Notices by Adrian Cochrane (alcinnz@floss.social), page 54

If there's no overlaps to consider only a fraction of that work needs to be done. Namely what I phrased as "decomposing the path & tracing the outline into the output bitmap", without need for its outerloop.

This is split into substeps: splitting the shape & generating the bounding box.

There's some legacy functions that are commented due to memory issues.

The vast majority of the code goes into efficiently computing different tweening functions. There's also debugging functions.

2/4?

Incidentally several of today's links came from: https://www.baldurbjarnason.com/2023/open-offices-links-notes/

There's a few more there you might find interesting.

FreeType's core logic to compute a signed-distance-field for a vector outline *with overlaps*, after additional validation & initialization, involves counting contours, allocating sidetables to store auxiliary info based on that count, iterate over the contours carefully geometrically decomposing the path & tracing the outline into the output bitmap, iterates again over all pixels tested against all contours to fill in their values, & cleans up.

1/3?

South Korea’s banking security: Intermediate conclusions - Almost Secure: https://palant.info/2023/02/20/south-koreas-banking-security-intermediate-conclusions/

Announcing Squire 2.0—Fastmail releases next generation of open-source rich text editor - Nell Jenkins @ Fastmail: https://www.fastmail.com/blog/squire-2-0-fastmail/

Defending against AI Lobbyists - Bruce Schneier: https://www.schneier.com/blog/archives/2023/02/defending-against-ai-lobbyists.html

Tech Was Supposed to Make Cars Safer. It Didn’t Deliver - Paris Marx: https://www.disconnect.blog/p/tech-was-supposed-to-make-cars-safer (Substack)

To be sure, I did not anticipate the surge of new renewables which (thankfully!) allowed us to reduce GHG emissions faster than we would have if we relied only on natural gas. But that transition was also market-driven. Cleaner is cheaper, and markets like cheap.

Wtf you can't enable unstable rust features on stable builds?

... what if these things stayed the same and this is an old version?

Use Your Illusion - Dan Sinker: https://dansinker.com/posts/illusions/
Simon Willison quoting it: https://simonwillison.net/2023/Feb/20/dan-sinker/

Conditional Tagged Templates - Baldur Bjarnason: https://github.com/baldurbjarnason/conditional-tagged-templates (GitHub)

Vibe Driven Development - Robin Rendle: https://www.robinrendle.com/notes/vibe-driven-development/

Notetaking, Tagged Templates, and How Deno is a Clear Improvement Over Node - Baldur Bjarnason: https://softwarecrisis.dev/letters/tagged-templates-why-deno-is-good/

What are fundamental skills for a web developer? - Go Make Things: https://gomakethings.com/what-are-fundamental-skills-for-a-web-developer/

https://rhapsode-web.org/ has been updated after my move to a self-hosted SourceHut! From a more rudimentary self-hosted CGit.

The updates I'll be making in a couple days to Haphaestus's & Argonaut's sites will have a bit more to them...

@alcinnz: I recall seeing jokes that Go was ALGOL 68+41.

@RL_Dane Tricky, and I definitely see the temptation.

Though the standard complaint is the fuzziness of the concept of "commercial".

I'll say (A)GPL is close enough!

What exists of my "Argonaut Suite" so far has been moved to my new codeforge!

Auditory browser
Rhapsode repository: https://git.argonaut-constellation.org/~alcinnz/rhapsode
Rhapsode issue tracker: https://todo.argonaut-constellation.org/~alcinnz/rhapsode
eMail: ~alcinnz/rhapsode@todo.argonaut-constellation.org

Webpage debugger
Amphiarao repository: https://git.argonaut-constellation.org/~alcinnz/amphiarao
Amphiarao issue tracker: https://todo.argonaut-constellation.org/~alcinnz/amphiarao
eMail: ~alcinnz/amphiarao@todo.argonaut-constellation.org

The code now compiles again!

But does it work?

It works! A little program for the grid, hastily written in Lua.

Why is this interesting?

The recent eye-popping advances in AI have come from models that scan huge datasets, either generated by humans (e.g. “all competitive Go games” or “all the text we could find on the web”) or by computer (“the AI plays itself a billion times”), and imitating the patterns in that dataset — with no underlying model of meaning, no experience to check against, no underlying theory formation, just parroting.
3/

The Go AI was trained by feeding it a huge number of Go games. It built a model based on w̶h̶a̶t̶ ̶h̶u̶m̶a̶n̶s̶ ̶d̶o̶. CORRECTION: KataGo is trained by playing against itself; the model input is “past AI games.”

The human beat it by doing something so obvious a human or sensible AI would never do it — so it wasn’t in the training data, so the AI didn't counter it.

(Basically, the human forms a conspicuous giant capture ring while distracting the AI with tactical battles the AI knows how to counter.)

#ai
2/

Computers have been beating the best human Go players since 2016. The Go world champion retired in part because AI is “an entity that cannot be defeated.”

But a human just trounced one of the world’s best Go AIs 14 games to 1: https://arstechnica.com/information-technology/2023/02/man-beats-machine-at-go-in-human-victory-over-ai

I think this news story is more interesting than it might first appear (without knowing details, so grain of salt). It isn’t just a gaming curiosity; it points to a fundamental flaw with “deep learning” approaches in general.
1/