Classifying Intelligence in Machines: A Taxonomy of Intelligent Control

Wilson, Callum; Marchetti, Francesco; Carlo, Marilena Di; Riccardi, Annalisa; Minisci, Edmondo

doi:10.3390/robotics9030064

Cited by 10 publications

(7 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To deal with this complexity, many organisations use multiple dimensions to classify AI systems (Wilson et al, 2020). For example, the OECD (2022) has recently published a Framework for the Classification of AI systems.…”

Section: The Matrixmentioning

confidence: 99%

The Switch, the Ladder, and the Matrix: Models for Classifying AI Systems

et al. 2023

View full text Add to dashboard Cite

Organisations that design and deploy artificial intelligence (AI) systems increasingly commit themselves to high-level, ethical principles. However, there still exists a gap between principles and practices in AI ethics. One major obstacle organisations face when attempting to operationalise AI Ethics is the lack of a well-defined material scope. Put differently, the question to which systems and processes AI ethics principles ought to apply remains unanswered. Of course, there exists no universally accepted definition of AI, and different systems pose different ethical challenges. Nevertheless, pragmatic problem-solving demands that things should be sorted so that their grouping will promote successful actions for some specific end. In this article, we review and compare previous attempts to classify AI systems for the purpose of implementing AI governance in practice. We find that attempts to classify AI systems proposed in previous literature use one of three mental models: the Switch, i.e., a binary approach according to which systems either are or are not considered AI systems depending on their characteristics; the Ladder, i.e., a risk-based approach that classifies systems according to the ethical risks they pose; and the Matrix, i.e., a multi-dimensional classification of systems that take various aspects into account, such as context, input data, and decision-model. Each of these models for classifying AI systems comes with its own set of strengths and weaknesses. By conceptualising different ways of classifying AI systems into simple mental models, we hope to provide organisations that design, deploy, or regulate AI systems with the vocabulary needed to demarcate the material scope of their AI governance frameworks.

show abstract

Section: The Matrixmentioning

confidence: 99%

The Switch, the Ladder, and the Matrix: Models for Classifying AI Systems

et al. 2023

View full text Add to dashboard Cite

show abstract

“…4b) considering also the uncertainties on the physical models. According to the taxonomy presented in [18], the proposed control approach can be classified as G0, E2, C1, since the goal is predefined by the user and it is followed by minimizing the tracking error (G0); the environment is defined but subject to time varying disturbances, modelled by the introduction of environmental disturbances and uncertainties (E2); the GP control law parameters are updated online by the NN (C1).…”

Section: Neural Optimization Of the Genetic Programming Control Lawmentioning

confidence: 99%

“…Many different Artificial Intelligence (AI) techniques have been used in the past decades for various IC applications [18], where the definition of IC can be summarized as follows: a controller can be defined intelligent if it can deal autonomously with unforeseen changes in the environment, in the control system or in the goals, by relaying on techniques pertaining to the fields of Artificial Intelligence, Operations Research and Automatic Control (Saridis [16] and Antsaklis [1]). Mainly three different classes of AI techniques have been used for IC both alone and hybridized: Fuzzy Logic (FL), Evolutionary Computing (EC) and Machine Learning (ML).…”

Section: Introductionmentioning

confidence: 99%

A Hybrid Neural Network-Genetic Programming Intelligent Control Approach

Marchetti

Minisci

2020

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

The proposed work aims to introduce a novel approach to Intelligent Control (IC), based on the combined use of Genetic Programming (GP) and feedforward Neural Network (NN). Both techniques have been successfully used in the literature for regression and control applications, but, while a NN creates a black box model, GP allows for a greater interpretability of the created model, which is a key feature in control applications. The main idea behind the hybrid approach proposed in this paper is to combine the speed and flexibility of a NN with the interpretability of GP. Moreover, to improve the robustness of the GP control law against unforeseen environmental changes, a new selection and crossover mechanisms, called Inclusive Tournament and Inclusive Crossover, are also introduced. The proposed IC approach is tested on the guidance control of a space transportation system and results, showing the potentialities for real applications, are shown and discussed.

show abstract

“…In fact, what was done so far in the literature consists in applying the GP offline to find a control law. On the other hand, if GP is used online, it is possible to define that control approach, IC [9].…”

Section: Introductionmentioning

confidence: 99%

Genetic Programming Guidance Control System for a Reentry Vehicle under Uncertainties

Marchetti

Minisci

2021

Mathematics

Self Cite

View full text Add to dashboard Cite

As technology improves, the complexity of controlled systems increases as well. Alongside it, these systems need to face new challenges, which are made available by this technology advancement. To overcome these challenges, the incorporation of AI into control systems is changing its status, from being just an experiment made in academia, towards a necessity. Several methods to perform this integration of AI into control systems have been considered in the past. In this work, an approach involving GP to produce, offline, a control law for a reentry vehicle in the presence of uncertainties on the environment and plant models is studied, implemented and tested. The results show the robustness of the proposed approach, which is capable of producing a control law of a complex nonlinear system in the presence of big uncertainties. This research aims to describe and analyze the effectiveness of a control approach to generate a nonlinear control law for a highly nonlinear system in an automated way. Such an approach would benefit the control practitioners by providing an alternative to classical control approaches, without having to rely on linearization techniques.

show abstract

Classifying Intelligence in Machines: A Taxonomy of Intelligent Control

Cited by 10 publications

References 61 publications

The Switch, the Ladder, and the Matrix: Models for Classifying AI Systems

The Switch, the Ladder, and the Matrix: Models for Classifying AI Systems

A Hybrid Neural Network-Genetic Programming Intelligent Control Approach

Genetic Programming Guidance Control System for a Reentry Vehicle under Uncertainties

Contact Info

Product

Resources

About