5 Reasons Why I’m Relaxed About the Rise of the Machines (And so Should You Be)
We’re all familiar with cautionary tales such as The Terminator in which artificial intelligence (AI) develops full sentience and aims to destroy and replace its human creators. While I strongly urge a note of caution towards AI and machine learning (ML) I also trust that readers of this blog are able to discern fact from science fiction. Most people’s fears of AI and ML are borne out of a misunderstanding regarding what these technologies actually are. While the idea of an intelligent machine will likely remain an existential anxiety for most, I’m relaxed about the development of AI and ML. In this article, I will aim to dispel some of the most common fears and anxieties by demystifying AI and ML. I’ll aim to highlight that their potential for good far outweighs any remote threat that they might pose to human existence. Here are five reasons why I’m relaxed about the development of these technologies and so should you be.
Reason #1. Machine Learning is as old as most common computer languages
Although ML has come to prominence in recent years, the term actually has a relatively long history. John McCarthy was the person who coined the term ‘Artificial intelligence’ back in 1956 at the first Dartmouth Conference to address this subject. During this period, the first computer languages were written, meaning that both AI and ML have their roots within the same time period as other technologies that we are familiar with.
Reason #2. ML is a process, not an entity
Both AI and ML are difficult and abstract concepts for most people to understand and I believe this is where much of the fear and anxiety comes from. While these two terms are often used interchangeably, defining them helps to demystify them. AI actually describes an entity – in other words, it’s an intelligence that is man-made. ML refers to a process by which AI is able to learn. This process can be understood by anyone. Once you understand that ML is a process, it is easier to overcome any anxieties you may have that it could lead to the creation of a sentient, malevolent being.
Reason #3. Computers must always be taught how to think
The neurons in the human brain are good at making intuitive connections that computers are incapable of. In understanding AI and ML, it’s important to understand that all computers must be taught how to think using what’s known as ‘training data’.
To understand this, let’s take the fictitious example of a fast food restaurant and an autonomous burger-making machine. The goal of AI and ML would be to train the machine to predict how many burgers it should produce and at what time. In order to accomplish this, human programmers would need to teach the computer how to make connections between individual data points, such as the time of the day and the number of people ordering burgers. This would be taught through trial and error and, over time, the machine could learn to extrapolate that data and make new connections itself.
Reason #4. ML applications still require significant oversight
An important part of demystifying AI and ML is understanding how machines that incorporate this technology are tested. The process involves identifying the ML machine that is most successful in predicting connections between different factors. This process still requires significant human oversight.
If we continue the fictitious burger-machine example, imagine that we have two identical machines, A and B. The job of an ML algorithm would be to make intuitive connections between the time of day and the number of burgers likely required. If Machine A predicts that demand for burgers will rise at lunchtime and Machine B comes to the conclusion that 3 AM will be the busiest time, it’s clear to see which machine will be discarded! That type of decision takes human oversight. It’s only through large trials of multiple machines that developments in AI and ML will come about.
Reason #5. AI and ML are just math and mechanics, not sorcery
I’ve provided a deliberately simplistic overview of AI and ML in order to demystify the process for the average reader. Further reading (and a lot of math) are recommended for those who want to deepen their understanding of these areas. There are certainly far more complex examples of the ways in which ML is being used than I have laid out here. However, my intention is to show that, beyond the jargon, AI and ML are processes of math and mechanics. These processes underpin all iterations of ML, from enabling computers to have conversations with humans, to helping doctors spot cancer and even teaching autonomous cars how to drive.
Deepfakes and the need for caution
While I don’t believe that the development of sentient, malevolent AI is likely within our lifetimes, I fully accept that we need to adopt a cautious approach towards AI and ML. One of the most frightening uses of these technologies to come into the public consciousness recently was the emergence of so-called ‘deepfake’ videos. These digitally manipulated videos harness the power of ML algorithms to create the illusion of someone appearing in a video they were never in. Some of the most popular deepfake videos are harmless fun, such as having Nic Cage appear in a range of popular Hollywood movies. As a huge Nic Cage fan, I’m all for that. However, there were a series of horrifying deepfakes that were used to sexually degrade women and, thankfully, they have now been removed from the most popular sites such as Reddit. As it’s impossible to remove anything completely from the internet, they serve as a cautionary tale of the problems that AI and ML can create.
This leads to the frightening proposition that we could see deepfake fake news and propaganda emerge in the future. While early deepfakes were created by programmers, the technology has developed to a point where non-technical users can create a decent deepfake without any programming skills whatsoever. What happens if and when someone turns their attention to world politics? Our only hope is that when the technology develops to a point of hyperrealism, programmers may be able to develop counter-deepfake programs that use AI and ML to help viewers spot deepfakes more easily. This area highlights a clear need for vigilance, caution and awareness.
Examples of the misuse of AI and ML algorithms such as the creation of deepfakes should give everyone pause for concern. This isn’t necessarily a bad thing; imagine if humans had paused to consider the long-term implications of fossil fuel usage before setting the planet on the path towards global warming with such vigor. However, I hope that people can discard the idea that magic or sorcery is behind AI and ML. These technologies are already revolutionizing the way we live and work and their potential for good is unmatched by any other technology. Today’s major tech players, including Apple, Amazon and Google are already chasing the AI prize via smart assistants such as Google Home and Amazon’s Alexa.
As I noted at the outset, a note of caution towards AI and ML is justified but the chances of a sentient destructive AI developing autonomously remains a remote possibility. I urge us all to promote the development of AI and ML for the benefit of everyone while regulating it so that it shapes, instead of hinders, our futures.