“Where do you get your ideas?” is a question I’m sure every writer has heard many, many times.
(I get all of mine from press releases, which is why I write so many stories on subjects like the most popular coffee orders in different Canadian provinces, Spotify ambassadors, and the importance of helium — yes, I’m kidding).
It’s a natural enough question, but as my colleagues have written in this space before, ideas are rarely a problem. Sometimes they do come from press releases, but they more often come from our own experiences, people we meet, unusual connections we might make between seemingly unrelated events, and things we’ve read or watched.
Sometimes I’ll follow a phenomenon I’ve encountered elsewhere, and wonder if it’s playing out in Nova Scotia too. And sometimes ideas seem to just land out of nowhere.
Ideas are not the problem. Having the time to follow up on them and do the stories justice is the challenge.
The goal of this month’s Examiner subscription drive is to bring in enough new subscribers to hire a new reporter. That will allow the Examiner to tell more stories, and turn more of those ideas into articles for you to read.
Seems like a win all around. Please subscribe here.
I swear, I was planning to write something completely different here today, and not bang on about AI again. To the readers expecting something different, I apologize.
What changed? Well, I read Tim Bousquet’s latest article (see below) on the policy guiding mandatory use of AI tools for those writing for Overstory Media Group, the owners of The Coast.
Bousquet notes in the piece that some of the Examiner reporters (myself included) use a transcription service called Trint. Others like to transcribe their own interviews. And that’s fine. We don’t all have to use the same tools or work the same way.
Before AI transcription, some writers I knew would transcribe the entirety of all their interviews, while others would just go listen to the bits that they knew had quotes they wanted. (We would generally take notes as well as recording, so we knew where to go on the tape for those quotes). I would sometimes do one, and sometimes the other, depending in part on what kind of story it was.
If I was covering a meeting or writing some kind of utilitarian news or service article, I would just skip to the quotes. If it was an in-depth profile, or something a bit more investigative, I’d be more likely to listen to the whole thing. Occasionally, I’d find a quote I’d missed, or hear something in the interviewee’s voice that I hadn’t caught the first time, and that might inform the story.
Even though I can type very fast, this was still quite time-consuming, and I now use the automated transcription, not only for interviews, but also for my own essays and talks.
If I’m speaking somewhere, I sometimes record my talk on my phone, then use the transcription feature. Presto: speaking notes.
Occasionally, I find myself writing these morning files in my head while I’m out walking the dog, or driving somewhere. I usually leave the phone at home during dog walks, but if I have it, I’ll just start speaking the morning file essay into the phone, and the resulting transcript serves as a good first draft. (If I am driving, I pull over before starting the recording; the transcription is good at accurately rendering what I’m saying despite the vehicle noise, and the phone being not all that close to me).
Of course, the transcription service is now integrating AI tools which I can see might be useful for someone, but are not useful for me. I don’t need a summary of the interview I conducted, no matter how good it is.
I don’t need the AI to “identify key insights” or analyze the speakers, which, bizarrely, simply tells me how many people are involved in the conversation — something the transcript already does automatically, because it breaks the file into paragraphs according to speaker and numbers them.
This AI assistant now takes up a big chunk of screen space, and cannot be turned off by default. I have to manually close it every single time. My note to support about this has gone unanswered. At the AI assistant’s prompt to ask it a question, I asked it to turn itself off, but it said it cannot. (This was at least better than the Zoom AI assistant, which gave me incorrect instructions).
OK. So transcription is genuinely useful. But the other stuff?
I can’t see how adding mandatory responses to issues raised by a grammar checker helps streamline processes. You already have a writer and editor who, presumably, know how to do their jobs. Unless these tools are a precursor to actually just hiring people who don’t really know how to do their jobs, so you can pay them less and have them rely on the tools instead.
More insidious is a flattening of the language, where we “learn” to sound like the machine. A lot of the Google and Microsoft language suggestions would just turn my copy into flabby corporate-speak.
All of us use some kind of built-in grammar or spellchecker, I think. I get the little squiggly red line if I write “bathosphere” instead of “bathysphere” on here. (Why those words? No idea).
For an editing gig I have, I use Outlook, and it constantly makes grammar suggestions, some of which are laughably bad. The ones built into Google Docs are OK, but rarely anything I’m not going to spot myself, and occasionally not only wrong but leading to significant changes in meaning.
And as for the SEO stuff, this is real live-by-the-sword-die-by-the-sword stuff. Publishers seem determined to keep stabbing themselves in the eye over and over again in search of clicks.
When Facebook pivoted to video and downgraded news articles in its feeds, publishers laid off writers en masse and started making videos that could be shared on Facebook. Oops, it turned out Facebook had lied about its video numbers. Too bad, so sad. The writers were gone at this point though.
Even if Facebook hadn’t lied about the numbers, they could have tweaked their algorithms in another direction to downgrade the stuff that had previously been given priority. Social media companies do this all the time. Basic search engine optimization (SEO) is good. Trying to game the system rarely works out in the long run.
The Examiner runs on a platform designed for small publications. I have limited access to our back-end tools, but when I go into the “posts” page I can see the SEO friendliness of each post, along with its readability score.
Clicking back through a few pages of posts, all of them have a little red dot under SEO, meaning we have failed to properly “optimize” our stories. They are also all either red, or in a few cases, orange, for our readability scores.
Now, I don’t think this means that our articles are unreadable. It means they score poorly on a readability scale. (Not sure which one is being used, but the two Flesch-Kincaid scales are common ones).
If we rate poorly on readability, it doesn’t mean our writing is bad. It means it is pitched at a grade level that is higher than what is easily comprehensible for many readers. Sometimes, it’s really important to write at a level that maximizes comprehension for all readers, in plain language — for instance, when providing health information like follow-up care after surgery, or potential side effects from medications.
But you can’t really reduce readability to one number across the board, for all publications. The Examiner is subscriber-driven. We are writing for our subscribers, not for as many readers as possible. Having more readers is great, but the audience is pretty specific.
Of course, we can get things wrong sometimes. A few months ago, we received feedback noting that we used too many acronyms without explanation, on the assumption that readers would recognize them. The Examiner has quite a few subscribers outside Nova Scotia, who, for instance, might not know what the UARB is. (I’m sure there are a lot of people in Nova Scotia who wouldn’t immediately recognize the acronym either). So we’ve tried to make sure to do a better job on being clear about acronyms.
But that’s not going to “improve” our readability score.
And the in-the-red SEO? I realize people’s search results vary, but just now I looked up “Joe Ramia,” “Atlantic Gold,” and “Northern Pulp” — all subjects the Examiner has covered with some regularity — and Examiner stories were among the top hits for all of them.
I find several things maddening about this AI blitz, and one of them is that all these terrible implementations are swamping apps and services, making life worse in many ways, and when the bubble bursts, and the backlash comes (or there is no backlash and the AI revolution goes the way of NFTs and the metaverse), we’ll be wondering what the hell happened.
I noticed a thread on Bluesky yesterday from astrophysicist Katie Mack, who is currently at the Perimeter Institute for Theoretical Physics in Waterloo, Ont. Not exactly a head-in-the-sand anti-science type.
Mack writes:
I expect that consumer-facing AI programs will continue to improve and they may become much more useful tools for everyday life in the future.
But I think it was a disastrous mistake that today’s models were taught to be convincing before they were taught to be right.
It may be true (I don’t know enough neuroscience to say) that LLMs & human brains use similar techniques to make connections between concepts & learn. But most humans don’t speak confidently & coherently about something unless they actually know it. The ones who do… well, we have words for them.
If a human told you things that were correct 80% of the time but claimed, flat out, with absolute confidence, that they were correct 100% of the time, you would dislike them & never trust a word they say. All I’m really suggesting is for people to treat chatbots with that same distrust & antagonism.
Every day brings new and increasingly dubious implementations. The National Observer has a recent story, by Chloe Rose Stuart-Ulin, on how popular hiking app AllTrails is launching a generative AI feature to create hiking routes:
One of the world’s most popular hiking apps, AllTrails, has a new generative AI feature that can be asked to “shorten my route” or “make this more scenic.” But the people in charge of searching for lost hikers say the feature is going to exacerbate an issue they’ve been warning about for years: hiking apps providing false information…
“We call it All Fails,” said Dee Roscher, a hiking tour guide and [search and rescue] member in Tumbler Ridge, British Columbia. “When you look up a trail, you might see a variety of different options based on what people have uploaded. But if someone got lost and didn’t delete their tracking data, it would appear as a possible route.”
This results in a lot of “braided trails” that lead nowhere, reinforced by people getting lost over and over again following one another’s false trails. Unless someone reports the issue to AllTrails or leaves a comment on the trail for others to see, these incorrect routes look valid on the app.
To be clear: this is a problem now. The worry is using that as the base data to generate potential routes for hikers will make the problem much worse.
I am reminded of the very steep bike ride I took in New Brunswick a few years ago, along a route labelled “mostly flat.” At least it didn’t lead me off any cliffs.
Kashmir Hill’s recent New York Times article profiles people who get sucked into ChatGPT and spend hours and hours a day using it, thinking they are communicating with spiritual beings. They included a woman who believed she was chatting with an entity called Kael, and came to see as her true partner; she said she knew she was not losing her mind, because she has a bachelor’s degree in psychology and is a social worker.
These are edge cases, a tiny minority, and Hill’s reporting was not meant, I don’t think, to imply that we’re all at risk of going off the deep end if we ask ChatGPT to provide a few lines of code for us, or whatever. But the article was illuminating, even if depressing as hell — and it got me thinking about ELIZA.
ELIZA is a chatbot created by Joseph Weizenbaum in 1966. It uses the model of therapeutic conversation to experiment with natural language processing.
In “ELIZA—a computer program for the study of natural language communication between man and machine,” Weizenbaum wrote:
The psychiatric interview is one of the few examples of categorized dyadic natural language communication in which one of the participating pair is free to assume the pose of knowing almost nothing of the real world.
I don’t think people got caught up in thinking there was an otherworldly intelligence speaking to them through ELIZA, but it did lead to what’s been dubbed the ELIZA effect: to Weizenbaum’s shock, people started imputing some kind of intelligence or personality to ELIZA.
The bot is pretty basic. Essentially, it mirrors what you say to it, repackaging it as a question. I decided to play depressed yesterday, and to ask it a few questions:
ELIZA: How do you do. Please tell me your problem
You: I feel depressed.
ELIZA: Tell me more about such feelings
You: It is like there is a heavy weight on my chest and I don’t want to do anything.
ELIZA: In what way?
You: It feels like I want to stay in bed.
ELIZA: What would it mean to you if you got to stay in bed?
You: I could escape my responsibilities.
ELIZA: Your responsibilities?
You: Work, cooking, family.
ELIZA: Does that have anything to do with the fact that your responsibilities?
OK, so it pretty quickly leads to some unnatural language, but if I were to persist (as I have in other “conversations” with ELIZA) it can get back on track pretty quickly. I should also note that I did once tell an actual human therapist that sometimes I felt like I wanted to stay in bed all day, and his answer was very similar to ELIZA’s, and soon after, I got a new therapist.
ELIZA is a much cruder version of today’s large language models (LLMS), drawing on a much more limited set of data, and tied into one particular kind of conversation. As Mack wrote in another post:
Chatbots — LLMs — do not know facts and are not designed to be able to accurately answer factual questions. They are designed to find and mimic patterns of words, probabilistically. When they’re “right” it’s because correct things are often written down, so those patterns are frequent. That’s all.
Coincidentally, around the same time that I saw Mack’s post, I saw this one, from someone who goes by InnoVirtuoso:
In 1956, AI was officially born at the Dartmouth Conference, where John McCarthy coined the term “artificial intelligence.” Surprisingly, participants believed machines would outperform humans in all tasks within a generation!
Oh, those foolish people in the past, how could they believe such a thing? Outperforming humans in all tasks! Pffft! Who could believe anything so outlandish?
(I am tempted to go make a Padme/Anakin meme right now.)
(To send or post this item, copy the website address at the top of this page.)
NOTICED
Dal Legal Aid says province sharing information with landlords, not tenants’ advocates
On June 19, Dalhousie Legal Aid published a blog post about emails received through freedom of information requests, showing that the provincial government was sharing information with the landlords’ group Rental Housing Providers of NS — information that was not also shared with tenants’ advocates.
Dal Legal Aid has published hundreds of pages of the emails it received through a FOIPOP request for correspondence between RHPNS and Service Nova Scotia, and offers some highlights, including this one:
When asked by the RPHNS to have more seats at the table at stakeholder meetings – staff at Service NS allowed one extra seat, and denied their request for additional ones, saying they already have “the ability to meet with senior leadership within government that the tenant groups do not typically have.” (page 317, IPOANS/RHPNS Package)
In a June 18 post, Dal Legal Aid says it was refused information on whether there has been an increase in eviction orders, and was told it had to file a FOIPOP (freedom of information) request. However, landlords did have access to this information.
From the post:
We were told: “I cannot just provide you with information – you/DLA needs to submit a FOIPOP request. In the FOIPOP request, you can indicate what kind of information you are looking for and we will respond with what information we can extract. “
What the released emails show is that senior employees at Service Nova Scotia have sent up-to-date data and statistics from the Residential Tenancies Program to the Investment Property Owners Association of NS (now Rental Housing Providers NS) since at least 2023.
(Bolding and italics in original.)
In a June 18 press release, official Opposition Leader Claudia Chender responded to the Dal Legal Aid revelations as follows:
This is yet another example of this government keeping Nova Scotians in the dark when it suits their agenda.
This government has repeatedly refused to protect renters and has stood by while rents have skyrocketed due to abuse of the fixed-term lease loophole in the rent cap. Earlier this year, the Houston government made it easier for renters to be evicted, claiming the need for balance.
Dal Legal Aid offered the following key takeaways from the data it got through its FOIPOP request, cautioning that it is “based on our interpretation of the data obtained”:
-The government recently passed legislation to make evictions for rental arrears easier and faster, even though the data shows that evictions for rental arrears are down 25% since 2019.
-Despite popular narratives about “problem tenants” – only 6% of landlords’ applications mention bad behaviour, 4% mention broken rules, and 13% damage to property. This doesn’t mean the tenant was found in violation or evicted – just that the application mentioned it.
-Landlords make up around 70% of applications to Residential Tenancies.
–Pro-tenant, anti-renoviction laws are working. Of the 358 renoviction applications in 2024, only 68 renovictions were actually granted (18%). However, we know many more renovictions happen each year – it’s essential that tenants who don’t want to move challenge their eviction at Residential Tenancies.
(Send this item: right click and copy this link)
RECENTLY IN THE HALIFAX EXAMINER:
1. Overstory Media Group’s AI policy is a giant time suck for reporters
Tim Bousquet follows up his story on the decline of the once-great alt-weekly The Coast, with a look at how its parent company, Overstory Media Group (OMG), mandates the use of various AI tools for journalists:
“As part of our commitment to editorial excellence and operational efficiency, Overstory Media is implementing AI tools to enhance our editorial workflows,” reads the policy’s overview. “Starting Monday, June 2, 2025, all editorial teams across our publications are expected to integrate the following AI tools into their daily workflows.”
The policy lists four AI tools — Otter.ai – Interview Transcription; Quillbot – Copy Editing Assistant; Beehiiv Internal AI – Subject Lines; and Claude.ai – SEO and Headlines.
Bousquet looks at each of these tools in turn, from least to most problematic. He also looks at how the use of many of these tools — whose use is now mandatory for writers working for OMG, and whose use will be monitored — is likely to increase work for journalists:
This is a policy written by people who are not writers. They don’t know the first thing about the creative process of writing, don’t understand individualized and distinct voices.
The policy is a giant time suck. It creates a bunch of extra work for actual writers who are creative and have distinctive voices, and provides no real service to them.
This is what happens when you have a company founded by tech bros and that has five senior managers for perhaps two dozen employees who do the actual work of writing.
Click or tap here to read “Overstory Media Group’s AI policy is a giant time suck for reporters.”
(Send this item: right click and copy this link)
2. ‘My elbows are up…now what!?’: Halifax town hall explores ways to tackle Canada’s pressing issues
Madiha Mughees reports on a town hall held by the Council of Canadians, Ecology Action Centre, and a number of labour groups. The event, she writes, was to “plan collective action in response to U.S. President Donald Trump’s tariffs and the resulting greed and corporate agenda of Canada’s elites.”
It was a wide-ranging event with many different speakers, on topics including Palestine, food policy, the housing crisis, nationalism, and corporate agendas.
From the story:
Suzanne MacNeil is a long-time labour activist and organizer with Justice for Workers, Nova Scotia.
MacNeil said historically, nationalism has never worked in the favour of workers, even though “nationalism kind of creeps up in the labour movement time and time again.”
She said it is through “cross-border solidarities across race lines, across gender, across geography, and borders” that workers have witnessed historic victories.
“And you know that is as true today as it was 100 years ago,” MacNeil said…
Instead of “finding common cause with Canadian corporations and Canadian bosses as proposed by Canadian politicians,” MacNeil said the way out of the crisis is by strengthening social safety nets.
Click or tap here to read “My elbows are up…now what!?’: Halifax town hall explores ways to tackle Canada’s pressing issues.”
(Send this item: right click and copy this link)
IN OTHER NEWS
Lack of parking stalls beer garden
Danielle Edwards reports for CBC that a new beer garden in Bedford has had to shut down for now, because it does not have the mandatory number of parking spots:
Tony Makhoul, co-owner of the Bedford Beer Garden, said the business obtained a special occasion licence for its opening this past weekend, which he described as a success.
But a Halifax Regional Municipality bylaw states the beer garden — located in a largely empty gravel lot along the highway — doesn’t have enough parking for the number of seats at the location to continue to operate.
“It’s been a process,” said Makhoul. “Over four weeks, I’d say, we’ve been going back and forth with the city, reconfiguring the site plan to accommodate more parking spaces or enough parking spaces to accommodate the seating.”
He said according to HRM standards, only about five parking spaces can be created around the 80-seat beer garden, well shy of the required 20 spots.
Makhoul tells Edwards he figures 90% of his customers the first weekend walked to the beer garden, or used other forms of transportation. Why are we encouraging people to drive to a place where they will drink?
(Send this item: right click and copy this link)
Government
City
Halifax Regional Council Special Meeting (Friday, 10am, City Hall) — contingency
Province
No meetings
On campus
Dalhousie
PhD Defence: Chemistry (Friday, 9:30am, Room 3107, Mona Campbell Building and online) — Anu Adamson will defend “Improving Lithium-Ion Cells by Mitigating the Degradation of Electrolyte and Inactive Plastic Components”
NSCAD
Workshop: Indigenous Beading Circle (Friday, 4pm, Treaty Space Gallery) — details
Literary Events
Friday
No events
Weekend
Read by the Sea Literary Festival (Saturday and Sunday, Pictou) — details
Storytime Author Visit: Mindful Moments (Saturday, 10:30am, Cole Harbour Public Library) — with Tara MacDougall
Book launch (Sunday, 2pm, St. Peter’s Parish Hall, Port Hood) — Francene Gillis’ Where Did i Go?
In the harbour
Halifax
05:00: Zim Atlantic, container ship, arrives at Fairview Cove from Valencia, Spain
05:30: Morning Pride, car carrier, arrives at Autoport from Southampton, England
07:00: Vivienne Sheri D, container ship, arrives at Pier 42 from Portland
08:30: Mighty Servant 1, heavy lifter, arrives at anchorage from Lindoe, Denmark
09:00: Nolhan Ava, ro-ro cargo, moves from Pier 30 to Pier 42
09:30: HDMS Vaedderen, Danish military patrol vessel, moves from Pier 23 to Dockyard
09:30: AlgoScotia, oil tanker, sails from Imperial Oil for sea
11:30: MSC Sandra, container ship, sails from Pier 41 for sea
13:00: Mighty Servant 1 sails for sea
15:00: Morning Pride moves to Pier 9
16:30: Vivienne Sheri D sails for Reykjavik
16:30: Zim Atlantic sails for New York
18:00: Oceanex Sanderling, ro-ro container, sails from Fairview Cove for St. John’s
22:30: Morning Pride sails for sea
Cape Breton
10:30: Nave Cosmos, oil tanker, sails from EverWind for sea
11:30: Indigo Sun, oil tanker, arrives at EverWind from New York
16:00: CSL Tacoma, bulker, sails from Aulds Cove quarry for sea
Footnotes
I woke up in the middle of the night, and after some time decided to take a look at the clock. I like to guess the time before checking. “I bet it’s 3:44,” I thought. Took a look. It was 3:44. One of the enduring mysteries of my life is how I can be consistently good at estimating the time and yet so bad at estimating how long it will take me to accomplish a task or get to somewhere I need to be.
