Artificial Intelligence & Machine Learning (Leaving Cert Computer Science): Flashcards

Systems performing tasks requiring human intelligence (learning, reasoning, problem-solving)

Systems designed & trained for a specific task only

Siri, Amazon Alexa

Systems with general intelligence across wide range of tasks

Systems surpassing human intelligence; speculative concept

AI subset: systems learn & improve from experience automatically

Model trained on labelled dataset to map inputs to outputs

Model trained on unlabelled data to infer natural structure

Model uses small labelled + large unlabelled data for training

Training agents via trial & error to maximise cumulative rewards

Systematic errors/prejudices leading to unfair/inaccurate results

Representative data, statistical methods, human oversight, fairness-aware algorithms

Techniques making ML model behaviour understandable to humans

Model's inherent ability to be understood without extra tools

Techniques explaining complex models after training