Published
Weekend Reading — 👌 6-finger deniability
This week we have a new search engine we’re actually excited about, the results from the 4-day workweek trial are in, and we have questions about craters.
At last, I managed to catch up, and this week's Weekend Reading actually gets published during the weekend. Phew. Will be in Las Vegas this Monday/Tuesday if anyone wants to grab a coffee.
Tech Stuff
Phind “AI search engine for developers”, which means you get AI suggestions alongside the familiar blue links. I’ve only been using this for a few days but this might be a keeper.
Phind summarizes search results, so it knows about current events, new blog posts, API docs, etc. It's far less likely to make things up.
For example, you can ask it “what’s new in Next.js 13?” and get a decent response. And it will patiently explain answers, including setup instructions, code samples, etc.
It only returns 15 search results for each query – I guess because the AI could only summarize that many — but I like this UI, it feels cleaner and simpler than Google.
And you can customize the search results by ranking the site it uses. BTW the default setup gives github.com a plus, and quora.com a double minus, so there’s that.
@clack/prompts Delightful UI and a simple API for building CLIs with prompts. Comes with text, confirm, select, multi-select, and spinner components.
He3: Modern Developer Toolbox Desktop app with a variety of tools: formatting Javascript, hashing text, encoding text, JSON to YAML conversion, etc
Rules for naming things:
Don't name a new thing "new", it'll be old before you know it.
Don't name an old thing "old", you'll end up having two or more old things.
Chrome’s Headless mode gets an upgrade I just learned that “the old Headless was a separate, alternate browser implementation that happened to be shipped as part of the same Chrome binary. It doesn’t share any of the Chrome browser code in //chrome”
The new headless mode is more compatible with regular Chrome, and also includes CLI options for dumping the DOM, taking screenshots, and printing to PDF.
You enable it using the --headless=new option.
(Apparently, the Chrome team didn’t get the memo about naming things "new")
Strudel REPL Live coding environment to write dynamic music pieces in the browser.
total-typescript/ts-reset A “CSS reset” for Typescript? It does fix some of the annoyances with TS:
fetch.jsonandJSON.parsenow returnunknown(vsanywhich causes subtle bugs)filter(Boolean)behaves EXACTLY how you expect!- some other things but really these two made all the difference for me!
Safari 16.4 Is An Admission Do we thank the EU for that?
This is the power of just the threat of competition.
Apple's representatives have offered browser-based claims in court and in regulatory filings to defend App Store rapaciousness. They argued that if developers don't like its generous offer to take only 30% of their revenue, there's always Cupertino's highly capable browser to fall back on.
Replicate Lets you run open-source machine learning models with an API request.
There's more happening in AI than ChatGPT. Much of it comes in the form of research papers, which are free as in "gratis" but also free as in “good luck figuring out how to use that”.
Replicate hosts these models and gives you a simple UI to try them out, and a consistent JSON/HTTP API to use them in production.
For example, if you wanted to experiment with ControlNet, which I cover later on, here’s the page with the models.
ChatGPT and AI could completely replace the need for StackOverflow for coders as long as it constantly told me my question was irrelevant, had already been answered, and berated me for not doing my own research first and then downvoted me.
Who called it “scrum master” and not “jiraffe” 🦒
Business Side
World’s biggest four-day working week trial hailed a ‘major breakthrough’
- The number of sick days taken dropped by around two thirds
- 39% of employees reported being less stressed
- 57% decrease in staff leaving (despite the “great resignation” period)
- However, concerns raised by some staff about increasing workloads and less sociability
- 92% of companies confirmed they will continue with the four-day working week
David Graeber got an economist to admit that he was not aware of single case where a company was fined more than the profit it turned breaking the law. He summarized this as the government saying: "Do all the crime you want, but if we catch you, you have to give us a cut.”
Machine Thinking
I swear I've seen this town before New AI generative tech just dropped: ControlNet uses a second input to control a diffusion model.
In addition to the prompt, you control the output with a sketch, pose, edge map, depth map, etc. This Reddit post shows different images generated from the same depth map.
You get more control than img2img or style transfer, and unlike fine-tuning — which requires a lot of processing — this tuning happens during diffusion, so you can do it on a personal device.
It also changes the conversation about generative AI art. We’re moving away from “write a prompt and hope for the best”, to where you have much more artistic input and control over the result.
People keep anthropomorphizing AI. Here’s why It’s human nature to anthropomorphize things: pets, plants, vehicles, appliances, anything and everything. We grow up reading about a tank engine that talks, we buy cars based on how their face looks, and say things like “the camera doesn’t like me” or “my computer is being stupid right now”.
I also think that we need new language to talk about AI, because AI is not at the same level as anthropomorphizing a toaster oven or a monstera plant. AI has the ability to mirror us — which not even dogs can manage very well — and that’s on a different level.
To summarize, we offer four thoughts. Developers should avoid behaviors that make it easy to anthropomorphize these tools, except in specific cases such as companion chatbots. Journalists should avoid clickbait headlines and articles that exacerbate this problem. Research on human-chatbot interaction is urgently needed. Finally, experts need to come up with a more nuanced message than “don’t anthropomorphize AI”. Perhaps the term anthropomorphize is so broad and vague that it has lost its usefulness when it comes to generative AI.
'It's Hurting Like Hell' Related:
The reality of users’ experiences is more nuanced than the company’s ads depict, however. Replika is a tool for many people who use it to support their mental health, and many people value it as an outlet for romantic intimacy. The private, judgment-free conversations are a way for many users to experiment with connection, and overcome depression, anxiety, and PTSD that affect them outside of the app.
Insecurity
Internet Crime Complaint Center (IC3) “The FBI recommends individuals take the following precautions: … Use an ad blocking extension when performing internet searches”
Criminals will start wearing extra prosthetic fingers to make surveillance footage look like it's AI generated and thus inadmissible as evidence.
Everything Else
ConsultingHumor Timeless advice:
“never underestimate the power of cc’ing someone’s boss”
I was doing an activity with my youngest and then told him I had to go back to work. He motioned to my laptop and said "Okay, go touch your letters daddy." which is such an amazing burn.
David Bisset "What's the definition of bad timing?”
What I love about this story is:
- the immediate & categorical response: 2 tail strikes within 6mins = shutdown all Alaska Air flights
- no bureaucratic process: One individual made the decision
- “That looks about right” gut check of automated data
- software sucks but it sucks worse when lives are at stake
- this sort of incident response is a credit to the aviation field and should be a model for all
Cats and Kittens “For rent .. boat with professional crew”
There is a worrying amount of fraud in medical research The replication crisis, but on a totally new level:
Yet journals can take years to retract, if they ever do so. Going by these numbers, roughly one in 1,000 papers gets retracted. That does not sound too bad. However, Ivan Oransky, one of Retraction Watch’s founders, reckons, based on various studies of the matter and reports from sleuths, that something more like one in 50 papers has results which are unreliable because of fabrication, plagiarism or serious errors.
daria-andrea “anyways, have a wonderful weekend Y'ALL 😁🎉”
