Questions about Medium’s approach to AI-generated writing
Like you, I don’t want to read nonsense articles generated by AI. But those who want to BAN ALL AI! must consider the bigger picture.
Article about Medium’s approach
Recently, Scott Lamb — VP of Content for Medium, posted a thoughtful article describing how Medium is handling content generated using artificial intelligence:
How we’re approaching AI-generated writing on Medium
We welcome the responsible use of AI-assistive technology on Medium. To promote transparency, and help set reader expectations, we require that any story created with AI assistance be clearly labeled as such.
That seems pretty reasonable.
Community response
In comments in response to the article, many people reacted quite strongly against the idea of AI-generated content being allowed on Medium at all. Some said they would not read AI-generated articles and block any author who used AI. Some commented that using AI to create articles for Medium is unethical and argued that all AI-generated content should be banned. There was even some name-calling, and a few people argued that anyone who uses AI is not really a writer.
Questions
Using articles on Medium as a focus for discussion, here are my questions about some of the many issues surrounding AI-generated content:
1. How will you even detect it?
As many people commented, you’re going to have a very hard time reliably detecting AI-generated content. And there’s no way you’ll be able to tell if an author used AI to assist with their rough creative- and writing process.
2. What if the AI is helpful or even necessary?
In response to that article, someone commented that English is not their first language, so using AI tools to polish their article is really necessary.
- If I use speech-to-text to dictate my article, I’ve used AI.
- If I use a screen reader to listen to my draft article, I’ve used AI.
- If I write an article in French and then translate it to English using machine translation, I’ve used AI.
- If I use tools like Grammarly to edit my article, I’ve used AI.
- If I use a Google search to find additional resources, I’ve used a variety of AI technologies all in one.
Many authors already do those things. Those tools have become so common, it’s easy to forget they use AI.
What’s newer are these scenarios:
- Using DALL-E to produce images to accompany an article.
- Using chatGPT to list three perspectives on an issue, and then you write about one of those perspectives.
- You’re having trouble articulating a point, so you generate a paragraph making the argument succinctly using chatGPT.
- You need to crank out 10 articles per week to feed the algorithm that promotes your stories or you won’t make enough money this month, so you write 3 quick articles summarizing books you’ve read.
- You need to crank out 10 articles per week to feed the algorithm that promotes your stories or you won’t make enough money this month, so you generate 3 quick articles summarizing books you’ve read using chatGPT.
3. Where do you draw the line?
How “pure” must an article be? Of the previous examples of using AI in the writing process, which ones are acceptable?
4. Will authors have to justify their use of AI?
If I can prove I can’t use a keyboard, is speech-to-text ok? If I can prove English is not my first language, is Grammarly ok? If I can prove I have a learning disability, is generating text using chatGPT ok?
5. How would you label AI-generated content?
Would you have to list all the AI tools you used in your writing process? Would you show a number indicating how AI your article is?
- 0 = You wrote it on parchment
- 5 = You used a grammar tool
- 10 = chatGPT wrote the whole thing
6. Will labelling content as “AI-generated” harm some authors?
Will that non-native English speaking writer be banished to the algorithmic wilderness because they need to use AI to polish their articles?
7. If the issue is quality, does it matter if AI is used?
If someone reads one of your chatGPT-generated book summaries, is inspired to read the book, and the book changes their life, does it matter that the article was AI-generated?
8. Will AI-generated content enable new possibilities?
If creators learn to use prompt engineering like a musician uses sampling or mixing, might there develop an entirely new form of content we haven’t seen before?
9. Will there be a lotta terrible AI-generated content?
For sure.
10. Will writers (platforms) not using AI be at a disadvantage?
Yes. Is this the root of the anxiety people feel about AI-generated content?
Conclusion
Luddites weren’t wrecking machines because they just didn’t like machines. Workers weren’t raking the fires and pulling the plugs of boilers in the Plug Plot Riots of the general strike in Britain in 1842 because they just didn’t like steam engines. They were fed up with miserable working and living conditions, angry about how having money gave capitalists so much power in society, and fearful for the future because of the pollution and environmental destruction from coal mining they could see all around them.
Sound familiar?
Today, what with a global pandemic, war in Ukraine, a world-wide recession looming, and the climate crisis hanging over us (just to name a few things) we can be forgiven for being alarmed and even angry at the thought of being replaced by machines for the profit of the ultra-wealthy — again.
