Hi,
My name is Sara Forni, AI Product Manager at Atex, and this is MyType, a newsletter dedicated to journalism, innovation and artificial intelligence.
Every fortnight we will explore a topic related to this evolving field to discover the pleasure to do journalism without repeating boring tasks and using technology as a helper, not a substitute. That's what we like to call ‘digital joynalism’.
If this mail was forwarded to you and you would also like to receive future releases, you can subscribe to MyType here.
Now, let's get started!
Everyone is talking about us
Don't worry, this is not a self-praising newsletter introduction to this, but rather a general consideration on the relationship between the media and artificial intelligence.
In less than two weeks there have been at least three striking announcements concerning journalism, and that is not to be undermined considering that, it’s usually more of a niche topic rather than a mainstream one.
Something is starting to change. One must be careful, however, to watch the direction things will take as at the moment everything is still superficial and the debate on this issue is still at an early stage. In fact, it’s not uncommon to read alarming headlines about how AI will take over, “writing articles instead of journalists” when, actually, efforts should be concentrated on developing functions and tools that will simplify, speed up and improve the journalist’s daily effort.
OpenAI flirts with newsrooms
The first announcement was from the Associated Press. The American news agent announced a collaboration with OpenAI, the well-known founder of ChatGPT and other language models. AP will license part of its journalistic archive to OpenAI, in return gaining access to OpenAI's technology and expertise. Both organisations will benefit from the sharing of their different expertise. Also, as Associated Press wrote in a press release:
“The collaboration builds upon AP’s efforts over nearly a decade to use automation to make its journalism more effective, as well as help local news outlets integrate the technology into their operations”
But OpenAI hasn’t stopped stop there. A few days later they announced another collaboration, this time with the American Journalism Project (AJP), a leading venture philanthropy working to rebuild local news, which will receive $5 million in OpenAI API credits “to help its grantee organizations assess and deploy emerging AI technologies within their organizations”, announced the company led by Sam Altman.
Even Google gets in on it
Imagine you have an idea to help a colleague and while you're explaining it to them , you directly do all the work for them. That’s about how the meetings must’ve gone between Google, The Washington Post, The Wall Street Journal and The New York Times, who met BigG last Wednesday to discover 'Genesis'. According to The New York Times, Google's new tool can ingest "details of current events" and other information to "generate news copy".
"Some executives who saw Google's pitch described it as unsettling, asking not to be identified discussing a confidential matter. Two people said it seemed to take for granted the effort that went into producing accurate and artful news stories"
Google's intentions are undoubtedly good – otherwise I don't think they would present such an innovation to the major American newspapers - but their approach is problematic, as it only fuels the fears described above.
If you're enjoying MyType, why not share it with friends and colleagues?
Finding Newsworthy Documents using Generative AI
Suddenly, technology starts scanning newsworthy information, documents and events and when it finds everything, it immediately sends you a notification. Would be useful wouldn’t it? The Associated Press has been using a tool developed by SAM (a Canadian social media solutions company) for a few years now, to detect newsworthy events based on natural language processing (NLP) of text-based chatter on Twitter and other social media venues.
SAM alerts expose breaking news events sooner than human journalists could track on their own through manual monitoring of social media. This technology could speed up the discovery of news items and also their publication. Of course, stories still need journalists to be revised and have their fact checked, but having this service available would be very useful.
However there are also those who have gone further. Professor and media expert Nick Diakolopolous, in a post blog on Generative AI in the Newsroom explained how to use AI to rate the newsworthiness of scientific abstracts in the Newsroom project, a case study published on the arXiv pre-print server, authoritative portal for free distribution of scientific articles. In particular, Diakolopolous paid two professional sci/tech journalists to rate 55 documents on a newsworthiness scale from 1 to 5. Then, he did the same with AI, applying three specific prompts to OpenAI's Gpt 3 and Gpt 3.5 turbo language models. The final result is not a complete victory but it is a starting point.
“We found a moderate correlation to expert ratings of newsworthiness for scientific abstracts. It was relatively cheap to develop and it would be cheap to deploy even at a reasonably large scale”
The right time context and biases
At the same time, there are things that can still be perfectionalized. One is the temporal context. As we know, these models are not trained on up-to-date information on current events topical information. For example, a scientific paper that talks about the latest news on COVID-19 won‘t have a different degree of newsworthiness for AI, simply because it was trained with information up to 2021. The second reason is related to the human biases that we are unknowingly transferring to AI models. Try to think about it, even if we think a news article is as close to the facts as possible, how often are we led to give weight to specific information that one of our colleagues has not even considered? Experimentation in this regard could continue. One method to avoid these distortions could be to add the editorial line of the editorial office we are writing for, to the prompt. In this way, the model could become more sensitive to what we consider to be news and what not.
Stories & Tools
AI ran all the programming on a Swiss radio station for an entire day
Wix will soon let you create an entire website using only AI prompts.
Meta, Google, and OpenAI promise the White House they’ll develop AI responsibly
Watch the 50+ sessions of the JournalismAI Festival on YouTube
A list of AI tools to detect disinformation
That's all for today!
There would be no 'digital joynalism' without sharing. We don't want this channel to be a one-way street. So for any ideas, suggestions and thoughts on the subject, don't hesitate to email me at sforni@atex.com!
Have a good weekend,
Sara
Dear Sara Forni, you've mentioned a preprint on ArXiv and included a link, but the link only points to ArXiv itself. Could you include the link directly to the reference you are citing? Thank you for the pleasant and interesting read.