AI use will only increase; The role of data centres? [#48]
Published about 2 months ago • 3 min read
Hello Reader,
When ChatGPT first came out, it scared me. Sure, its output was often boring - and occasionally a little weird. But I could see how it significantly narrowed the gap between what I could create and the average writing out there. And it's only going to get better.
I’ve always believed in tackling challenges head-on. So even though I could write circles around the formulaic output of GPT-3.5, I did exactly that by using AI across a wide range of my writing assignments.
AI can be hard to master
This is a lot harder than it sound, however. Here's one personal story I often share to illustrate.
I had two stories to write, based on interviews with two different experts. For the first piece, I used AI to answer highly specific questions in bullet point form. I then had AI write them up into short blocks. I chose the most promising snippets and manually rewrote them into a cohesive piece, then edited it further. I loved the result - I estimated I saved an hour.
Enthused by that success, I used the same technique for the second piece. It didn't work. The AI made incorrect inferences - a real problem for the niche, highly technical cybersecurity topic. In the end, I had to rewrite everything to make it work. Frustration aside, I estimated probably would’ve saved 1.5 hours if I’d just written it myself from the start.
As I wrote this week in "AI can make your prose look good, even when it's not", AI can help you write better, but it won't make your arguments stronger or supply the logic you haven't developed yourself.
AI is being underestimated
Yet the latest models have advanced significantly. As AI influencer Ethan Mollick (professor at The Wharton School) recently shared, Robert Ghrist, a mathematician and associate dean of undergraduate education at the University of Pennsylvania, has been doing some fascinating things with it.
Last month, I also attended a presentation by GovTech's Chang Sau Sheong at Apidays Singapore (Read), where he demonstrated using n8n, a workflow automation platform with agentic AI capabilities to quickly build an app that sorted support tickets.
I suspect Robert Ghrist has a point: we're underestimating what AI can do.
Workflow of n8n demonstation set up by Chang Sau Sheong
AI and its impact on the data centre
Of course, you might be wondering: What bearing does the future development of AI have on data centres? This coming week, I’ll be presenting to a small group of senior data centre leaders at a private function.
Here’s what I plan to tell them:
Data centres are splitting into AI and traditional facilities.
Soaring costs will continue to shrink the pool of frontier AI players.
Fast followers in AI benefit from a massive cost advantage.
Inference hardware will start making serious inroads into data centres.
Everyone is using AI - they just aren't admitting it.
Past trends are a poor guide for forecasting AI data centre demand.
What do you think? Am I missing anything crucial?
Finally, despite doubling every year, DC Byte says just 11% of global data centres take-up from 2019 to 2024 are AI data centres. It's something worth thinking about in "APAC data centres poised to grow 2.5x by 2030."
As usual, you can reach me by replying to this email, too - I respond within the week.
Regards, Paul Mah.
Leave a comment!
You can now comment on stories published on techstories.co. If you received this newsletter from me, your account was created automatically.
All you need to do:
Scroll to Member discussion and sign in with your email for this newsletter.
An email message with a "magic link" is sent immediately. Click on it to log in.
Comment away - no password needed!
Spotlight
Google unveils CDU to unlock 1MW data centre racks
Fifth-generation CDU architecture will be contributed to OCP this year.
