Mr. Brightside
In 2018, we live in a world where political entities use ill-gotten data to "whisper in the ears" of individual voters using targeted advertising. It's scary, and you probably haven't heard the worst of it.
The "whispering" in question was still written by a human. Which means it wasn't optimally persuasive, just as persuasive as a human could make it. The amount of it was also limited by the time a team of humans takes to write persuasive content. AI is about to change all that.
It's incredibly important, right now, that we make the right decisions about the AIs we develop, what data they have access to, and who uses them for what. Much more important and subtle than this, we need to envision the kind of future we want and the role AIs will play in it.
The good, the bad and the AI
What will it feel like when an AI tries to persuade you of something? Imagine watching HD video of a world leader saying something they never said. Imagine credit card phishing that feels exactly like a conversation with your spouse.
Accepting that this technology will exist and will be used, we have to begin thinking about two things:
1. How do we imagine and design towards good outcomes, rather than focus on the bad ones
2. How do we defend against the bad outcomes that some people will design anyway
What quickly becomes apparent is that these two things are really the same thing. The future we can imagine is the one we will make happen - so it's important we imagine a good future.
Imagining the good
Imagine we lived in a world of perfect communication. One where the most important information always spread fastest, people were always equipped with the facts, and having made up your mind on an issue or decision, there was an efficient means of clearly acting on your convictions while maintaining exposure to compelling alternative views.
Let's break this down and explore ways of getting to a world of better, if not perfect, communication.
Spread the most important information
In a social-first world where fake news designed for our biases racks up shares and travels faster than headlines, it's important that popularity is not the only fuel for online reach.
Transparent consumption. If the reader profile of every article (or the slant of every publisher) was publicly visible, people would be faced with the bias of the content they consume. Better armed with this knowledge, people could choose to read around multiple sides of an issue.
Establish facts
In a world where AI might enable HD audio or video of events that never occurred, it's critical that people can tell fact from fiction.
A media blockchain. If every piece of media - such as video of a public interview - was issued a token upon creation, distribution could be tracked back to the time & place any event occurred. Using this technique, "original sources" of raw content could be easily established and used to verify all news
Frame the key issues
In a world where AI might produce a universe of concocted arguments - think conspiracies for everything - it's critical that the issues that matter are prioritised.
Issue selection. If publishers collaborated on the creation of content around a central list of core issues, which could be actively voted up and down by people, arguments and counter-arguments could be better navigated. Additionally, policymakers could easily enter the discourse or propose action.
Arm advocates on multiple sides
In a world where AI can use the information about us to make messaging hyper-persuasive, it's critical that this technology is not monopolised by a select few.
Access equity. Persuasive AIs which target, edit or generate language and visuals to make narratives appealing to specific people cannot be ruled out. So they must be made available equally to all sides of a debate or issue.
What we're doing about this
Persuasive AIs have immediate commercial use for brands as well as a social responsibility to people and governments. We're focusing on building an AI that yes, helps sell stuff, but also helps people understand the value of things and make better decisions about what to do, including what to buy.
We hope that this technology will one day be deployed in the service of not just products, but ideas that need to be heard, understood and acted on.
