Page 145 - Computer - 7

It might be tempting to think of ML as doing all the hard work, but the ML algorithm can only follow
the precise instructions of its creator. This section highlights the art of the AI engineer. They harness the
power of concepts from a range of disciplines – most notably computing, logic, statistics, and calculus –

while balancing a range of considerations about the problem itself and the context of its solution.

FACTORS ENCOURAGING ML

The factors that make Machine Learning AI the more relevant approach towards AI are:

Massive Datasets

Machine Learning algorithms tend to require large quantities of training data in order to produce high
performance AI models. For example, some facial recognition AI systems can now routinely outperform
humans, but to do so requires tens of thousands or millions of labeled images of faces for training data.

When Machine Learning was first developed decades ago, there were very few applications where
sufficiently large training data was available to build high performance systems. Today, an enormous
number of computers and digital devices and sensors are connected to the internet, where they are
constantly producing and storing large volumes of data, whether in the form of text, numbers, images,
audio, or other sensor data files. Of course, more data only helps if the data is relevant to your desired
application. In general, training data needs to match the real-world operational data very, very closely
to train a high-performing AI model.

Increased Computing Power
To a much greater extent than Symbolic AI Systems, Machine Learning AI systems require a lot of

computing power to process and store the large volumes of datasets. During the early 21st century,
computing hardware started getting powerful enough and cheap enough that it was possible to run
Machine Learning algorithms on massive datasets using commodity hardware.
One especially important turning point around 2010 was developing effective methods for running

Machine Learning algorithms on Graphics Processing Units (GPUs) rather than on the Central Processing
Units (CPUs) that handle most computing workloads. Originally designed for video games and computer

143