The future is always uncertain – that’s what makes it both so exciting and so terrifying. Long term, we don’t know much about what’s going to happen. However, what if we could predict the future in the short term? The editorial team at the Economist took a unique approach to getting answers for 2020. Aside from their own predictions, they also looked to AI to let them in on some secrets. The takeaway: want your fortune told? Consult AI.
About a year ago, non-profit OpenAI announced that it had built a neural network capable of natural language processing, named GPT-2. The tool was not immediately available to the public, for fear of misuse, and was released in phases over a number of months.
Rather than feeding the neural net garbage, they used articles that had received more than three upvotes on Reddit (which doesn’t exactly guarantee high quality, but whatever).
The dataset was made up of around 40GB of text. 1GB of text is about 130,000 double-spaced MS Word pages – so the neural net had a lot to work with.
GPT-2 has no understanding of the words that its generating. It uses language patterns and word sequences to build a guidebook. It then uses that guidebook to answer questions, or predict what will come after certain word sequences.
GPT-2 was used in the Economist article. Although they describe its answers and “unedited,” deputy editor Tom Standage clarified their methodology in a Medium post, and it definitely involves some tweaks here and there.
Standage wrote an intro, and asked the neural net the following question to kick things off: “Greetings, GPT-2. Have you done an interview before?”
He then configured the tool so that it would respond with 5 answers, and they would pick the best one.
Standage and GPT-2 continued:
“Do fake news stories, generated using AI, pose a threat to democracy? Are we likely to see this tactic being used in the 2020 American presidential elections?
“A: Yes, we’ll probably see them. It’s just a matter of when. Fake news stories are generally generated by political campaigns, and have the potential to have a huge impact on the outcome of an election. This is because they spread misinformation, often with no basis in fact.
“Q: How do you see the economic outlook for 2020?
“A: I think it is very unclear. The outlook for the UK economy is uncertain, but the American economy is very strong. I think we are going to have a lot of turbulence in the world economy.”
Not real predictions
So, if you want your fortune told, consult AI, right? Clearly, the neural network has no real capability of predicting the future. What GPT-2 has gotten good at is predicting which words will appear sequentially, and that’s not too shabby. With that being said, AI was built by humans – and much like us (or most of us), it has no fortune-telling abilities. Still, I did have a good chuckle about going to visit an AI fortune teller at a carnival – imagine that.