What’s In Store for 2020? Here’s What AI Says

March 2, 2020 - 8 minutes read

What does the near future hold for humanity? Who will lead our countries? What technologies will transform our society? And how will the global economy fare through it all?

Each year, the editorial team of The Economist takes a swing at answering these questions. For 2020, they decided to change things up a bit. Besides making their own predictions, the group enlisted the help of artificial intelligence (AI) development. More specifically, they asked some hard queries to an AI system that has stirred up quite a bit of controversy.

The AI’s answers to questions about politics, technology, and economics may surprise you.

The AI That Knows All?

Approximately a year ago, the San Francisco-based developer and non-profit research organization OpenAI announced that it had created GPT-2, a neural network for natural language processing (NLP) that could “write like a human.” The news was met with both dispute and fear.

None of this was alleviated when OpenAI also declared that GPT-2 would not be released to the public immediately due to its immense potential for misuse. Instead, it was released piece by piece over the next several months.

GPT-2 wasn’t just trained on any low-quality text. Its creators only used articles which received more than three upvotes on Reddit. Granted, this isn’t exactly a dependable bar for high quality. But it’s (probably) better than nothing. GPT-2’s training dataset ended up being 40GB of text. To put this in perspective, 1GB of text is roughly 130,000 double-spaced pages in Microsoft Word!

The tool itself doesn’t actually possess any understanding of the text it generates. It depends on patterns in language to derive statistical associations between words and phrases. By doing so, it builds a “guidebook” for itself (similar to the ones you’d use to learn a foreign language). From there, it leverages this guidebook to answer questions. GPT-2’s developers have pointed out that its replies become more convincing when it’s trained on narrower datasets for specific applications.

The Economist deputy editor Tom Standage conducted the interview about 2020 with GPT-2. He’d write an introductory paragraph followed by his main question. GPT-2 would then generate five responses. Standage would select the best one.

The AI Q&A Session

Okay, let’s get to the moment you’ve been waiting for: What burning questions did GPT-2 answer? Let’s go over a few of them right now.

The Future of Technology

When asked “Which technologies are worth watching in 2020?”, GPT-2 admitted to having trouble narrowing down the list. But it did say that AI was the most important one due to its versatility. And it’s only becoming more powerful every day.

In a follow-up question about the future of AI, GPT-2 emphasized that the technology is a tool that must be used responsibly. Rather than concerning ourselves with potential ill intentions or consequences, we should instead focus on developing the technology as correctly as we can.

When asked if it represented a step towards artificial general intelligence (AGI), GPT-2 remained diplomatic in its response: “I am not a step towards AGI. I am interested in understanding the origins of language. The only thing that I can do is answer questions correctly.”

AI Concerns

Speaking of AI concerns, Standage’s next question for GPT-2 was “Are you worried that AI technology can be misused?” GPT-2 replied that this was definitely a possibility. To stymie it, the neural network said that powerful organizations such as governments and companies must ensure their AI creations are utilized responsibly.

Next, Standage broached the scary topic of job automation. GPT-2 pulled no punches with this one and claimed that automation would be circumstantial. In industries like finance and medicine, AI would inevitably replace people. But in many other sectors, machine intelligence will simply aid in solving hard problems.

Economics & Politics

Moving away from only tech questions, Standage and GPT-2 delved into the topics of economics and politics. Standage asked if fake news is a threat to democracy and if we’d be seeing this tactic employed during the 2020 U.S. presidential elections.

GPT-2’s response? Probably. In fact, “… it’s just a matter of when.” The neural network buttressed its reply by saying that fake news is typically generated by political campaigns and that they can have a monumental effect on election outcomes. When asked about Donald Trump’s chances of winning a second term, GPT-2 was straight to the point: “I think he will not win a second term. I think he will be defeated in the general election.”

Lastly, was GPT-2’s economic outlook for 2020 positive or negative? The machine predicted turbulence for the global economy. More particularly, the UK economy’s outlook is uncertain, and the U.S. economy will continue to be strong.

Want to see the full list of questions and answers? Check out The Economist article here.

The Crystal Ball Will Become Clearer

Since AI can process and synthesize vast amounts of information faster than any human possibly could, it’s being used to predict all sorts of things, like crime trends and epidemic outbreaks. Asking GPT-2 to predict the future based on a Reddit-curated past is certainly something new. Some of its answers were insightful, while others were quite vague. Still, one can’t deny that it was an interesting endeavor.

As AI technologies and machine learning applications improve, it’s likely that we’ll start seeing more tools like GPT-2 making more and better predictions in the future. On an optimistic note, when asked if it had any advice for readers, GPT-2 gave a concise one-sentence reply: “The big projects that you think are impossible today are actually possible in the near future.”

What do you think about this AI system’s predictions? Do you think that it’s on point, or that it was simply speaking nonsense? Would you rely on an AI to predict your future? Let us know your thoughts on this captivating topic in the comments below!

Tags: , , , , , , , , , , , , , , , , ,