Page 69 - Raspberry Pi as a Foundation for Boosting Computer and Technology Literacy

Machine learning 5.3
• Low power consumption (but higher than microcontrollers) and
can draw power from a wide range of external sources, ranging
from traditional power supplies to batteries to solar panels
• It contains no moving parts and is silent
• Highly customizable and flexible compared to commercial solu-
tions

5.3 Machine learning 5.3
Artificial intelligence refers to the data science technique aimed at rep-
licating human intelligence, which encompasses thinking and acting
both humanly and rationally (Norvig and Russell 2021, 19–20). As a
result, the subject of AI and its derivatives is extraordinarily complex,
often leading to significant confusion and misconceptions. This con-
fusion, in turn, frequently results in terms such as ‘artificial neural
network’, ‘machine learning’, and ‘deep learning’ being inaccurately in-
terchanged. While these terms are certainly related, they each denote
distinct subfields within the vast domain of AI. Thus, AI includes fields
such as natural language processing and computer vision, as these
fields aim to equip machines with the abilities inherent to human in-
tellect – namely, speaking and seeing.
Thus, machine learning is also a subfield of AI since it enables the
machine to learn specific tasks by constructing models grounded in
observed data and acquiring the ability to predict potential future sce-
narios (Norvig and Russell 2021, 669). Furthermore, deep learning rep-
resents a variation of machine learning that utilizes neural networks,
encompassing both supervised and unsupervised approaches. A neural
network, in turn, comprises individual ‘neurons’ that abstractly resem-
ble the structure of neurons within the human brain (Norris 2019, 213).
These neurons are interconnected by connections and structured into
multiple layers. Thus, the machine’s learning process can be adjusted
by changing the number of layers and the weights of the connections
(Norvig and Russell 2021, 801–2).
Machine learning algorithms are generally intended to deal with
various classification, detection, recognition, and prediction prob-
lems (Norris 2019, 214). As a result, they are frequently used in image
recognition, text-to-speech, text-to-image processing, and the devel-
opment of large language models. The rising popularity of such areas
has led to the ubiquitous integration of machine learning algorithms

69