An interactive panel held at this year’s Maine Press Association (MPA) Conference on Oct. 14 explored the distinct technological transformation that has been evoked in the presence of Artificially Intelligent (AI) bots and Large Language Models (LLM). Director at Associated Press, Dwayne Desaulniers, described how forums such as ChatGpt can be used as a tool in news writing while still upholding journalistic integrity.
AI forums have taken the world by storm in recent months, with the extent of its potential unclear. Every day, AI grows stronger as more information is being programmed into it. Its range of capabilities expands constantly as people worldwide contribute to this development.
“For me, it’s like being back in 1995, when the internet happened, and we had this crazy thing called Netscape navigator… between then and now, we’ve seen so much tech in our newsrooms. But I gotta tell you, there’s something different about this one that is intimidating. It’s uncertain and unknown where the impacts are going to go,” said Desaulniers.
ChatGpt has created a more streamlined approach to online research but also serves as a potential threat to the legitimacy of student work in academia and the strive to sustain journalistic integrity. With the line between using AI as a reference and committing plagiarism growing exceedingly thin, Desaulniers and many other journalists have been educating themselves on its increasing role in reporting.
There is a dichotomy between embracing new technologies to streamline production and reach a wider audience versus allowing robots to overtake the scope of this pivotal profession. In doing so, journalists would be relinquishing the torch, so to speak. It is necessary to be cautious in accepting AI and for each news organization to set strict limitations that specify appropriate use.
Another concept to be wary of is the spreading of misinformation. ChatGpt, for example, does not weigh its information based on credibility as Google’s search engine does. Based on the keywords used, AI may be pulling from random blogs equally as it is from organizations such as the Associated Press, which is known for its fact-based reports.
Desaulniers suggests requesting the forum to cite its sources as verification of what it’s telling you per the process of due diligence. In that sense, AI acts as a sort of middleman between you and the site information. Some say it is more beneficial to visit them directly and be a part of the readership, as online organizations do not receive any sort of compensation when their words have been included in a grand summary.
“Understanding how to use these things is a really great skill for journalists. Keep in mind the whole thing could blow up because these LLMs have scraped all of our journals, and that’s where all the giant database’s information is coming from, it’s our reporting. There’s lawsuits just waiting in the wings,” Desaulniers said.
The annual revenue for ChatGpt is $1.6 billion, with much of that amount coming from user fees, as the paid version costs $20 per month. The New York Times is seeking $150,000 from Open AI for each news story that information was gathered from. If content creators are granted payment for their work, AI could face its demise as it functions by taking various bits of data from hundreds of thousands of online platforms.
None of this is to say that journalists are under obligation to reject this innovation. Since they are incredibly keen on coming up with relevant inquiries, its use may prove more advantageous than not. ChatGpt’s exceptionally fast-generation abilities could be an excellent tool inside newsrooms nationwide.
Some of the numerous methods for beneficially using AI to the advantage of journalistic practices are as follows:
- “De-jargonning” documents full of scientific terminology
- Translating articles into any language. Despite being highly proficient, confirming accuracy through Google Translate
- Scanning long reports for news. By programming the computer to think like a journalist, it can assume that persona and pick out which sections are most relevant to readers.
- Brainstorming interview questions
- Writing headlines in preparation for publication or summarizing articles post-publication
- Social media posts after pieces are finalized
- Podcast or radio scripts
- Suggesting local topics to reporters that relate to national or international occurrences
By using these forums in accordance with the rules of journalistic integrity, we can collectively influence the newer models so that they become increasingly more harmonious with our ethics.
“An algorithm that says, ‘Here’s a question, look at the sequence of the words in that question and then go back to this giant repository of information and find similar patterns.’ Pattern recognition is really what it is. The machines aren’t sort of thinking. They’re saying literally ‘10 times out of 10, this word followed that word,’ and that’s basically how the answers are constructed,” Desaulniers said.
The research and data capabilities of Open AI are currently limited but still developing at a frighteningly rapid rate. Part of a journalist’s job is to authenticate information and accuracy. One way to do so is by asking questions to which you already know the answer. That way, you can test the waters to determine whether or not the answers are properly executed.
Desaulniers exemplified the impressive speed of ChatGpt’s generated responses by having it scan two photos and answer a related question. The first was a parking sign with multiple overlapping time regulations that were difficult to comprehend. He asked simply, “Can I park here on Monday from 4 to 5 p.m.?” Within seconds, he was given the correct answer without the hassle of interpreting the entire sign.
Secondly, he added a grocery store flier with nearly 100 unrelated coupons. Desaulniers informed ChatGpt that he has $15 to spend and wants sufficient food for the week. It managed to quickly compose a list of all the possible purchases that can be made using some of the coupons with the budget remaining below $20. It not only included the price of each item but also its nutritional value.
“At AP, we struck a licensing agreement with Open AI a few months ago, and our hope is that we will be able to make our voices heard about what we see that may be concerning. How can Open AI and ChatGpt be more transparent when folks read the output from that engine?” said Desaulniers.
This inquiry leads to an even broader discussion. Does AI threaten the future of good journalism, or will it remain a useful tool? Furthermore, to what extent should we participate in its evolution?