LinkedIn has an AI problem and it doesn't care

Plus shiny happy ugly AI people

After a planned break for a holiday in Cork, and an unplanned break for half my family being sick, Explainable is back.

Charles Bukowski, the David Foster Wallace for guys who want to romanticize their drinking, cropped up last month in a thread from a hustle bro about expanding your mind. It got outsized attention because Bukowski is as far from LinkedIn grindset culture as you can imagine. It got rightly mocked, but it also prompted responses like the following:

“Intriguing! Bukowski's raw and unfiltered view of life certainly has a way of gripping the soul. It's fascinating how his words can transform perspectives and challenge the status quo”.

That’s the most AI-generated post I have read in the past 12 months, and I have read a lot of AI-generated posts. Maybe the writer just has a certain AI-style cadence. But it got me thinking about the one style of writing that AI has mastered, soulless LinkedIn-style comments.

I share this newsletter on LinkedIn most weeks. It tends to get more engagement there than on Twitter/X. Many journalists I know have reluctantly developed a LinkedIn strategy since Musk torpedoed their previous online audience. And anyone I know in media looking for work - so a lot of people - uses it. LinkedIn is leaning into news reads one report, it’s going for short-form video reads another. While every other social media platform is battling to stay relevant LinkedIn is having its moment. 

But, and this may pose a slight problem for LinkedIn, it’s a truly awful platform. I’m posting on LinkedIn because it may boost my traffic and it may lead to work. The driving concept behind anyone posting on the platform is to earn more money, whether that’s securing a new job, building an audience, or building a network. Which is fine, we’re all entitled to want better pay and greater job security. But it makes for a grimly necessary social media platform rather than a fun one. And LinkedIn short-form video or news content (especially news content) won’t change that. 

If Facebook was originally the place to share things with friends before slowly morphing into the social equivalent of Homer’s web page, then LinkedIn is threatening to go down the same route, except with a less fun initial USP (look at me casually deploying USP as a term, so LinkedIn)

Which brings us back to AI. This newsletter cannot be written by AI. No piece of feature writing can be written by AI. No decent fiction can be written by AI. Some functional writing can be written by AI, with heavy human input.

But what AI can currently do, is a serviceable job of writing LinkedIn content. LinkedIn even flags it as a feature, not a bug, through an AI writing tool .

Why can it do a good job with LinkedIn content? Because its writing that is stripped of humour, of rough edges, of honesty. It’s a learned language, one we’ve all learned from the day we first had to compose an ‘offend no-one’ reply-all email.

LinkedIn has two options here, they can say ‘screw it, who cares’ and let everyone post their AI-generated filler, safe in the knowledge that people still need LinkedIn. Or it can go hard figuring out how to detect AI writing, despite no current silver bullet tools to detect it. It could assign a value to people crafting their own message and communicating it in their own flawed way. There is little to no chance they will choose the latter option.

Will that matter to their bottom line? Maybe. We will likely continue using LinkedIn for its core function, but begin to avoid the newsfeed (does that sound like another social platform you know?).

And, maybe, it’s a long shot, maybe in this sludge of AI-generated ‘great insight’ the ability to write well with some wit - not great writing, just fun-to-read honest writing - will have a premium placed on it. I’m choosing to cling to this hope. If this was a LinkedIn post I’d sign it off with a misappropriated Charles Bukowski quote but it’s not so I’ll just link to him writing about writing.

---

Small Bits #1

Once you weed out the AI bros, It’s a small pool of creative people showing how they are using AI tools to enhance their craft, rather than mask lack of craft. I like Martin Nebelong’s stuff because he’s honest about AI limitations (he frequently cites the random ‘slot machine’ nature of text-to-image/video tools). Plus, he’s an actual digital artist, not someone coming to the form post-Dall-E. You can read/watch him on 3D art here

---

Small Bits #2

This is trending on Chat GPT Reddit. A user claimed that GPT readily created an image of an ugly man but pushed back at a prompt to create an image of an ugly woman. You can see the full post here. It’s trending because, well, there is no shortage of very online men eager to believe they are discriminated against. I tried the same prompts, along with some variations, and GPT made me ugly pictures of both genders. It did push back on the use of the word ‘ugly’, but the responses were the same whether I entered ‘man’, ‘woman’ or ‘person’. This won’t stop the above claim from getting traction on some dark corners of the internet.