BusinessDesk is now making use of artificial intelligence to near-instantly publish NZX news.
We’re using ChatGPT to write articles from company announcements as they are released, cutting the time it takes to publish an article from half an hour to under a minute.
This allows our newsroom to concentrate on what we’re great at – discovering, writing and publishing news others do not have.
Artificial intelligence (AI) stories will be clearly marked as automatically generated with the byline 'BD AI' (I created the reporter image using the beta version of Adobe’s new AI creation tool, Firefly). Articles will not be emailed as News Alerts or appear in the Daily Digest.
AI stories cannot and will not replace any of our in-depth reporting based on company announcements, such as this article on Fletcher Building's leaky pipe problem published this week, which provides the context and analysis an AI summary does not.
We are also supercharging our share price graph to include the AI-generated summaries of key announcements. Look for the diamond on the horizontal axis then click (or tap) to show the summary.
We believe this is a first for the classic share price chart featured on business and financial news websites the world over. It moves beyond just showing movements in share price to include information to help readers understand why those movements have occurred.
The AI tool is the brainchild of our data editor, Andy Fyers, and the development firm that we’ve worked with for years now, Turkey-based PlusClouds.
We’ve also had support from Google News Lab as part of a project on newsroom experimentation we’ve been part of for the past six months, alongside the Sydney Morning Herald and Australia’s National Indigenous Times. Google contracted consultancy Bastion Transform to help with project management and creating a model for how we as journalists can experiment successfully.
I’ve spent time lately playing with ChatGPT and image-creation tools like Midjourney. I see no way they can replicate the bespoke journalism created by our newsroom. Our investigations team – Victoria Young, Murray Jones and Oliver Lewis – can spend months teasing out information for stories from contacts, official documents and other sources. AI cannot do that.
But I see a use for AI summarising things such as NZX announcements. We’re not asking ChatGPT to source additional information by using its “knowledge” to expand articles. We’re taking one piece of content (an NZX release) and summarising it to be more like a news article.
I see this as freeing reporters to do more valuable work. We’re not replacing journalists with AI. Another advantage is that we can cover much more of the market.
Our editors still look at all announcements and will sometimes choose to produce versions of these articles with more information. They will still commission analysis, interviews and commentary.
We’ve had the NZX AI summaries in beta (pre-production) since late last year and we have refined the model along the way. We’re confident in what we’re producing.
Better search engine
Our next AI-related project is a significantly enhanced search engine. That will be released in the next month or so. We’re working with fledgling Auckland startup Silky Insights to build that.
I’ve been using the beta version and it is a huge improvement on our currently woeful search tool. You’ll be able to find the articles and information you want far more effectively, and even ask questions based on our previously published stories.
For example, asking, “What was Synlait’s profit in 2019?” gives the answer and then shows the story that covered the result.
Finally, we’re nearly ready to release a new app. Again, I think it is significantly better and faster.
Want to know more about the future of artificial intelligence? Read Niall Ferguson’s excellent overview of the ethics of AI.