classifier
In plain wordsA program or model that sorts inputs into categories, for example flagging an email as spam or a chatbot message as harmful. Many AI safety filters are classifiers.
A function or trained model that maps an input to one of a set of discrete classes, usually by computing a score for each class and applying a decision rule. Probabilistic classifiers estimate a probability for each class, as logistic regressionlogistic regression, n.: A simple, widely used method for yes-or-no predictions that outputs a probability — say, a 70% chance a customer cancels — by passing a weighted sum of inputs through an S-shaped curve. and softmaxsoftmax, n.: A formula that turns a list of raw scores into probabilities — all positive and adding up to one — with the biggest share going to the highest score. networks do, and a label is then chosen by arg max or by a threshold tuned to the relative costs of errors; others, such as support vector machinessupport vector machine, n.: A classic method that separates two groups of examples with the widest possible gap — like laying the broadest possible road between them. A mathematical shortcut lets it draw curved boundaries too., produce uncalibrated scores or labels only. Classifiers are called linear or nonlinear by the shape of their decision boundarydecision boundary, n.: The dividing line, or surface, where a classifier switches from predicting one category to another. Simple models draw straight lines; flexible ones draw curves., and discriminativediscriminative model, n.: A model that learns only to tell things apart — spam or not spam, cat or dog — without learning how to produce new examples of those things. or generative according to whether they model p(y | x) directly or model how inputs arise from each class. In deployed AI systems the word often denotes a separate model that screens inputs or outputs, such as a safety classifiersafety classifier, n.: A separate AI model that checks messages going into or coming out of a chatbot and flags or blocks those that break safety rules, like a security guard at the door. or an AI detectorAI detector, n.: Software that scores how likely it is that an AI, not a person, produced some text or media; for writing, its verdicts are wrong often enough that many schools no longer trust them..
The classifier flags about 2% of prompts for human review at a 0.8 threshold.
Origin
Used in statistics and pattern recognition since at least the 1950s–1960s for trainable decision devices such as the perceptron; Nils Nilsson's book 'Learning Machines: Foundations of Trainable Pattern-Classifying Systems' (1965) treated them systematically. The safety-filter usage spread with content-moderation models in the 2010s and LLM guardrails after 2022.
Usage note
A classifier's raw scores are not automatically trustworthy probabilities; see calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time.. In casual use, 'classifier' sometimes means the whole moderation pipeline rather than a single model.
Related
- classificationclassification, n.: Sorting things into categories — spam or not spam, which animal is in a photo, which language a text is written in — by learning from examples that are already labeled.
- decision boundarydecision boundary, n.: The dividing line, or surface, where a classifier switches from predicting one category to another. Simple models draw straight lines; flexible ones draw curves.
- logistic regressionlogistic regression, n.: A simple, widely used method for yes-or-no predictions that outputs a probability — say, a 70% chance a customer cancels — by passing a weighted sum of inputs through an S-shaped curve.
- support vector machinesupport vector machine, n.: A classic method that separates two groups of examples with the widest possible gap — like laying the broadest possible road between them. A mathematical shortcut lets it draw curved boundaries too.
- discriminative modeldiscriminative model, n.: A model that learns only to tell things apart — spam or not spam, cat or dog — without learning how to produce new examples of those things.
- safety classifiersafety classifier, n.: A separate AI model that checks messages going into or coming out of a chatbot and flags or blocks those that break safety rules, like a security guard at the door.
- calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time.
Sources
- Nilsson, N. J. (1965). Learning Machines: Foundations of Trainable Pattern-Classifying Systems. McGraw-Hill.
Mentioned in
classificationclassification, n.: Sorting things into categories — spam or not spam, which animal is in a photo, which language a text is written in — by learning from examples that are already labeled., content moderationcontent moderation, n.: Checking what people post, or what an AI produces, and removing, limiting or labeling whatever breaks the rules or the law — done by people, by AI, or by both together., decision boundarydecision boundary, n.: The dividing line, or surface, where a classifier switches from predicting one category to another. Simple models draw straight lines; flexible ones draw curves., deepfake detectiondeepfake detection, n.: Using software, often itself AI, to tell whether a video, photo or voice recording has been faked or generated by AI rather than captured from reality., discriminative modeldiscriminative model, n.: A model that learns only to tell things apart — spam or not spam, cat or dog — without learning how to produce new examples of those things., safety classifiersafety classifier, n.: A separate AI model that checks messages going into or coming out of a chatbot and flags or blocks those that break safety rules, like a security guard at the door., randomized smoothingrandomized smoothing, n.: A way to make a classifier provably harder to fool: ask it about many noisy copies of an input and take a majority vote, which small deliberate changes cannot easily flip.