Hello Reader,

I had a most intriguing call this week with the founder and former CEO of a unicorn firm that got me thinking more about AI and the future of work.

At the cusp of the discussion is the rapid advancements in AI, and what the future bodes. More specifically, can AI be harnessed to produce content with human-like nuance with the inexhaustible output of large language models?

AI writing

Pfft, AI writing as well as humans?

When it comes to using AI to write, journalists I've met fall mainly into two camps: The deniers and the skeptics. The former won't touch AI with a ten-foot pole, while the latter are often quick to cite newsroom guidelines against AI use or bemoan the insipid nature of AI output.

The situation is more varied with content writers. Here, the group that stands out would surely be the true believers, often found on LinkedIn touting prompt packs to change your life. Unsurprisingly, some in this group also market their use of AI as a supposed differentiator.

What do I think? Well, using AI to produce exceptional prose isn't as simple as it looks. I speak from deep experience, having used AI practically every single day since the public release of ChatGPT.

AI scared me

If you've read my LinkedIn posts on writing, you would have known how AI scared me despite 15 years of full-time writing. While I remain confident of producing clearer, sharper content, it would be the height of folly to pretend AI has not dramatically increased the average writing quality of everyone.

Certainly, the gap between average and what I can produce is now much narrower than in the past. And it continues to be whittled down with each AI model released.

As I wrote last year in "The job losses have already begun":

I'm a realist. The generative AI genie is out of the box and there's no putting it back. Neither am I going to hide my head in the ground and wish everything away, though this might perhaps be influenced by the fact that retirement isn't an option for me now.

Which is why I am determined to grasp AI by its horns, metaphorically speaking.

AI is getting better

By the way, have you seen how good the latest AI models are? Sure, pre-training scaling, which kicked off the massive data centres builds we see today, and even reinforcement training appears to be stumbling and offering diminishing returns. But new test-time techniques appear to be unlocking further improvements.

I've also been using "reasoning" models such as DeepSeek R1 and Claude 3.7 Sonnet since their release. When paired with Internet data (Perplexity), they produce very compelling outputs indeed. And I'm not even going into the new "Deep Research" modes.

So, AI continues to get better, slowly but surely expanding its capabilities across more complex tasks and domains.

Is native AI writing next?

Back to my call.

Today, I tap on a range of AI tools when I write. Like how a chef uses a variety of knives and utensils for different culinary tasks or how a master craftsman selects the right tool for each specific job, I seek to write better and more efficiently with the appropriate AI technique or service.

Somehow, I don't think this is the end point.

Much like how it took a while for cloud applications to evolve into today's cloud native approach, I believe a "native AI" approach to writing could be just over the horizon.

Not the lazy, mundane copy created from a single prompt or three. But a new way of writing that melds genuine creativity with AI assistance to produce original content in an extraordinarily efficient way - much like how modern digital art software fundamentally transforms the creative process.

What do you think?

PS: There are quite a few data centre stories this week. My favourite story is probably my visit to SuperComputing Asia here. As usual, all feedback welcome - simply hit "reply" to this email.

Regards,
Paul Mah.

Spotlight