Hope you’re all doing well considering the British Summertime heat feels like being trapped inside a microwave, and BigFinish are continuing to make their website and app worse.
Speaking of Briggsy’s Glorious Solar Empire with CD warehouses situated on the moons of Deimos and Phobos, someone in my Discord server recently shared this Reddit theory with me.
Just wondering what people’s thoughts are on this. Considering how appallingly put together the new site and app are, it wouldn’t surprise me if the people hired to make it were lazy enough to do it all with generative AI.
It’s just the hard glace cherry on top of the soggy trifle really, isn’t it?
Of course this is speculation but it is depressingly plausible. Many strange things going on in BF land right now. I hope they pull themselves together soon
I thought that it’s always raining in the UK? Do you even know what the words “sun” and “heat” mean?
Joking aside, and back to topic!
Sadly, I believe the decision to use AI has largely to do with cutting costs (which would also explain the rumours that BF contracted an external company to build their new website and app and picked the cheapest option). BF is a much smaller company than many of us believe, and they don’t have millions upon millions to spend on things, so they try to save a penny whenever they can. However, if this is the outcome, I question whether it was a wise decision.
Surely, they could ask one or two of their many competent writers to write a simple description for their new genre pages?
It’s very tempting to use AI for things because it can save time and money.
I’ve often thought of some cool things I could do with AI on my site, but so far I’ve completely avoided it. Nothing on TARDIS Guide was produced by AI.
From a technical standpoint I find it really cool and there are some cool ideas I could implement which would be impossible to do without it.
But I’ve avoided it because of the huge backlash, the fact that we are in an industry all about human creativity and imagination, and also the environmental aspects - LLMs use a lot of energy which I am appalled by. Not to mention, as evidenced above, the results aren’t always great.
It’s very frustrating to see Big Finish do this, and I hope these are just placeholders they did for speed and will be replaced by something actually written by a human.
There’s a Twitter thread here that casts some doubt on whether or not this is actually the work of GenAI: x.com
I certainly get why people feel that it is, and I definitely suspect it’s the work of a third-party as speculated above, but I wouldn’t want to bet either way on if that third-party used AI.
I certainly think BF are wise enough not to directly use Gen AI, and what we’ve heard about, e.g. provisions in their contracts, backs that up.
I really am not sure how accurate those “Detect GPT” things are, or even can be, but I tried it myself and am getting different results to that Twitter thread:
I thought BF would be wise enough not to launch a new website and app that are broken and unfinished but leave my library completely empty for a week and counting but here we are…
The link in the OP is interesting. If it’s true that BF is strapped for cash, I don’t understand why they’d spend money on a website no one asked for. I know you have to spend money to make it, but considering they’re a small company with a niche fanbase, I would think upgrading the site would be the least of their concerns (the complaints I see about the company are about everything but this). I’d love to see their financial statement, just for curiosity.
I remember hearing rumous that they’ve had cashflow issues ever since the Monthly Range got scrapped honestly, and that the Classic Doctor Adventures box sets haven’t been selling as well as they liked - particularly the 1DAs and 7DAs.
Think that’s why we got stuff like ‘Dark Gallifrey’ and ‘Sontarans VS Rutans’ coming out as individual releases again rather than the box set format - trying to remind people of their glory days.
Not to sound like a stuck record here, but I genuinely think resurrecting the old Monthly Range would solve a lot of their cashflow issues.
I remember reading an interview Briggs gave about why he cancelled it. Apparently he wanted to make it easier for people to jump into any range with any Doctor without worrying about massive story arcs.
It’s the curse of boxsets– if you’re interested in one particular story which is being sold on its own, you can easily buy it, but if it’s part of a boxset I for one am much less likely to buy an (expensive) boxset for the sake of one story. They’re great for arcs but problematic for individual releases.
That’s especially true with the death of the Monthly Range. I’m sure loads of people would be happy to pick up a single Five or Seven story if it looked cool, but when it’s bundled into a three-disc set with a story you’re not interested in, you’re much less likely to buy it or even pay attention to it. I much prefer the individual releases myself.
I really think that the main problem that causes their money problem is that they oversaturating their own market. With 10+ releases eatch month and a backlog with 1500+ stories, it easily gets overwhelming for new users and makes a lot of stuff feel unimportant due to the big back catalog. Who will buy box twelve of River Song when it is new if you did not buy the first eleven? The production costs to make this many audios a month must also be kind of high. I have not seen their numbers and might be completely wrong here but this is my theory.
Not only are AI detectors extremely unreliable, they have a bad tendency to categorize text by non native english speakers as AI.
Similarly, programs like Glaze to try and make it so that people can’t train ai on your art are easy to get around and don’t work half or more of the time anyways. It’s all modern snake oil.
That said, I think the main issue with using AI isn’t the AI itself, but that people are just uncritically slapping its output in without any thought. Sure, you can generate some boilerplate text with AI, but generate a few versions and edit it so it sounds good.
Similarly, if you’re going to use AI art, spend some effort on the AI art! Use things like controlnet and inpainting while working on it, and fix up by hand all the bits that it gets wrong…
To be honest, that tends to be a bit of an issue these days, too, where people tend to assume AI when something sucks, when humans are equally capable of sucking. Not to mention some artists end up getting accused of AI artwork due to having a style similar to what AI uses.
Though this case is indeterminate, as it could be the human copywriter was the one using AI.
(For art and pictures, look at characters hands and make sure they have the right number of fingers. Also look at background details and look for areas where things blend together in a way they wouldn’t normally. Shadows not matching what is supposed to be casting them is typical, too. Style wildly changing within a picture is also a sign. Gibberish text, as well, as AI tends not to be good at it.)
I would honestly assume the human copywriter used AI with descriptions like that
This is a lot less common more recently, but yeah background details and blending together is still definitely a thing, and hands being weird is still somewhat common, but they’ve usually got the right number of fingers
