This morning, UK Defence Journal trialled a new experimental tool aimed at automating the production of short-form news bulletins. We quickly found it was a terrible idea.
The system used a WordPress plugin configured to pull in RSS feeds and metadata from newly uploaded defence imagery. The first test focused on recent F-35 activity, with the aim of generating concise updates of around 100 words.
The concept was to assess whether automation could help accelerate the publication of routine, fact-based updates. While the technical integration functioned correctly, the editorial output did not. The system frequently hallucinated content, often inventing details and misreading context. It also generated fabricated quotes, despite explicit instructions not to do so.
As a result, we have ended the trial and will continue to rely solely on human editorial input. We apologise to our readers that this tool, evidently, did not work.
We remain committed to improving our coverage and maintaining a balance between timely news updates and in-depth reporting. If you have suggestions for how we might increase the volume of accurate and trustworthy bulletins, allowing our team to focus more on long-form investigations and analysis, we welcome your input. Please get in touch via our contact page.
No problem George. You and the rest of the team do a excellent job. Sometimes the human touch is still better. 👍
I know everyone would be against AI content in any context.
But even if you did use it, it doesn’t make sense not to proof read, and it shouldn’t be posted under your name as the author. There should be a separate account that makes it clear it’s an AI summary of other sources.
A far simpler solution would be to copy the MOD press release directly via RSS, which is what it’s intended for. Then you can edit in an updated analysis later on like TWZ do.
George
Good move. The point about ‘AI’ is that it’s not (intelligent). My eldest and his eldest are both computer whizz kids and they say it’s just mindless search engines going round and round interrogating each other in ever decreasing circles, and a new bit of garbage input onto one of them soon becomes established ‘fact’ on all of them. Their recommendation – stay human.
There is a basis to that certainly, but it would be wrong to take that as an exact description of the way they work, plus they are developing on a monthly basis to the point that many prominent experts including the father of Ai claiming they will achieve AGI within a year or so. You could judge humans in a similar manner as the example you employ and easily claim them to thus to be unintelligent. Humans are deemed intelligent animals despite most being thick compared to scientists and professors, yet even some of those can make some deeply ignorant statements. I mean last week more than one average human being rang up radio phone ins on the Bob Vylan affair thinking it was Bob Dylan they were criticising which shows how we spend more time mouthing off than actually checking the most basic of facts. So too easy to jump to the conclusion that such things are indicative of a non thinking entity, especially when we are generally biased towards our own species. We are still in the Stone Age of artificial intelligence presently, progress has been made that even Ai experts thought was at least 50 years away a decade ago so best not to under estimate the potential even if it’s still not clear how intelligent and how quickly it all happens. I mean what is intelligence? It takes many directions while we tend to judge it by human standards only as the true scale. Go listen to Ameca responding to questions, yes she isn’t sentient but some of the answers show an iSight that you cannot simply gather from complex study of search engines especially in a few seconds, more is going on than that and in the end we have to accept that human intelligence started out in taking and associated input through its ‘sensors’ and bringing it together to initially be useful in hunting and general survival, became instinctive but over time became a lot, lot more than that pretty much as a side effect nature never intended, it simply uses too much fuel. Oh and that took many, many thousands of years to accomplish too.
Hiding in the shade today! East Anglia is sizzling🌞🥵
Clearly, AI is not intelligent
It’s not sentient by any means, at best it’s really a sophisticated pattern matching/filtering algortihm.
AI is not the answer to everything. You need to keep a human in the loop to filter out the misinformation.
It was probably over twenty years ago that I read a comment in a book saying something like “artificial intelligence has had a long history of over-promising and under-delivering”. Even truer now than it was back then. (By the way the term artificial intelligence was first coined in 1956 at a conference which concluded that they’d get the whole AI thing finished in a summer! Seventy summers later…)