‘The 600 series had rubber skin. We spotted them easily.’
Kyle Reece. The Terminator
Yesterday, NaNoWriMo released a statement declaring that discriminating against people using AI was both ableist and classist. They did this while also being sponsored by an AI-powered grammar checker and it went exactly as well as you think. I’ve used NaNoWriMo twice, got two useable zero drafts out of it and defended it more than once from whichever pro author’s turn it was to say ‘EVERY MONTH IS NANOWRIMO WHEN YOU DO THIS PROFESSIONALLY’ that year. To see it screw up this badly was infuriating, and I said as such. As did about ninety percent of my social media bubbles.
My buddy Brendon had a slightly different take on it, as he often does. He noted that, just maybe, this is a communications failure before an ethical one. I pulled that string, and it led me to a conclusion about something pretty fundamental. To get there, we’re going to break the situation down into the players. Starting with NaNoWriMo.
NaNoWriMo is a yearly event where authors sign up to hit a specific wordcount across November to create a zero draft of a novel. It’s basically a self-sustaining accountability network. A community tool designed to help you get the terrible draft on the page where you can fix it. It’s a fun system but they’ve been in trouble for a while. As Jason Sanford reported a while back the organization took months to remove admin privileges from a moderator who had been behaving inappropriately towards children. They’ve accepted sponsorship from predatory vanity presses. They’ve, in every way that matters, faceplanted. So much so that earlier this year, their interim Executive Director sent this email to members;
“We listened to a lot of feedback in November—some of it was hard to hear, but taking action to improve the organization was necessary. Since then, we’ve paused some of our standard operations in order to think through how to move forward, holding our mission in mind and community safety as priority number one.”
It’s a laudable sentiment. In exactly the way this, released yesterday, is not.
Classism. Not all writers have the financial ability to hire humans to help at certain phases of their writing. For some writers, the decision to use AI is a practical, not an ideological, one. The financial ability to engage a human for feedback and review assumes a level of privilege that not all community members possess.
Ableism. Not all brains have same abilities and not all writers function at the same level of education or proficiency in the language in which they are writing. Some brains and ability levels require outside help or accommodations to achieve certain goals. The notion that all writers “should“ be able to perform certain functions independently or is a position that we disagree with wholeheartedly. There is a wealth of reasons why individuals can’t “see” the issues in their writing without help.
General Access Issues. All of these considerations exist within a larger system in which writers don’t always have equal access to resources along the chain. For example, underrepresented minorities are less likely to be offered traditional publishing contracts, which places some, by default, into the indie author space, which inequitably creates upfront cost burdens that authors who do not suffer from systemic discrimination may have to incur.
- At first blush, this looks like they’re saying that barring authors using AI is classist, ableist and racist. If so, that is bad across multiple axes.
- Using disadvantaged groups as human shields isn’t lifting them up, it’s using them as human shields.
- Embracing this technology absolutely, one hundred percent, guarantees a tidal wave of worthless ChatGPT regurgitated pseudo manuscripts of the sort that almost killed Clarkesworld.
- That choice puts incredible pressure on staff already battling the internal culture of the organisation and, worse, renders meaningless the very point of the month itself.
Like I said, all of this is very, very bad. Then, because they’ve decided all I will do this week is edit this piece, they updated the article.
Note: we have edited this post by adding this paragraph to reflect our acknowledgment that there are bad actors in the AI space who are doing harm to writers and who are acting unethically. We want to make clear that, though we find the categorical condemnation for AI to be problematic for the reasons stated below, we are troubled by situational abuse of AI, and that certain situational abuses clearly conflict with our values. We also want to make clear that AI is a large umbrella technology and that the size and complexity of that category (which includes both non-generative and generative AI, among other uses) contributes to our belief that it is simply too big to categorically endorse or not endorse.
So basically, Brendon was right and they are supporting non-generative AI which is good while also criticising Generative A.I. which is bad. But they aren’t criticising it because the situation is too big and complex and anyway the check has cleared.
There are, against all laws of God and Science, two good points made in that paragraph. The first is that bad actors are infesting the A.I. space and they absolutely are. The second is that A.I. is an umbrella term, and it is. When my research hit this point yesterday, I realized I had some knowledge gaps. I asked the internet for help. They DELIVERED as you’ll see in a moment.
Before we get there though, it’s worth pointing out that Generative A.I. creates really bad, generic writing that’s excellent as a first draft but is in no way close to being completed, successful or good. writing. As is pointed out in this thread, it reads like the white guy who didn’t do the reading.
As has been pointed out elsewhere, it also creates the horrific situation of the very authors whose style pings as slightly outside the self-appointed ‘norm’ being decried as using A.I. by the A.I. powered detection systems designed to stop plagiarism. That means authors using assistive programs face the real possibility that what they think is streamlining them is actually tripping them up. Which brings us to A.I. itself.
Charles Babbage’s Difference Engine
The two types of A.I. we most commonly see are called Generative and Analytical. Both are powered by enormous data sets but do wildly different things with them. Generative A.I., as its name suggests, takes those enormous data sets and makes something that looks new but is actually made from other people’s data arranged in a slightly different way. If you’ve seen any awful, weirdly varnished looking artwork in the last three years, it’s been Generative A.I. If you’ve wondered who popped an actor you admired in the back of the head with a 2 by 4 as they babbled about how soon MOVIES WILL BE MADE BY COMPUTERS, that’s Generative A.I. It’s Sam Altman’s Folly. Grey Goo generators. Thief Machines. The bad actor under the A.I. umbrella and the one which is being pushed by venture capitalists because, presumably, Hell has invested in OpenAI. They’re also the basis of a lot of spellcheckers, programs which have been around for over a decade. Programs like ProWritingAid, which currently sponsors NaNoWriMo this year. A firm which, in fairness, is over a decade old, rather than one of the endless flying bored apes that make up the recent A.I. crop.
The fact they’re a decade old speaks to the larger problem everything under the A.I. umbrella faces. Generative A.I. is being used so persistently for such remarkable acts of cultural vandalism that we’re officially into ‘No one trusts anyone anymore and we are all very tired territory,’ ‘A.I.’ has become synonymous with shitty art, overpromising, under delivering and that unique brand of pseudo-Randian techbro callousness that’s deeply affronted we aren’t dying sooner or applauding louder. By itself, that’s a big problem, including for companies like ProWritingAid who have a decade’s worth of (presumably) happy customers. But when you see Analytical A.I. tarred with the same brush, that’s a tragedy. Because Analytical A.I. isn’t as loud, evil or sexy. It’s just very, very useful. Yesterday I asked the internet for examples of Good A.I. usage on Twitter and BlueSky. Here’s what I got;
AI mapping of protein targets in pharmaceutical development, Astronomy, specifically star and planet detection, cancer cell detection, finding proteins, tailoring drugs to specific users, assistive technology, Project Gutenberg’s audio book collection, science based studies and research in general, cancer detection, fraud detection, process acceleration, video and audio editing, pharmaceutical R&D, NPC behaviour (NOT scripting) in video games, mapping and statistical count of the remaining trees on the planet, editing and audio, summarisation and research assistance, traditional search engine ranking, auto response suggestions, photo modification, MRIs, CT, X-Rays, Medical imagining, sorting baked goods and detecting cancer cells (THIS IS SO COOL), drug repurposing, documentation assistance, coding, machine translation, speech recognition, internet moderation with specific reference to CSAM detection, vaccine detection, biochemistry, closed environment data retrieval, testing biology, chemical and genetic coding, software development drafting, clinical decision support, epidemiological modelling and forecasting, astronomical classification of known objects, transcription, archaeology, anomaly detection in data sets, weather forecasting, molecular biology, animation inbetweening and business documentation, quick explanatory writing/debugging of computing code (not always the best option for students), structural drafting of academic writing (can be a useful tool for students, so long as they know what aspects are (a) useful toolbox, and which are (b) plagiarism, avoid!), QA in call centres, employee briefing and P&WC Engine maintenance
MASSIVE thanks to the close to a hundred folks who responded to those threads. Please click through they’re all great follows.
Three interesting points to pull from that list. That’s the first time in months I’ve had equal responses from BlueSky and Twitter. The arts are almost nowhere in that list and most interestingly, of all? Analytical AI is cited time and time again as a force for real, measurable good. A force for good that’s continually obfuscated by the thief machines and the damage they’re doing. And obfuscated further, as Friends of the Lid Queen of Swords pointed out, by the complex environmental cost of A.I. in general. Here’s the Harvard Business Review on that:
Even putting aside the environmental toll of chip manufacturing and supply chains, the training process for a single AI model, such as a large language model, can consume thousands of megawatt hours of electricity and emit hundreds of tons of carbon.
There’s always nuance and remembering that is an enormous challenge. The thief machines lack it because they’re passable impersonations of real people from a distance. The venture capitalists and their minions lack it because the line goes up and if it doesn’t go up it’s made to go up in their vertical religion of ephemera. And now, we lack it. Because we’ve all been under siege from these things for so long, we’re all so tired and so very, very Angry
But now we know what we’ve lost, we can get it back. We have to get it back. Like fiction’s best known robot puncher said: ‘The future has not been written.’ Let’s write it ourselves.
Sources
https://aws.amazon.com/what-is/large-language-model
https://jasonsanford.substack.com/p/genre-grapevine-for-february-2024
https://today.ucsd.edu/story/7-ai-powered-technologies-you-should-know-about
https://slate.com/technology/2022/01/ibm-watson-health-failure-artificial-intelligence.html
https://www.nature.com/articles/d41586-024-00478-x
https://zapier.com/blog/best-ai-grammar-checker-rewording-tool
https://gamerant.com/the-last-of-us-2-enemy-npc-tattoos-detail
https://cohost.org/bluetinge/post/6306107-difference-between-p
https://futurism.com/ai-ceo-chatbot-convince-woman-euthanize-do
https://danieljosolder.substack.com/p/nanohellno
https://www.threads.net/@jomorgansloan/post/C_Z2EnPxEcC
Further Reading