The Tick-tock of AI starts with the tick-tock of a clock.
“The career of a young theoretical physicist consists of treating the harmonic oscillator in ever-increasing levels of abstraction.”
When I was 18, my first-year mechanics class introduced me to the concept of the so-called simple harmonic oscillator: the mathematical description for something that moves back and forth in regular fashion, like a swinging pendulum in a grandfather clock or a hanging crate on a spring. Like most long-term relationships, I didn’t think much of this first encounter; after all, it was just a simple section in an introductory text — right?
I was wrong.
The next decade of my life would be learning to use (and, perhaps somewhat begrudgingly, appreciate) the harmonic oscillator as it showed up again, and again, and again, and again. From classical mechanics to quantum mechanics to quantum field theory, the harmonic oscillator was always there.
When I moved into the world of data science (first the field, then the company!), I was enamoured with all the cutting-edge tech: generative adversarial networks (GANs), deep learning, large language models, recurrent neural networks, the (buzzword) list goes on. Apparently, however, history does tend to rhyme, because no matter what avenue I went down, each and every tech was ultimately constructed upon the shoulders of the same giants: statistics, linear algebra, differential equations, and regression.
The AI Boom
Now, unless you’ve been living under an extremely large boulder, you’ve likely been exposed to the hype of AI beginning to reach a fever pitch: with experts (who happen to be members of AI company C-suites) proclaiming grand utopian visions — available (for a price!) by no later than 2027, counterbalanced by mocking nay-sayers that claim chatGPT is merely a slightly updated version of Clippy. However, this discussion actually misses the point entirely. The conversation revolves around some point in the future: it’s about what will be, not what is.
Years ago the Pandora’s box of AI was thrown open and has already had a drastic impact on our lives and culture; the ghost in the shell is already loose and, for better or for worse, we’re dealing with the consequences.
Such bold claims must be backed up, of course. It’s crucial that we’re all on the same page as to what I’m talking about and one of the main difficulties when discussing AI is that it’s an incredibly broad term: it includes both fully sentient physical robots (like Sonny in I, Robot) and simple loops of code that fit a straight line to a small collection of data points.
For the purposes of this blog, we’re going to be focusing in on machine learning (ML), a subset of AI that focuses on understanding and building methods that let machines “learn”: training data is leveraged by algorithms to improve the performance of models. These models can describe anything that you are trying to predict or classify based on underlying characteristics. For example, you can design models to predict the sale price of a house based on features of the property and the current state of the market, classify individuals as likely voters, estimate expected lifetime value of a customer, or detect anomalous purchase patterns with a credit card. The point is that these models improve themselves as they take in more data: they learn what the best* answer is.
*Footnote: If we’re being extremely pedantic, we often can’t get the truly optimal answer — just a better one.
Who is the potter & who is the clay?
Although this might seem rather benign (and a far cry from the AI of science fiction), the impact of ML is difficult to understate. Every major social media platform makes heavy use of ML algorithms, deployed with the purpose of either maximizing customer experience, advertiser engagement, or profit, depending on what the current business goals of the platform are. The impacts of algorithm design and selection are massive, with adjustments able to change traffic flow to content by orders of magnitude. As much as we might like to think of ourselves as independently minded and able to reasonably evaluate content, the environment that we exist in heavily affects our perceptions of reality — not just in the context of what we believe when it comes to various issues, brands, and people; but also in which of these we think about at all. Although the grooves of a rock tell a stream how it must flow, eventually the water carves away even the hardest stone; we may create inventions, traditions, and beliefs, but eventually they sculpt us.
From the early-2010s, we began to see the impact of ML on culture as curated feeds and recommender engines became the heart of the titans of tech: Google, Amazon, Facebook, Shopify, Netflix, Spotify, and, eventually, TikTok. During their rise, each of these rivals were able to provide better search results, products, and videos, often due to far better algorithm design than those that came before. However, this transition away from chronological feeds and simple subscriptions came with a host of externalities, some good, others terrible.
The political world has been forever changed: grassroots and astroturfed movements alike can grow incredibly rapidly — recommender engines tend to love engagement and stories (both true and false) that elicit strong reactions, which causes them to get pushed to more people, which causes more engagement, etc. As the Covid-19 pandemic rather cruelly revealed, public health and the dynamics between governments and their populations is a completely different arena than in past centuries. The corporate world is heavily impacted by this too, with many businesses being wholly dependent on remaining in a top search result position.
In a sense, these algorithms are the specters of the modern era — ephemeral and invisible, but nevertheless impactful. We don’t tend to think of these blocks of code as AI, but they’re the most real example of that we have: they feed into, and, in a very real sense, define some of the limits of our reality. This, of course, leads to a very natural question: in this brave new world, how do we operate? As individual, groups, businesses, and governments, how do we navigate a landscape filled with ML?
In order to plan, we first need to understand. In the next blog we’re going to take a journey back to where we begin to pull the curtain back on AI and machine learning and take a glance at all the machinery that makes things work.
Read Part 2 of Dr. Data Sciences’ guide to AI and Machine Learning now!