AI, paperclips and the end of the world…
Recently a game about making paperclips went viral on the internet… In the game, you play the part of a super intelligent computer, whose sole objective in life is to manufacture paperclips…
You’re probably thinking this sounds dumb? How could this ubiquitous paper fastener contribute to the end of the world?
In fact, the game is analogous to a very important principle within Artificial Intelligence. It serves as a cautionary tale for companies trying to implement AI, and the perils of chasing a metric through automation.
The aim of the game is very simple: increase your paperclip inventory no matter the obstacle. This is achieved by completing a series of puzzle and tasks; like mining resources, creating machinery and setting up new factories.
Next your AI persona expands into the stock market. You start managing a portfolio of investments, and even a trading strategy. Your ROI increases and your profits sky-rocket. Soon you have taken over the world’s paperclip market.
This is when the game takes a turn…. you start to run out of the required materials. It’s not long before you have stripped the planet of all its natural resources. The only way to proceed further is to set off into space…
By the end of the game you are managing an infinitely expanding space fleet, set to exploit every inch of the galaxy in the pursuit of the materials for making paperclips….
The moral of the story is; even the most benign applications of AI could end in unforeseen disaster. No matter how smart your technology is, if you give it a narrow-sighted objective even the best intentions could end in catastrophe for your company/brand (/planet).
The paperclip scenario might sound ridiculous. However, there are countless examples of this phenomenon already.
One of the most pertinent examples is the programmatic ad bots – they’re designed to scrawl the internet placing adverts and maximising reach, but have resulted in family brands featuring on porn sites, ISIS videos and Jeremy Kyle clips.
Then there are the algorithmic pricing bots, on sites like eBay and Amazon. These have been known to inflate prices to the point that no one purchases the product anymore.
Let’s also not forget the high-frequency trading algorithm that lost Knight Capital 440 million dollars, in an impressive half an hour…
So, what’s happening here? Is AI dumb? No, not at all.
We are in a renaissance of artificial intelligence at the moment. An abundance of data and open source communities are creating leaps and bounds in the field. It is getting better and better every day, and will be replacing humans in more and more tasks.
AI’s strength is in very specialised tasks. However, it’s currently weak when it comes to generalised intelligence, or “common sense”. An algorithm will not have the sense to ask, “When should I stop making paperclips?”.
For marketers, this means you need to work on your brand and proposition. There are great efficiencies that can be gained from using AI, but you can’t rely solely on an algorithm to find your audience.
At the end of the day, the human brain and computers are two vastly different architectures, excelling at different kinds of things. Humans and intelligent systems will forever be interdependent, bound together in a continual exchange of information and goals, a “symbiotic autonomy.”
Richard Brooker – Data Scientist, Mindshare UK