so the network cannot lean too heavily on any one of them and learns sturdier patterns. masks and sampling, and reporting results over several seeds is standard practice for showing that a finding is not a fluke. In image and video generation, the seed determines the initial noise from which a diffusion modeldiffusion model, n.: An AI that learns to create images, sound or video by practicing noise removal: shown pictures buried under static, it learns to clean them up, then builds new ones starting from pure static. or flow matchingflow matching, n.: A way to train generators that learns, for every point between random noise and real data, which direction to move, so that noise can be steered smoothly into new examples. model starts: with the same model, prompt, samplerdiffusion sampler, n.: The step-by-step routine a diffusion model follows to turn noise into a finished picture; the choice of sampler and number of steps trades speed against quality., step count and settings, the same seed reproduces the same image, while a new seed gives a different composition. Users therefore lock a seed to vary one setting at a time, and share seeds so that others can recreate results. Exact reproduction can still fail across hardware, software versions or batch sizes, because parallel floating-point arithmetic is not always deterministic.
Same prompt, seed 1234, and the castle comes back — only now in snow.
Origin
Seeding goes back to the earliest pseudorandom number generators, such as John von Neumann's middle-square method of the late 1940s. In generative AI, exposing the seed as a user setting became standard with the open-source Stable Diffusion tools of 2022, and seeds are now routinely shared alongside prompts.
Usage note
Language-model APIs sometimes offer a seed parameter, but providers generally promise only best-effort reproducibility; see nondeterminismnondeterminism, n.: The fact that an AI model can give different answers to exactly the same question, partly because it deliberately chooses words with some randomness and partly because of tiny computing differences..
Related
- nondeterminismnondeterminism, n.: The fact that an AI model can give different answers to exactly the same question, partly because it deliberately chooses words with some randomness and partly because of tiny computing differences.
- diffusion samplerdiffusion sampler, n.: The step-by-step routine a diffusion model follows to turn noise into a finished picture; the choice of sampler and number of steps trades speed against quality.
- samplingsampling, n.: Picking some items at random from a larger set or from a range of possibilities — surveying 1,000 voters, say, or letting an AI choose its next word by a weighted lottery.
- stochasticstochastic, adj.: Involving chance, so the same starting point can lead to different outcomes. AI chatbots are stochastic: ask the same question twice and you may get two different answers.
- temperaturetemperature, n.: A setting that controls how adventurous an AI model's word choices are. Low temperature makes it pick the safest, most likely words; high temperature makes it more varied and surprising, but more error-prone.
- weight initializationweight initialization, n.: Choosing the random starting values of a neural network's internal numbers before training. Good starting values keep signals from fading away or blowing up as they pass through many layers.
Mentioned in
AI artAI art, n.: Images, music or other artworks made with the help of AI — from early computer drawing programs to today's prompt-based image generators — and the debate over whether and how such work counts as art., MidjourneyMidjourney, prop. n.: A popular paid AI image generator, first used by typing prompts into the chat app Discord, known for its polished, painterly look; it can also animate images into short videos., text-to-image modeltext-to-image model, n.: An AI system that creates a picture from a written description, such as 'a fox reading a newspaper, in watercolor', usually within seconds., diffusion samplerdiffusion sampler, n.: The step-by-step routine a diffusion model follows to turn noise into a finished picture; the choice of sampler and number of steps trades speed against quality., nondeterminismnondeterminism, n.: The fact that an AI model can give different answers to exactly the same question, partly because it deliberately chooses words with some randomness and partly because of tiny computing differences., reproducibilityreproducibility, n.: Whether other people, or the same people later, can get the same results by rerunning an AI experiment. Many published AI results have proved hard to reproduce, which makes real progress harder to judge.