Machine Learning in Artificial Intelligence: Towards a Common Understanding

Abstract

The application of "machine learning" and "artificial intelligence" has become popular within the last decade. Both terms are frequently used in science and media, sometimes interchangeably, sometimes with different meanings. In this work, we aim to clarify the relationship between these terms and, in particular, to specify the contribution of machine learning to artificial intelligence. We review relevant literature and present a conceptual framework which clarifies the role of machine learning to build (artificial) intelligent agents. Hence, we seek to provide more terminological clarity and a starting point for (interdisciplinary) discussions and future research.

Unterstützung der Wissensarbeit durch künstliche Intelligenz – Anforderungen an die Gestaltung maschinellen Lernens

“…Dies impliziert nicht nur, dass das Assistenzsystem insgesamt den sonst üblichen Sicherheitsanforderungen, die an IT-Systeme zu stellen sind, genügen muss. Insbesondere ein System mit einem "learning backend" (Kühl et al 2019), das also kontinuierliches Lernen zulässt, ist so auszulegen, dass es sich robust gegen Eingriffe von außen selbst dann verhält, wenn die Manipulationsversuche auf zulässigem Wege, also über die Lernkomponente des Systems erfolgen. So wurde 2015 Microsofts Kommunikationsbot "Tay", der aus der Kommunikation mit Twitter-Nutzern lernen sollte, von einer Gruppe gezielt mit Hasskommentaren gefüttert, so dass er von Microsoft bereits nach wenigen Stunden aus dem Netz genommen werden musste (Sickert 2016…”

“…Künstliche Intelligenz (KI) und Anforderungen an das maschinelle Lernen. Rückkopplung laufend weiter adaptiert werden(Hirt et al 2019). Beiden Ansätzen gemein ist, dass auf Basis existierender Datensets Modelle gebildet werden, die dann auf weitere Sachverhalte in Form neuer Daten angewandt werden.…”

Intelligent Tutoring Systems: Why Teachers Abandoned a Technology Aimed at Automating Teaching Processes

“…The expectation is that the artificial intelligence (AI) applications used in ITSs can support teaching and learning through machine learning and personalized learning. Machine learning is based on algorithms that enable a digital system to automatically learn, often by using training data, to make decisions or predictions [9]. Personalized learning is believed to make learning more efficient and relevant to a student's needs when individual feedback and assessment are used to tailor the instructions [10].…”

Easy and Efficient Hyperparameter Optimization to Address Some Artificial Intelligence “ilities”

“…Analytics and machine learning (ML), colloquially termed Artificial Intelligence (AI) [1], are becoming increasingly ubiquitous for classification and prediction across a broad spectrum of applications due to their ability to learn nonlinear patterns in data [1]. Fundamentally, AI/ML are complex algorithms that automate procedures based on statistics and nonlinear optimization [1] [2]. However, as a result of their complexity, issues exist in broadly adopting AI solutions [3].…”

“…As illustrated in Figure 2, craft involves highly experiential work, which involves tricks of the trade, and various and haphazard approaches to transmit 1 For example, DataRobot (https://www.datarobot.com/), AutoML (http://www.automl.org/), and SigOpt (https://sigopt.com/) knowledge. When a discipline moves to more established procedures, production, and developing applications for sales, then one has moved to a commercial stage.…”