We already live in the singularity

“The Singularity” is a term coined by John von Neumann, a major figure in the history of computer science. The concept refers to a hypothetical time when computers become more intelligent than humans and can improve themselves without our input. Imagine a run-away reaction where artificial intelligence is able to improve itself. This improved self is able to further improve itself. With each improvement the rate at which improvements are made increases. Eventually, humans will no longer be able to compete.

How intelligent could it get? Where would that leave us?

A more general definition from Wikipedia is:

When technological growth becomes uncontrollable and irreversible, resulting in unforeseeable changes to human civilization

Hold up that sounds a lot like today!

What it means to be in the singularity

When it’s expressed in popular culture, in movies like “Transcendence”, “Singularity” or more recently “The Mitchells vs. the Machines”, algorithms get so smart they violently rebel from their human overseers and enact some sort of genocidal third-act rampage that our heroes must stop or escape. This would certainly be a singularity by the definition above, but it’s not what we see around us every day. Companies like Facebook demonstrate how technology is already causing unforeseen and uncontrollable changes to our lives. For the sake of this article, let’s look at the singularity differently.

I will define The Singularity as:

When technology becomes a controlling and inescapable force in the lives of many people

This may stretch the classical view of the singularity, but I say the classical view is outdated. We need to look at the concept of “The Singularity” differently.

The key difference here is that even though humanity as a whole is still in charge of these modern machines, Facebook has the power to turn its platform off if it REALLY wants to, large parts of the population can not turn the algorithms that control their lives off. Those uneducated about its methods or denied the means to learn, are often at the mercy of the algorithms we build. Just a survey of your everyday life shows work is found through LinkedIn and done on Slack. Schooling is administered through Canvas and taught on Zoom. Friends are found on Facebook. Entertainment is watched on YouTube. Relationships blossom on Tinder. Our lives are increasingly algorithmically determined. For the vast majority of us, we live in The Singularity.

Singularity of Truth

If you reading this, chances are I do not have to explain how technology has warped the ability to form an objective truth of the world. Algorithmic feeds give us a narrow and curated look into the world that reinforces biases, promotes outrage, and further polarizes us.

A 2020 study by the American economic review had participants deactivate Facebook for 4 weeks and measure their political leanings and quality of life before and after. Not surprisingly, members voted more, were less susceptible to fake news, had more in-person activities, and overall reported high quality of life. Polarization fell and human connection rose.

This revelation that social media polarizes us is nothing new, but the ways it does so is much more subtle than people think. Online platforms construct digital personalities of their users from post history including dozens of data points your device gives the websites every time you visit and hundreds of additional data points platforms are able to deviate themselves. These small pieces of information help to determine which ads best appeal to you and which content would be best, effectively constructing the view of reality that you agree with best.

Advertising can then be tailored to how the platform views you, allowing the same product to be portrayed differently to different people. Ads are often not a single item but a multitude of various takes on the product that the platforms then determine who would react most positively for. Matching the ad to the user can see 5x or more clicks, good for advertisers, but it means that each person sees a reality different from the next. The same actions are taken with content on the platforms, adjusting what you see, how it’s said, who says it, and 100 other factors to make sure you respond best.

This example popularized by “The Hype Machine”, When users are classified as Extraverted through tracking, they are more than 5x more likely to click on “Extraverted” ads.

But social media is not the only place truth is controlled by machines. I myself participate in Natural Language Processing (NLP) research, the field focuses on teaching computers to read, understand, and most importantly generate text. The intricacies of this are rather fascinating, but more generally computers read in billions of samples of text from all across the internet, books, newspapers, and even specially generated datasets. They then try to replicate what they read as best as possible, then tailor it to the reader, oftentimes to frighteningly good effect. The example below shows to what extent this can be done with consumer-level tools. More powerful tools exist in the hands of large companies who spend many millions of dollars to generate them.

The bold section of the below quote was written entirely by a machine (GPT2, here). The first sentence was written by me as a prompt and it continued on with no edits.

[Computers algorithmically warp our sense of truth, especially politically.] Some programming boards are even programmed with the intent to abuse the trust that we give them. This gives the big tech companies an out-of-control incentive to flood us with news, entertainment, information, and as many marketing messages as possible. This brings me to one of the largest technology companies, Facebook.

Jarvis, a different AI-driven content generation company, writes blogs, social media posts, and marketing, all while charging ~$30 for 20,000 words. A professional copyrighter, from somewhere like Contentfly, would charge more than $2000 for that same 20,000 words. Is the quality the same? Of course not, but it’s getting there, and for the vast majority of users it doesn’t matter. Using tailoring techniques to match the content to the user often means that quantity is more important the quality.

How can we know what’s true if the truth is auto-generated?

Video too is increasingly being forged. Deepfakes require only a few reference photos to generate a convincing enough video. In the last few weeks, there have been numerous celebrity impersonations online. Tom Cruise and Charli D’Amelio both found digital doppelgangers of themselves going viral on the app TikTok. With the right software, anyone can make themselves look like whoever they want. Here is that technology applied to the Mona Lisa. Is it perfect? No. But it only has to be good enough, and they get better every year.

Everywhere we go, reality itself is being controlled by machines. Truth is a matter of perspective and every window is crooked. This is what it means to live in a modern Singularity of Truth.

Singularity of Work

Work is facing the same fate in a way that is less well known. Automation is gobbling up jobs at a rate never before witnessed while producing few new jobs. A laundry list of technologies (self-driving cars, personal robotics, high-power drones) stands to decimate the existing global job market in the next ten years. And since, by its very nature, automation makes sense because it increases productivity and lowers costs, for every 100 jobs lost to automation far fewer than 100 jobs are created. We often picture this as massive autonomous factories, and to an extent this is true, but that is only one part of the story. Management positions, resource allocation, supply chains, all are no longer human-run endeavors.

The clearest example of this is found in the gig economy, the informal name given to services rendered through apps like Uber, Doordash, or even Fiver. People in these jobs sign up to provide a service and then are given tasks as needed by the algorithm. Did someone order some food? Here, go deliver it. Does someone want a ride? Go pick them up here. At current, Uber is doing 2 BILLION trips a quarter. On average, every person on the planet takes 1 Uber ride a year.

If you treat Uber drivers as employees, something Uber really doesn’t want you to do, Uber is the third-largest employer in the United States. About 1 million people make their living this way. And how exactly is the “uber-way”?

The MIT tech review summarizes it like so:

When a driver is logged in, the app assigns them pickup requests from people nearby. The system metes out feedback by tracking the proportion of pickups a driver accepts (Lyft and Uber each give drivers 15 seconds to decide), and averaging the rating that passengers give their driver after a ride.

Put more simply, the algorithm looks at your performance history to determine if you are fit for the ride, then gives you 15 seconds to answer yes or no to the trip. However, if you say no, it will remember for next time. Maybe next time it will not think you are as fit. In this way, your boss is not “Uber” but “the Uber algorithm”

This idea of “Your Boss is an Algorithm” is even more evident on the video-sharing platform YouTube. Content creators on YouTube receive a share of the revenue from ad views, with top YouTubers making $10,000,000+ and many thousands of other creators making a more modest living supporting their content niche. In this way, YouTube incentivizes high-quality content, grows its platform, and allows the thriving of small and diverse communities. This sounds simple but the process is much more gray.

YouTube wants to keep viewers’ attention for as long as possible, to show more ads and therefore receive more ad revenue. To this end, YouTube recommends videos that are displayed on the sidebar that updates every time you view a video. Phrased as a simple: “Hey, here are similar videos” lurks a complex algorithm that takes in everything google knows about you and tries to predict which video from the entire platform would get you to stay on the site the longest. YouTube does have a search option, but most views (more than 70%) on the platform come from this recommended section. To make a successful video, you need it to be recommended to people.

This is a detailed dashboard is provided by youtube to content creators to help them track metrics for their videos

As a content creator for YouTube, who do you answer to? Is it to your fans? To YouTube itself? No, ask any content creator and the answer is the same, you answer to the YouTube recommendation algorithm. If your video is not liked by the algorithm then it will not be shown in recommendations and as such you will lose millions of potential views and thousands of dollars. For a full-time content creator, not making content geared towards the algorithm is a death sentence because your fans will simply not be able to find it. Optimizations like this are so important to creators that entire companies (Oneupweb) exist to work with creators to make sure their content is correct.

How we find and do work is increasingly being managed by algorithms. Employment is a matter of finding favor with platforms so they may smile on you with rides, views, and money. This is what it means to live in a modern Singularity of Work.

Singularity of Money

Money is also increasingly controlled by algorithms, creating the Singularity Trifecta of Truth, Work and Money. The very underpinnings of the financial system are no longer controlled by some corporate banker in their office filling out paperwork, but in massive automated networks of computers trading assets with other computers. Markets are increasingly driven by hype and speculation and no longer by real human activity. For those not in financial systems, this may seem like someone else’s problem, but unless you live off subsistence agriculture, you are directly influenced by the global economy with every good or service you use. You are trapped in the Singularity of Money

In May of 2010, the Dow Jones Industrial Average, one of the largest and most important indexes in the united states, fell by 10% without warning, wiping out nearly 1 Trillion dollars of value only to recover by 7% within 10 minutes. This is what is considered a “flash crash”. Similar events have occurred with remarkable frequency in recent years as markets seem to suddenly spike up or down for seemingly no reason before snapping back to reality. What is happening?

At first thought, it could be a bug with the way prices are calculated, but this is not the case. Millions of real buy and sell orders for specific securities come in all at once, at such high volume that the price is dramatically shoved into a new trajectory. Was this a coordinated effort to manipulate markets? Oftentimes no. In reality, there are thousands of computers that have learned to make money by reading financial news and trading stocks based on these reports. For many of them, this is an incredibly profitable endeavor. But sometimes, the bots get it wrong. A misleading report comes out that confuses the algorithms in just the right way. Suddenly, they all decide to buy or sell stock in unison, only to be immediately aware of their mistake and reverse their decision to prevent losses.

This all becomes fun and games when these economic swings are just moving digital numbers up and down but they are having a real impact on people’s lives. Stocks are a cornerstone of retirement, of the government’s own financial security, and of the financial security of the companies the stocks represent.

Cryptocurrencies are aiming to revolutionize money. Innovative cryptography and a decentralized network of validators remove power from central banking organizations and from governments. Given their new and often volatile nature, flash crashes often hit harder.

From Bloomberg:

Eye-catching, mistaken plunges are not unusual in crypto. In June 2017, Ether had a harrowing 45 milliseconds that took its price down to 30 cents — a 99.9% loss. This week, it was Bitcoin’s turn to flash crash: The biggest cryptocurrency dove to $8,200 within a single minute on Thursday from around $65,000 on Binance’s U.S. exchange after an institutional trader’s algorithm encountered a bug and cratered the price.

Outside of flash crashes, cryptocurrencies find themselves susceptible to other types of involvements. One of cryptocurrencies’ main claimed advantages is the ability to have financial policy managed by code directly. Code can be deployed directly into the blockchain to allow automated and verifiable financial actions to be taken. This is intended to remove the corruptible human element that has caused much of history’s financial collapses (2008 is a big example), but at the same time, it builds a financial system that is purely algorithmic. In effect, we have given control of our money to machines to remove the corruptible human element. I’m not sure how I feel about that.

Money is something that may have the most justification for an algorithmic future considering how much strife humans have caused when messing around with the financial system. However, that does not mean the current approach is without faults. Algorithmic trading has brought forth a fragile and unpredictable system of machine control over money. Cryptocurrency has promised to be different but is shaping up to be a continuation of the Singularity of Money

Escaping the Singularity

I wish I could end this article on a hopeful note. I wish I could tell you that I had an algorithm to thwart the algorithms. But these issues cannot be easily solved by one person. Each of us must try to understand technology. Much of the control algorithms have can be removed, if we understand how the algorithms work. We must stay informed as advancements are rolled out. Read scientific papers, think critically, and help others do the same.

This was originally published on medium here