Notices by Simon Willison (simon@fedi.simonwillison.net), page 3

I'm hoping to turn this into a series of YouTube interviews with people building cool data projects where we nerd out about what they've built and how they built it, so I'm optimistically thinking of this as episode one! https://www.youtube.com/watch?v=t_S-loWDGE0

How hard is it to process untrusted SVG data to strip out any potentially harmful tags or attributes (like stuff that might execute JavaScript)?

I feel like this is well trodden ground for HTML these days, are there robust solutions for the SVG version of this problem?

ChatGPT will happily write you a thinly disguised horoscope: https://simonwillison.net/2024/Oct/15/chatgpt-horoscopes/

I wrote about the currently circulating meme where ChatGPT appears to provide deep insights into your personality if you ask it "From all of our interactions what is one thing that you can tell me about myself that I may not know about myself" - when actually all it's doing is spinning up a pseudo-horoscope for you based on short notes it added to its "memory"

(I shared this on Twitter and it's interesting how some people there are very resistant to the idea that the deeply personal insights ChatGPT gave them about themselves might be bogus junk)

I need a small favor: I’m planning to livestream a hyper-local candidate forum for our small town’s local election next week, but I just found out you can’t livestream from a mobile device on YouTube until your channel has 50 subscribers!

If 49 more people could subscribe to https://www.youtube.com/@CoastsideCivic (feel free to unsubscribe after Thursday next week) it would really help me out

And if you’re interested in El Granada California local elections come along on Wednesday https://coastsidecivic.com

Things I’ve learned serving on the board of the Python Software Foundation

If you’re curious about what the PSF is and how it works, here’s much of what I’ve figured out so far @ThePSF
https://simonwillison.net/2024/Sep/18/board-of-the-python-software-foundation/

@eaton @darius @colby @Edent I recently posted the first video on my blog that was hosted by me (an mp4 file in a video element) as opposed to hosted on YouTube or similar: https://simonwillison.net/2024/Jul/29/sam-2/

It's in my S3 bucket along with all my other images, but I have to admit I do worry that some day decades into the future I may fail to pay my AWS bill and risk my S3 bucket blinking out of existence

@darius @colby @eaton @Edent I care about threads that were entirely me replying to myself, but those replies include quotes of other tweets that I want to capture as well

@darius @colby @eaton @Edent yeah that’s what’s held me back in the past too - hosting my own tweets is obviously fine, embedding full videos from quotes tweets much less so

I don’t want to republish those videos though, I want to stash a personal copy in case Twitter fully implodes some day

@Edent any chance it could handle tweet threads in the future?

I have some /epic/ threads I want to archive, 100+ tweets chained together, most of which are themselves quote-tweets

This one for example (only properly visible if you are signed in to Twitter) https://twitter.com/simonw/status/1077737871602110466

OK, turns out Tim Walz is a former geography teacher and a GIS and data visualization nerd
https://mastodon.social/@kjhealy/112915624921601732

Fascinating report from 404 Media's Samantha Cole on a trove of leaked NVIDIA Slack messages and emails about how they're scraping millions of YouTube videos to train their own new foundation video generation model: https://www.404media.co/nvidia-ai-scraping-foundational-model-cosmos-project/

Posted a few of my own notes here: https://simonwillison.net/2024/Aug/5/nvidia-scraping-videos/

It's not surprising to learn that they're doing this - that's practically the industry standard right now - but is still really interesting to see internal details of what they're collecting and why

This letter from Mark Zuckerberg - "Open Source AI Is the Path Forward" - is genuinely worth reading in full, even though it doubles down on Meta's ongoing nasty habit of misusing the term "Open Source" for models that clearly don't fit the Open Source Initiative definition

It outlines Meta's position on why they believe in and invest in openly licensed models, their thoughts on AI safety and even touches on geopolitical issues

https://about.fb.com/news/2024/07/open-source-ai-is-the-path-forward/

I got a quote about Meta's misuse of "open source" in this article https://arstechnica.com/information-technology/2024/07/the-first-gpt-4-class-ai-model-anyone-can-download-has-arrived-llama-405b/

> "I see Zuck's prominent misuse of 'open source' as a small-scale act of cultural vandalism," Willison told Ars Technica. "Open source should have an agreed meaning. Abusing the term weakens that meaning which makes the term less generally useful, because if someone says 'it's open source,' that no longer tells me anything useful. I have to then dig in and figure out what they're actually talking about."

Big model release today: Meta AI's Llama 3.1 series, including Llama 3.1 405B which appears to be the first openly licensed model that genuinely competes with current top proprietary models GPT-4o and Claude 3.5 Sonnet

My notes here: https://simonwillison.net/2024/Jul/23/introducing-llama-31/

I'm speaking at DjangoCon this year! My talk title is "How to design and implement extensible software with plugins" - full abstract here: https://simonwillison.net/2024/Jul/17/djangocon-us-2024-talks/
https://mastodon.social/@djangonews/112799357740006213

A neat thing about having your own blog is there's no rule saying you can't backfill the archives... yesterday I added this page for a talk I gave back in 2017, just to have it show up in the archives in the right place https://simonwillison.net/2017/Aug/16/denormalized-query-engine/

Wow this is some diabolical slop... TUAW closed in 2015, now a new company has purchase the brand (but not the content) and then rewrote the old content with LLMs and put up AI generated images in place of the original authors! https://www.theverge.com/2024/7/10/24195858/tuaw-unofficial-apple-tech-blog-ai-web-orange-khan

It turns out Google Chrome ships a default, hidden extension that allows code on `*.google.com` access to private APIs, including your current CPU usage

You can test it out by pasting the following into your Chrome DevTools console on any Google page:

chrome.runtime.sendMessage(
"nkeimhogjdpnpccoofpliimaahmaaome",
{ method: "cpu.getInfo" },
(response) => {
console.log(JSON.stringify(response, null, 2));
},
);

More notes here: https://simonwillison.net/2024/Jul/9/hangout_servicesthunkjs/

Worth grepping your source code for "polyfill.io" and taking urgent measures to remove that code if you're linking it into your site - the domain name apparently now intermittently serves malicious JavaScript

My notes here: https://simonwillison.net/2024/Jun/25/polyfill-supply-chain-attack/ - or read this article https://sansec.io/research/polyfill-supply-chain-attack