preserves

Disclosure: I work on VMs at Google Compute Engine :)

This was a really, really good write-up. I appreciated the breadth of VMs tested and the spread of benchmarks. A few random observations:

1. Turin is a beast.

2. The data on price-performance makes Hetzner look really fantastic, especially for small scale projects where region placement doesn’t matter much and big bursty scaling isn’t required.

3. I think the first ever cloud VM I ever provisioned was on DigitalOcean. I was surprised at how old their fleet was, but I guess they have some limited Emerald Rapids offerings now: https://www.digitalocean.com/blog/introducing-5th-gen-xeon-p...

A great example of Hyrum’s Law in action (https://www.hyrumslaw.com/).

This book supported the curriculum of my introductory computer science course at Tufts University back in 2013. It was hard but amazingly valuable. By the end of the semester, we were doing natural language classification using binary search trees and higher-order functions [1]. I think learning a Lisp-based language first gave me some very powerful foundational problem solving skills that have made me a successful software engineer today.

[1] https://www.cs.tufts.edu/comp/50PSS-2013f/homework/trigrams....

This is exactly the type of problem that a good parser should be able to solve, and training a parser on lots of data and throwing neural nets may indeed be a viable solution. Why wouldn't it be? The article describes how their architecture can help make sense of ambiguity.

In terms of a basic probabilistic model, P(meow | rug) would be far lower than P(meow | cat), and that alone would be enough to influence the parser to make the correct decision. Now, if the sentence were "The cat sat on the rug. It was furry", that would be more ambiguous, just like it is for an actual human to decode. But models trained on real-world data do learn about the world.