"You spend the entire week learning how to install CUDA and after that masochistic exercise you still barely see any improvements."

10 years since ImageNet and you still have to download a PDF to install CUDA.

Expand full comment

"Every hardware vendor is the cheapest, consumes the least power, has the most innovative design and runs BERT the fastest. Every dev tool is the fastest, has a modular minimal design, has the best community and most industry usage."


You know the product doesn't offer much when they tell you features not benefits.

Go to the homepage of almost any GCP or AWS service and every single one with tell you: scalable, zero-downtime, no servers and my favorite...

"networking capabilities for hybrid and multi-cloud scenarios"

What do any of these even mean?

Where can I run my ML model code?

Expand full comment
Oct 19·edited Oct 19

Brilliant article. I did have a little cry when it got to close too describing my actual life, but blinked away the tears and carried on. If I had a pound for everytime a technical pre-sales rep tells me their cloud platform does "out of the box" a difficult ML task like model drift (which requires ground truth and is literally never out-the-box), I would have around five or six pounds. But still, that's quite a lot seeing as there's only 3 main cloud providers.

Expand full comment