The 167th Block: Text perplexity, news as music, and newsletter growth in Brazil
And more guidelines for AI in the newsroom
This week…
Your reading time is about 6 minutes. Let’s start.
I had a busy week and did not have a lot of time to visit and drink from all of my usual Internet firehoses for this edition of the newsletter. It’s one of the reasons I like the newsletter format — the stories come to me instead. Let me know if there are newsletters that you enjoy that you think I might. It does not have to be about media analysis and tech culture. My interests are diverse, and right now, for example, I would like to keep up with the upcoming FIFA Women’s World Cup. What’s the goss, lads? I would also like to have more newsletters that are audio/visual driven, although short essays and observation-based pieces are welcome.
And now, a selection of top stories on my radar, a few personal recommendations, and the chart of the week.
ICYMI: The Previous Block explored online news without tech giants, at least in Canada, and tracking and privacy on Apple devices and Meta’s Threads app. CORRECTION NOTICE: None notified. Writing guidelines for the role of AI in your newsroom? Here are some, er, guidelines for that
Hannes Cools and Nicholas Diakopoulos for Nieman Lab:
The guidelines we analyzed vary in specificity, and are sometimes named differently as “editor’s note,” “protocol,” “principles” or even “deontological charter.” The tone of some of the guidelines are restrictive where specific uses are banned. Other documents are more examples of governance, where news organizations are committing to specific responsibilities to make AI less risky. Below, we will discuss some of these overarching patterns with examples from specific guidelines.
A long read, but a good read. The authors analysed a range of organisations, mainly in Europe and the US. Still, you can get in touch with them if your organisation has published guidelines on generative AI that they should add to their list.
Programs to detect AI discriminate against non-native English speakers, shows study
Ian Sample for The Guardian:
Writing in the journal Patterns, the scientists traced the discrimination to the way the detectors assess what is human and what is AI-generated. The programs look at what is called “text perplexity”, which is a measure of how “surprised” or “confused” a generative language model is when trying to predict the next word in a sentence. If the model can predict the next word easily, the text perplexity is ranked low, but if the next word proves hard to predict, the text perplexity is rated high.
Large language models or LLMs like ChatGPT are trained to churn out low perplexity text, but this means that if humans use a lot of common words in a familiar pattern in their writing, their work is at risk of being mistaken for AI-generated text. The risk is greater with non-native English speakers, the researchers say, because they are more likely to adopt simpler word choices.
Surprise.
News as music: Schibsted explores AI-based innovation to attract GenZ
Neha Gupta for WAN-IFRA:
The team hired 10 teenagers, aged 16-19, from a youth recreation centre in one of the most crime-ridden areas in Stockholm. This 10-week long paid assignment invited them to come into Schibsted and work in prototyping future news experiences.
Below are three news products these youngsters designed through the course of the programme:
News Clinic: This product is about news media taking more of an active approach to safeguarding the mental health of its news consumers. This speculative product aids consumers in their news consumption by an on-demand news therapist, which is an AI avatar. This AI-therapist can be activated either automatically or manually.
It’s Time: A news service that challenges the logic of optimising for time spent. The idea here is that news businesses should optimise for informing consumers in as little time as possible. This service provides news consumers a weekly budget, where they can only consume the product for seven hours per week, one hour per day, 30 minutes in one go.
Once they consume the product for 30 minutes, it automatically self-destructs and cannot be opened for the next two hours.
News Sound: This concept asks a basic question – what if the news was music? The idea here is that publishers convert written news stories into AI-generated music.
Did it work?
Inside Brazil’s growing indie newsletter economy
Nathália Pandeló Corrêa for Inbox Collective:
In the expanding world of Brazilian newsletters, one name has captured the attention of readers and industry insiders: Gaía Passarelli. A former MTV host and media personality, Passarelli has harnessed her existing popularity by tapping into the power of a growing medium. After the launch of her newsletter, ”Tá Todo Mundo Tentando” (“Everyone is Trying”), in the midst of a nationwide lockdown, Passarelli emerged as a success story in the era of pandemic-born newsletter writers.
As Brazil witnesses a surge in independent media and the power of social media engagement, Passarelli’s journey exemplifies the rise of newsletters in the country, where writers and journalists now forge direct connections with their audiences.
Brazil’s media landscape has grown more diverse and digital, and newsletters have emerged as a key way for writers, journalists, and organizations to engage with audiences. While traditional media outlets have faced financial constraints, massive layoffs, and struggled to adapt to digital platforms, independent journalists and writers have seized the opportunity to build direct relationships with their readers.
Where are all the non-English newsletters? Please recommend. I am open to anything.
What I read, listen, and watch…
I’m reading “To converse well,” an essay by Paula Marantz Cohen for Aeon. “Since conversation is, by definition, improvisational, it is always bringing to the fore new or unforeseen aspects of oneself to fit or counter or complement what the other is saying.”
I’m listening to “Google and Meta Are Fighting with Canada” on Tech Won’t Save Us by Paris Marx, who is joined by David Moscrop for this episode.
I’m watching Center for Humane Technology co-founders Tristan Harris and Aza Raskin discuss the “AI dilemma.”
Other curious links:
“The race to put Indigenous land on the map” by Liani MK for Rest of World.
“European Court extends ‘right to be forgotten’ from search engines to news sites” by Charlotte Tobitt for Press Gazette.
“Scientists used ChatGPT to generate an entire paper from scratch — but is it any good?” by Gemma Conroy for Nature.
“El interés en ChatGPT empieza a caer y su alianza con Microsoft apenas come terreno a Google” por Carlos del Castillo en elDiario.es.
« Bien préparer son brief créatif » par Christelle Perrin dans Médianes.
Chart of the week
Statista’s Anna Fleck reported preliminary data from the World Meteorological Organization showed that the world experienced the hottest week on record earlier this year (July 3-9). Fleck visualised absolute temperature records being broken around the globe since 2017.
And one more thing
