This is the thing that frustrates me about ML/AI.

It’s *trivial* to make a system that predicts IQ/personality etc from an image. AI/ML is just a function approximator. Data in, prediction out.

Making it *accurate* is the problem.

BECAUSE IT’S NOT A SIMPLE FUNCTION.

Today in shitty machine learning startups, this company claims to predict IQ, personality, and violent tendencies by applying deep learnin…
twitter.com/jbensnyder/status/

invest in stocks

chicken, beef, vegetable

you'll be a bouillonaire :blobcat:

This the first article trying to "bury" REST that I find actually making sense: tbray.org/ongoing/When/201x/20

Because it doesn't really bury it. It's a high-level overview of alternatives for cases where REST doesn't cut it. And, well, Tim Bray knows what's he talking about, and is one of the most coherent tech explainers on the planet :-)

@veer66 @yogthos they're indeed processes by default. There are a few other types of workers in Gunicorn, including threads. uwsgi (another popular Python app server) is similar.

@yogthos But all of that is beside the point, because server-side Python is mostly parallelized either with forks or with async I/O. It was never a problem, really. Looks like the author simply doesn't know that Python is one of the major languages for implementing Web servers and only heard of its data science applications.

@yogthos no, it's the other way around. GIL *doesn't* block on I/O, so many times it's perfectly normal to use threads in a networked Python applications. However it *does* block on everything the interpreter does, so a multi-threaded program doing something interesting in the actual Python code would suffer from GIL. (1/2)

@yogthos I actually stopped reading after he said Python was slow because of the GIL, and that it's not suitable for the server side because of that. I mean, you can be wrong in many details when you're reviewing 5 languages, but that one was a fundamental blunder!

@freakboy3742 snow at 1°C is a pretty shitty thing… Take it from me :-)

@plsburydoughboy thing is, it's "if", not "when". There's no real reason or incentive for it to happen.

@yogthos that would've been handy for my deployment pipeline… if it wasn't discontinued a few weeks ago :-(

I still don't understand how Mac became a de-facto "hacker" platform, if any unix-related things are done so poorly they would shame a 20 year old Linux setup… Like, packages do use `/usr/local/etc/git_completion.d/` to add their custom scripts, but bash doesn't read this directory.

PSA for all of you Mac fans: you'd be *amazed* by the attention to details on any modern Linux system, where everything pretty much works out of the box.

You don’t schedule having a shower in the morning. it’s basic hygiene. You shouldn’t be scheduling tech debt management. It’s basic sustainability for your codebase. - @kll@twitter.com

"In RPython you don't write JITs, you write *hints for the JIT generator that generates a JIT for you*"

(This is about Pixie, a -inspired lisp implemented in RPython.)

youtube.com/watch?v=1AjhFZVfB9

Show more
Mastodon

Follow friends and discover new ones. Publish anything you want: links, pictures, text, video. This server is run by the main developers of the Mastodon project. Everyone is welcome as long as you follow our code of conduct!