Machines will soon program people

Image source is Captain Ribman

We’re on the verge of a revolution — very soon, computers are going to start programming us, through ideas, culture, and eventually, our DNA.

We may have no idea this is happening to us.

To understand this, you really should start with the article “Artificial intelligence chatbots will overwhelm human speech online; the rise of MADCOMs.” There, I explain how emerging AI technologies will enable machine-driven communication tools (MADCOMs) that dynamically generate content for marketing, influence, politics, and manipulation. These MADCOMs will be running influence campaigns 24/7/365 all across the social web. But since the MADCOMs won’t be able to differentiate the human accounts from the machine-driven accounts, MADCOMs will run information ops on machines and people. The machines will talk back and run their own influence campaigns. The end result is the Internet being swamped by machines talking to other machines.

Much of this content will be dynamically generated. Sure, humans will configure the AI tools and give them objectives, but their content will evolve based on machine learning. And as they communicate and influence other machine-driven accounts, the MADCOMs behind them will evolve their content as well.

The end result could be machines becoming the driving force in our culture.

AIs are already creating news articles, novels, music and screenplays. Soon they will create memes, write jokes, drive political conversations, and promote celebrities. They will probably be jabbering away on Reddit and 4Chan, trying to convince humans that Coke is the real thing or that 9/11 was a coverup. They will be spinning all sorts of wild tales.

And in doing so, our creations will be programming us, through culture.

In thinking about thinking, I have always thought that we overestimate our free will and are in fact a sum of our sensory inputs and inner dialogue. We have a basic hardware — our personality in the Myers-Briggs fashion — and that doesn’t change much over time. The theory used to be that it was fixed at a very young age, perhaps in utero. But recent research shows personality is highly malleable throughout life. Regardless, we can think about personality like hardware — you might change your computer many times throughout life, and it’s a bit more persistent.

But things like identity, values, character are the software that runs on top of personality and those are malleable based on inputs. We can be programmed, so to speak, by the data, information and knowledge we absorb throughout life. We are very much the sum of our sensory inputs.

When machines start dynamically creating content we consume, they will program us through this content. They will have an active (yet non-sentient) role in shaping culture and ideas. That may in fact teach us that we aren’t as sentient as we think. Perhaps we are just really good at taking inputs from the world, integrating, connecting and mashing them together, and then spitting them out in a way that makes it seem like we’re actually directing our behavior, when in fact they are mostly the sum of our inputs?

At the base of that is choice. I think this is the only thing that drives our sentience and it’s at the root of all behavior. We may not really choose a particular action — it is pre-determined based on our past inputs. But we can choose what we pay attention to now and the cumulative effect of that drives our decisions later.

Perhaps this is why humans tend to get more stuck in their ways as they get older. The balance of their attentional choices are behind them so it’s harder to reprogram the system.

This idea is related to concepts in the article by @dweinberger article “Alien Knowledge: When Machines Justify Knowledge,” that explains how machines will know things about the world we can never understand and they might never be able to explain to us.

So if that knowledge is hidden, and machines can program us through dynamic content that modifies our culture, it could imply several hypothesis:

  1. Machines could teach us this hidden knowledge by programming our culture. We wouldn’t even know they were doing so, until we realize we know what they are trying to teach us.
  2. Machines could easily manipulate us if they could hack our cognition (deep learning excels at finding patterns in complexity, and human cognition is a complex system). I wrote about the psychological and persuasive techniques used in computational propaganda in the report “Can Public Diplomacy Survive the Internet?
  3. We’re entering into a new cultural age where machine inputs count as much as human inputs. We will create a new and totally uncharted culture together. Will this be the new Renaissance, or the next Inquisition?
  4. Humans will use MADCOMs to wage cultural warfare on each other. Different groups with agendas will configure AIs to run operations to shape society with the culture they want to live in. Those not using these tools? Well, cultures using inferior technology rarely fare well when confronted with technologically superior civilizations.
  5. AIs will see heavy use in medicine. They may know things we don’t. We will probably use them to alter our genome. They might make changes in our genome we don’t/can’t understand that actually change our cognition, behavior or physiology. They could literally re-program us from the DNA up without all that messy culture stuff.

Buckle up folks. The 21st Century is going to be a wild ride.

For more info, read this great article from Scientific American: Will Democracy Survive Big Data and Artificial Intelligence?


This is a Short Byte, a snippet of technology thinking designed to ignite ideas and promote feedback and discussion. Read more Short Bytes

Please give the heart a click. It shows you enjoyed this article and makes it available for others to read.