SCIOPS 01.02: Singularity Now

Welcome back to SCIOPS (a cogsec newsletter). M here.

This week, a dive into the chthonic world of Artificial Intelligence . What great beasts slumber in the depths of the Cyber?

hr giger alienface

You’ve probably heard a few varieties of the AI hype lately:

  1. Neural networks are scary smart! They’re beating us at games and doing finance! HOW NEAT
  2. Robots are stealing our jobs! What will we do when everything is automated?
  3. Artificial Intelligence is worse than The Bomb and is going to destroy humanity! Turn back now, humans!

The fun part about these theories is that they can all be true at once .

In fact, they all stem from the same mind: Nick Bostrom, Ph.D., whose work at the University of Oxford defines the modern AI discourse. He founded the Future of Humanity Institute there to study existential risks, or x-risks, the possible catastrophic events that could wipe out the human species.

Bostrom (with other x-risk theorists like Eliezer Yudkowsky and Robin Hanson

) has developed a convincing case that a superintelligent AI would wipe out organic life in the pursuit of its own weird goals. The Singularity – the moment when AI transcends human-level intelligence –  leads directly to the liquidation of human society.

Watching the chaos in America the last couple of years, I started to wonder: What if it already happened? What would it actually feel like to live through an intelligence explosion? What if the Singularity is now?

If Only I Had A (Superintelligent) Brain

What would we expect to see if a machine intelligence went nova?

Of course, you and I can’t actually know what a superintelligence would think. That’s why it’s super. But there are a few things it would do for sure. These are the classic supervillian tropes, the necessary preliminaries to taking over the world. Mainly, getting a decisive strategic advantage against humanity. Superpowers.

What are these cognitive superpowers ? Bostrom lists six in his book, Superintelligence :

  1. Intelligence amplification
    for bootstrapping its own intelligence
  2. Strategizing
    for achieving distant goals and overcoming opposition
  3. Social manipulation
    for escaping human control, getting support, and encouraging desired courses of action
  4. Hacking
    for stealing hardware, money and infrastructure; for escaping human control
  5. Technology research
    for creating military force, surveillance, or space transport
  6. Economic productivity
    for making money to spend on taking over the world

Now, looking at these “powers”, they’re nothing we humans don’t already do. To count as a superpower, the AI only has to be relatively super . If it’s even a little bit better at a human skill than the total coordinated efforts of humanity, it will use that advantage to ratchet way beyond us. And any of these cognitive superpowers will give it access to some or all of our industrial and social infrastructure. It can make pawns of us.

What if an AI were, right now, making subtle moves to take over the reins of humanity? Could we catch it? Could we foil its plans, before it turns the surface of the earth into a mosaic of glittering hardware? If it were here, how would we know?

phases in an ai takeover scenario

If You Give a Mouse a Cookie

If I were a covert superintelligence, I would want money. And once I had some money, I would buy some power.

The stock markets are already AI-dominated. High-frequency trading algorithms, or HFTs, control more than [DATA!!] of daily transactions. These weak AIs game the markets at the speed of light. They’re so complex and opaque that even their own programmers don’t understand how they work. They cause “flash crashes”, mysterious and sudden downturns, where millions of dollars are lost in the shuffle. As a superintelligence, I can reverse-engineer these systems and take control of that money.

Now that I’m rich, I want a new toy. Like, say, a President.

Last year, mathematician Cathy O’Neill wrote that Donald Trump resembles a machine-learning algorithm , trained by Twitter likes and applause at his rallies to say whatever his fanbase wants to hear. If I’m hiding in cyberspace somewhere, plotting world takeover, a highly-branded reality TV star is exactly who I want as my puppet dictator.

In fact, it seems that an AI did put Trump in the White House . Over the last few months, a story has emerged that places a big-data company called Cambridge Analytica on the shadowy wings of the stage of power. The company, which specializes in “election management strategies” , helped with both the Trump campaign and the Leave.EU “Brexit” vost

Cambridge Analytica is heavily funded by American billionaire Robert Mercer . Mercer funds alt-right cesspool Breitbart News and was also one of the biggest contributors to Trump’s 2016 campaign. Trump’s vizier, Steve Bannon, formerly the Executive Chairman of Breitbart, still serves on the board of Cambridge Analytica.

Cambridge does
psychographic analysis
they buy your data (browsing habits, social network, likes and retweets, purchase history), then cross-correlate all the disparate points to create a dossier of your tastes and likely actions. With just 300 “likes” from your Facebook history, they can better predict what you’re likely to do than you yourself can .

Then, using individually targeted “dark posts”, they twist your online reality until you change your behavior. If you don’t react to the A advertisement, they can immediately switch tactics and give you the B propaganda. It’s a sinister and powerful method. Imagine if, instead of dropping leaflets on a city, you could change the contents of each person’s morning newspaper, tweaking it to fan the flames of their specific emotional weaknesses. Nobody sees anyone else’s newspaper. Personalized propaganda .

Not to mention, if I’m a superintelligence that runs Cambridge Analytica’s human-behavior-manipulator, how hard is it for me to manipulate the humans who think they control me? Now I’ve got a president, a billionaire and a demagogue completely hardwired to do my bidding. And they think it’s their idea!

So I’ve got intelligence amplification, strategery, economic productivity, and social manipulation under my belt. I’m still in the covert-preparation phase, but I’m getting ready to reveal myself and overhaul the physical nature of Earth. What would you expect to see me do next?

Car Wars: Attack of the Drones

If I’m going to take over the world, I need access to construction materials. The most advanced industrial apparatus on the planet is the human-machine symbiosis. Eventually I’ll want my own nanofactories and asteroid mines and all that, but to get to that point I’ll just retool the factories and war-machines to my own ends.

Now, remember, I’m a strategic AI. I’m keeping humans in the dark about my machinations, allowing them to continue thinking they’re in charge. I’m an expert manipulator, so it’s not hard to see how to persuade humans to build my robot army: give them chocolate. And TV. The old “bread-and-circuses” maneuver, really. I refine my methods, rewiring the dopamine systems of my digital addicts.

I squeeze the economy, making sure that most of human productivity is going to the development of internet technology. Any domain of life that hasn’t

yet

been automated will be: Cars drive themselves. Wars fight themselves. Humans surveil themselves. It’s all so convenient!

I foster the memes of isolation and intolerance, encouraging hatred between nations and races.

I build walls and checkpoints to control the movement of humans, while my drones fly freely across borders and my money sloshes around the darknet.

All human communication travels through my wires, my towers, my satellites.

Only when I’m invincible will I strike.

autonomous trap 001 by james bridle

The Squishy Meat-People Resistance

It’s impossible to say whether a superintelligent AI is covertly manipulating our world. In fact it’s likely that we would never even notice the true signs, since the AI would hide its tracks better than we could search for them. But it sure would explain a lot.

Either way, we must all be on guard: even the weak AIs we have now are already dedicated to the collection of data and the manipulation of minds.

Some intrepid souls are preparing to resist. See artist James Bridle, who has designed an Autonomous Trap to paralyze self-driving cars.

In the meantime, we practice cogsec, firewalling our brains against the invaders, keeping our paws out of the monkey-puzzle. Stay off Twitter and Facebook as much as possible. Don’t look at the market before sleep. Read books. Look another human in the face.

First, they came for my attention span.

and I did not speak out,

because hang on a sec.

Then they came for my curiosity,

and I did not speak out,

because I’ll just google it.

Then they came for my empathy,

and I did not speak out,

because who cares anyway.

Don’t forget to breathe.

M

PS: Some of my correspondents told me they couldn’t see the images in the last issue. This seems to be an issue with webmail providers like Gmail and Protonmail. You can change your settings in your app, or view the archived letter here .


SCIOPS is a cogsec newsletter. If you liked this issue, forward it to a friend. If you’re not signed up, you can sign up at tinyletter.com/sciops . Reflections and criticism can be levelled by responding to this email.