Humans are no longer required to write basic news
The Olympics are massive, and it is almost impossible to catch everything. There are 306 events throughout the two-week long event, 11,000 athletes in play, and 206 countries represented. It’s a lot to cover, even for news outlets with hundreds of employees and international reach like the Washington Post.
To combat this, the Post decided to employee a rather unusual method of covering all the events and reporting the results – it is using an AI program to write some of the stories.
The program, known as Heliograf, is overseen by three full-time engineers. It is responsible for short blasts, which in theory allows its human reporters to cover the larger stories and actually give some additional details.
“We’re not trying to replace reporters,” said Jeremy Gilbert, the Post’s head of new digital projects. “We’re trying to free them up.”
The posts will be purely informative, and they will probably focus on things like medal winners in sports where the Post’s primary audience – Americans – won’t have a vested interest. In other words, expect AI-written news like the short clip below:
China’s Long Qingquan wins weightlifting gold in men’s 56kg with a lift of 307 kgs, beating North Korea’s Om Yun Chol by 4 kgs. Sinphet Kruaithong of Thailand wins bronze with a lift of 289 kgs.
Handing off the grunt work to the AI allows the Post’s human team to go a little more in-depth and write about competitors like Hungary’s Katinka Hosszu, who absolutely shattered the 400 IM world record by nearly two full seconds and earned a feature article. In this instance, for the Olympics it makes absolute sense to use an AI-like program. It’s a little sad that the Post won’t bother to just syndicate the news of someone like Quingquan from other outlets – he may have a fascinating life story Post readers will never know – but it makes sense.
It also sets a troubling precedent.
If all you want is results, AI news is fine. For anything bigger though, AI can’t possibly hope to catch anything beyond the facts. For some, however, that might seem like a good thing.
Imagine using AI to cover politics – the software might simply identify a politician’s actions – maybe they gave a speech to a crowd in a certain location. That would potentially make political coverage a little more even, but it would also be incomplete. An AI couldn’t discern things like how the audience reacted, or what new policies the politician mentioned. It also wouldn’t be able to discern the differences in inflection, nor could it fact-check the candidates.
It’s an appealing idea to eliminate any human bias from news, but it’s also an impractical one. It also doesn’t eliminate the problem with modern news. Most news outlets aren’t biased in terms of ideology, they are biased in terms of putting out the most entertaining or attention-grabbing story for their particular demographic. Sometimes that means ratcheting up fears, sometimes it means pointing out the fallacies of others. It may even be in finding a new angle to tell the same story, like the feeling of the crowd or what other countries think about foreign news.
Sometimes though,real news gives us a better look at the truth than a completely analytical and unemotional view can offer.
At the moment, this is only a prototype using a rudimentary form of AI to report simple stats and results. Hopefully that’s where it stays.