Why I write my posts by hand in the age of AI content

Social media is full of grandiose posts designed to attract "likes." The problem? Many are completely wrong and end up misleading thousands.

I keep coming across posts making outlandish claims or touting conclusions that are plain wrong. And it's getting worse. It's Saturday, but I couldn't wait for another UnfilteredFriday to talk about this.

Context is everything

I'm not an expert in everything. But some posts about data centres and AI deployments rile me up. Here are two such posts this week.

The first claim: Meta's longtime chief AI scientist Yann LeCun plans to leave and launch his own startup, showing that the future of AI isn't with the tech giants. Another claim: An older story about China's undersea data centres with a low PUE "proves" it has solved the cooling problem to outcompete Western firms in AI.

When making an argument, the context matters. Here's how it looks once I add in what I know. For the first claim, the missing context is this: Meta has spent multiple billions this year to build a new "Superintelligence" AI team headed by prodigy Alexandr Wang. And oh, Alexandr was made Yann's boss. Make what you will of that, but the previous claim is clearly untenable.

For the second claim, the missing context: I'm no fan of the chip war. But the facts are clear - the barrier to AI leadership isn't cooling but access to chips. Anyway, at a 9% failure rate of GPUs per year according to Meta's data, putting GPUs in such a hard-to-access area would be foolish.

Unwitting untruths

Now, I don't think there's an intent to mislead. But when you use AI to repurpose a single story into a sensationalised post for "likes," this is what you get. That's why I write differently. I avoid covering topics I don't know much about. I do additional background research. And I write my posts by hand.

I'm not averse to using AI, but I've found that it struggles to pack insights into a concise post that's both clearly expressed and easy to read. So instead of pushing the cognitive load onto readers with AI-written text, I frontload the effort on my end so that readers get it immediately.

Read selectively

We are constantly advised to read broadly. This remains good advice.

But in the age of AI, where stringing together smooth-reading text is no longer the barrier to entry it once was, I'd suggest readers be discerning and read selectively, too.

And don't just believe everything you read online.