Working with around 50 local newsrooms in the U.S., we’re continuously mapping how and why AI is used. It’s clear that there is a spectrum of use cases, which basically relate to how much human oversight is required at any given instance.
Before generative AI became widely available (with the launch of ChatGPT late 2022), the AI used in the news industry consisted of algorithms based on conditions, so called rules-based AI. This is the technology used, for example, in recommender systems and personalization. It’s also what United Robots deployed to build our text robots and what companies like AP and The Washington Post used in their early content automation efforts. We use rules-based AI for the data analysis which identifies the story in the data set, as well as to generate the article that tells that story. With rules-based AI, the human involvement can basically be limited to the development of the system.
In United Robots’ case, that means that we build a text robot, using quality, structured data as the “raw material” — for example, all the facts in a text comes from a verified data source. We can also adjust the set-up of the robot to follow the editorial guide of a given newsroom. Once all of this is in place, the robot-generated articles are ready for auto-publication. Many of our clients in the U.S. leverage auto-publication, which means, for example, weather warnings and wildfire updates reach the news site — and the public — within minutes of the warning being issued by authorities.
As generative AI improves and as publishers’ AI usage matures, United Robots is currently testing whether generative AI might help improve our efficiency, output quality and time to market for new products. United Robots existing content products use Rules based AI for data analysis and text generation (green). We are also testing improving efficiency and quality of output by letting generative AI support our processes (yellow).
As generative AI improves and as publishers’ AI usage matures, United Robots is currently testing whether the newer technology might help improve our efficiency, output quality and time to market for new products. While staying committed to delivering trust-worthy, rules-based content to our newsroom partners, we’ve started experimenting, on the side, with different ways of using generative AI. It’s still early days, but our hypothesis is that generative AI can be useful for us at tasks like turning unstructured data into structured data, at translating existing rules-based text robots into new languages and at adding context to granular data feeds from traffic data, for example.
Come see us at MEGA!
Antonia Mariassy and Cecilia Campbell will be in Austin, and are looking forward to exploring how automated content can help your newsroom free up reporter time and expand your coverage. Email us to set up a meeting!
cecilia.campbell@unitedrobots.se
Learn more about our U.S. products and clients here: https://www.unitedrobots.ai/products/us
