Home › Technology › Artificial Intelligence & ML › Inference

Inference

IN-fuh-rens /ˈɪnfərəns/

nounC1GRETOEFLIELTS

formaltechnicalacademic

A conclusion reached on the basis of evidence and reasoning; (in AI) the process of using a trained model to make predictions or decisions on new data.

In AI, 'inference' is the operational phase where a deployed model takes new, unseen data and applies its learned patterns to generate an output, such as a prediction, classification, or recommendation. It's the 'doing' part after the 'learning' part.

The story behind the word. This word comes from Latin `inferre`, meaning 'to bring in' or 'to introduce.' It combines `in-` (into) and `ferre` (to carry, to bear). So, to `infer` is to 'bring in' a conclusion based on evidence and reasoning. In AI, it's the process where the trained model 'brings in' its learned knowledge to make new predictions.

Word relationships

conclusion: A general term for the end result of reasoning.
deduction: A specific type of inference, moving from general premises to specific conclusions.
prediction: The specific output of an AI model during inference.

Commonly confused with

assumption: An assumption is something taken for granted without proof, whereas an inference is a conclusion based on evidence.

Word family

infer (verb): To deduce or conclude (information) from evidence and reasoning rather than from explicit statements.
inferential (adjective): Relating to or involving inference.

Collocations

draw an inference
logical inference
statistical inference
real-time inference

Example sentences

"From the available evidence, the only logical Inference was that the suspect was innocent." In a philosophical discussion.
"Once the Deep Learning Model is deployed, it performs real-time Inference on incoming sensor data to guide the autonomous vehicle." Discussing AI deployment.

Memory hook

Think 'IN-formation' leading to a 'FERENCE' (reference) or conclusion. You bring in information to make a conclusion. Bring In Conclusion

When not to use

Do not use 'inference' to describe the learning process itself (which is 'training'). Inference is about applying what has been learned.

Fun facts

In logic, inference rules are used to derive new statements from existing ones.
The speed of inference is critical for real-time AI applications like self-driving cars or conversational agents.