Page 65 - Monocle Quarterly Journal Vol 3 Issue 2 Spring

in neural networks. As opposed to most supervised methods, which can use smaller sets of training data that prescribe the response variable as the correct outcome for the network’s output, an unsupervised neural network does not use a predefined training set as a guide, but rather relies on large volumes of data to refine its own outcomes through the continuous refinement of predictions.
In fact, it is this flood of new, rich data that has been created in recent years – and not just the improvements in processing power – that has allowed AI deep learning processes to experience a new dawn. Without the pervasive yet extremely effective information-gathering techniques used by Big Tech, the unsupervised deep learning systems that require this extreme depth and breadth of data to prune and perfect their neural networks – much as the human brain strengthens useful synaptic connections whilst pruning unnecessary ones – would not be able to function optimally. And whilst this growing body of personal and public information has effectively facilitated very promising advancements in AI, the problem is that in capturing and storing this immense data, a few giant corporations effectively control almost all of the data enabling advancements in deep learning.
Each of “The Four” capture this critical resource of raw data that feeds their continued growth in quite
different ways. Galloway breaks down these data gathering methods using a metaphor relating to cravings of the human body, namely Google as appealing to
BIG TECH AND THE COLONISATION OF DATA
In every passing second, Google processes an average of
40 000 searches, totalling 3.5 billion searches per day,
capturing a share of over 80% of searches worldwide.
the mind, Facebook appealing to the heart, Amazon appealing to the stomach, and Apple to the genitals – or as he puts it, “Google is God, Facebook is love, Amazon is consumption, and Apple is sex.” And when appealing to the most basic human desires, Galloway notes that these companies do not need to drag the data out of their customers, but rather it is freely handed over in almost every action made by their adoring and overly-trusting users.
Let us contemplate Google for a moment. In every passing second, Google processes an average of 40 000 searches, totalling 3.5 billion searches per day, capturing a share of over 80% of searches worldwide. To envisage how much data this company has collected since their
63