Hi Reader,

Last week, I got an unusual message on LinkedIn from someone at a company called Mercor.

They were looking for experienced data scientists for an "ongoing project", with a premium daily wage.

I was intrigued, so I looked at the role. It involved completing data science tasks to help train AI models.

Experienced data scientists and data engineering are among the highest-paid professionals, among others.

In a way, I totally get it...

AI model trainers are running out of high-quality Internet data to train their models. So now, they're paying premium to get access to the brains of real experts, in each domain area, to make their models smarter.

As the old data science adage goes: garbage in, garbage out. I’d add: diamonds in, diamonds out. (well, I made that one up 😏)

But honestly, I just don’t know what to think about all this.

Yes, AI models are very useful, doing the grunt work that used to crush your soul. It's such an exciting time to witness the progress.

At the same time, there’s something quite icky, unsettling about helping train a kind of “digital overlord” that could outsmart highly skilled people at their own jobs...

...the people who’ve spent years and decades honing their craft.

I kept wondering:

Who am I without the "expertise" and "talents" I’ve always thought of as my own?

I wish we had more honest, in-depth conversations about what all of this means, not just about what the latest AI tools can do.

***

A few days ago, I chatted with a data scientist friend whose team is building an automated engineering platform.

The goal... is to automate engineering work and reduce the company’s engineering headcount by 60% 😨.

It’s hard to picture what the world will look like in a few years. And I especially share concerns with people who are just starting their careers. How will they get enough practice to build their problem-solving skills and judgment?

Oh by the way, you might be giving AI chatbots free training data without even knowing it...

I recently made a video (below) about how to protect confidential data when using AI tools like Claude Code to analyse it. The video got amazing engagement.

​

It turns out, many people were not familiar with solutions like generating synthetic dataset when using AI.

This also makes me think local AI and data sovereignty will become a big topic in the future.

👩🏻‍💻 A little update

I’m giving a masterclass on GraphRAG in London on 11 Nov. I’ll be talking about how graph structures can help a RAG system connect information and reason better.

The session is 2 hour-long. And I’m terrified, to say the least... If I can barely keep people engaged in a 10-minute video, how am I supposed to do it for 2 hours? 😅 And let’s hope the live demo doesn’t break!

I'll be in London for 2 days only, but if you're around, let me know!

Have a great week,

Thu

PS: Should I take that Mercor gig? 🤣