The prospect of robots taking over the world has been floating around in science fiction and popular culture for decades. Many of us might imagine some sort of machine uprising or super-intelligent revolutionaries gone out of control. But at least for now, these fears have yet to become reality.
But we shouldn’t write them off either. There is already an alarming amount of influence that robots—artificial intelligence (AI) algorithms, rather—wield over the world and they do so in uncountably many ways. Merely one of them, which has grown significantly in recent years, is natural language generation.
If you’ve ever used Google Translate, chatbots, or a voice assistant, you know exactly what this is. It’s AI trying to generate sentences—to write. Write fluently, that is.
And this fluency is what has made natural language generation increasingly powerful in recent years—so much so that many news articles you read today are already being written by algorithms rather than humans
In fact, even well-established organizations like The New York Times, The Washington Post, and Associated Press have been using the technology for many years, helping them churn out stories and articles with unparalleled speed.
And quality-wise? Well, the writing by AI has progressed far beyond the formulaic or robotic rigidity that we might expect from robot writers. Even poetry and creative writing are not too much for them.
Learning What It Reads
Trained on millions—even billions—of web pages, books, and just general text online, many of these algorithms use a structure loosely based on neural networks in the brain to improve themselves, mimicking whatever words, phrases, and sentences they come across. In a sense, the way AI writes is a reflection of how humans write.
In 2020, OpenAI’s GPT-3 language model produced text that was sometimes completely indistinguishable from human-written text, causing researchers to restrict the accessibility of the model to the general public. This was doubtless done in the hopes of preventing malicious individuals from using the model to endlessly generate text that could influence people in a dangerous way.
If AI can already write simple news articles, then it isn’t a stretch to see how it could also write misleading articles or propagate harmful information online.
GPT-3 the Artist
To test its creative writing ability, researchers gave GPT-3 the title of the poem, “Shadows on the Way”, and asked it to compose in the style of Wallace Stevens. This is what it wrote:
Nobody will come to this place. It is a road that leads nowhere.
The solitude is deep. The mountains are high.
But they are desolate, and they turn the traveler’s face
Towards the North. All the sounds of the world are far away.
When the wind rises above the trees,
The boughs bow to the ground.
Even the birds that inhabit the tangle of weeds
That is the roadside cover, are silent. One listens,
But hears no roar of the forest. One is alone.
One will be taken.
One will be taken.
There is no utterance, there is no conversation,
But one is uneasy all the same . . . .
There is a thin blue mist,
A darkness rising like smoke,
And within that darkness
A possession of the heart.
One will be taken . . . . It was here, and it will be here again—
Here, under this sky empty and full of light.
What to Worry About
AI writers have provided us with summaries of articles in an instant, entertained us with games like AI Dungeon, or helped us write through predictive word generation (see Google Docs’ and Gmail’s Smart Compose). Likewise, numerous companies are already using the technology to generate short financial reports or create marketing material in a fraction of the time—and cost—it would take a human.
Just as this technology can act to our benefit, however, it can also act to our detriment.
Influencing public opinion and spreading misinformation could be done simply through the massive amounts of text that can be generated in mere minutes, flooding the internet with AI text designed to read like writing from a trustworthy source.
Ultimately, it’s important to recognize that despite the increasing influence and ubiquity of these robot writers, they have yet to escape human control. We probably aren’t going to see a machine uprising—like those imagined by science fiction—anytime soon. What we will see, however, is more and more malicious actors using these tools in dangerous ways.
And that is something to look out for.
That should be a source of worry.