Newsletter 44 – Look under the hood

I’m always apprehensive about showing people how I use AI apps such as Whisper and Marker. These are free and powerful apps that I use to transcribe audio and convert PDFs to text.

However, they use a command-line interface, which for us Gen X geeks is known as “the 1980s DOS interface”. You have to type text commands to get things to work. The personal computer revolution happened when Apple and Microsoft popularised the graphical user interface, where we pressed simple graphical buttons instead of memorising cryptic commands like “cd c:\user\my_documents” just to look for a file.

Things got more convenient with touchscreen phone interfaces, where buttons were replaced by swiping actions. Computing devices have become very easy to use, but in the process, fewer people understand how they work.

Whenever I show people a gaming graphics card, aka a GPU (graphics processing unit), in my AI workshops, very few actually know what it is. Or that ChatGPT runs on similar hardware, which in turn explains why GPU-maker Nvidia is the most valuable company in the world.

It seems that the more knowledge and technology we have access to, the less we know about what is under the hood. Everything has been glossed over by UI and UX (user interface and user experience), and the human is content to just press buttons and swipe on screens.

This is one of the reasons why many people don’t understand why Gen AI doesn’t give them consistent answers or why it hallucinates. Once you understand that Gen AI strives to give you probable answers, rather than fixed answers, you will learn to apply a healthy dose of skepticism whenever you use it.

Using Google AI Studio, I created a simple demo of how it figures out the next probable word to generate in any sentence. Try it and you’ll see how Gen AI guesses what to generate next, rather than actually “knowing” the right thing to say. Because its output often sounds so polished – sometimes better than what we can write ourselves – we tend to take its answers for granted.

We should always try to understand how things work instead of accepting them blindly. This is fundamental to critical thinking and figuring out how to survive the disruption that is coming our way.

What I wrote recently

You need to apply love when communicating with different audiences.

Experimenting with GaN chargers for my work laptop.

68% of SG youth are open to changing jobs, says a survey. I think the percentage should be higher.

Why are people using templates when trying to make connections on LinkedIn?

Gen AI companions are all the rage. But if you are AI-literate, you will know the robot doesn’t care for you.

I turned the latest Advanced AI Masterclass group photo into a Chinatown oil painting. The dates for our 2026 Masterclasses will be 6th Jan (Fundamentals) and 20th Jan (Advanced). More details here on ST Skills.

OpenAI launched AgentKit to help you make agents. It’s not as easy as vibe-coding though, you do need some coding knowledge.

Unlocked NYT articles

A Plea From Doctors: Cool It on the Supplements

Recruiters Use A.I. to Scan Résumés. Applicants Are Trying to Trick It.

10 Small Things Neurologists Wish You’d Do for Your Brain

That’s All, Folks. Thanks For Reading!