This reminds me of those videos where the bar staff try blind pouring a shot, and it's wild how good some people are. Would love to see a similar competition, re: can the most senior store members be accurate to 1° :)
I'm imagining a spectrum between "has to be slow" and "needlessly slow", with a middle slider for that one razor where things take as much time as you give them.
Intentionality is a big theme in math research (so i've heard), where solving "useful" problems isn't the ideal goal. The goal is to solve interesting problems, which might seem useless, but along the way achieve results with much wider implications that would have been impossible to discover directly. Or, how inventions like toothpaste came from space travel research.
(rhetorically) Does an indirect result "justify" a longer, slower project? Is speed an inherent property of the problem, or is it only knowable once it's complete? Or both, in the cases of misused funds?
After reading this, I am now of the belief that software is, actually, a good thing. And that programming can be enjoyable.
Seriously, this does such a good job of capturing the feeling of MAGIC that code is capable of -- both in its process and in its output. Textbook "craft". It's hard to experience that sometimes when surrounded by dependency hell, environments, build systems, certain dynamic programming languages, and the modern web ecosystem.
What is the benefit of this over something like Pkl[0]? Pkl compiles down to JSON, YAML, etc., but the language itself is user-friendly. This way you get best of both worlds: readable and editable for humans, and parsable for computers.
That's a pretty egregious mistake for a designer to make -- and that's not even mentioning the lack of accessibility. WebAIM's contrast checker says it's a 1:1 contrast ratio!
If someone is releasing a model that claims to have a level of reasoning, one would hope that their training dataset was scrutinized and monitored for unintended bias (as any statistical dataset is susceptible to: see overfitting). But if the graph on the announcement page is literally unreadable to seemingly anyone but the creator... that's damning proof that there is little empathy in the process, no?
> It may not sound like much, but this was a lot of work. I spent days just thinking about this problem, trying to understand what is happening now and how to improve it.
I find these kinds of problems the most fun and the most educational! I tried building a grid layout system from scratch in SwiftUI, and it was similarly tricky to map out:
- what the "ideal" behavior one expects is,
- what edge cases exist,
- how to handle the edge cases,
- maintaining ideal behavior while handling edge cases.
(It was tricky b/c SwiftUI lays out its children, then its parent -- so the parent needs to ask its children for its view size, and iterate and set rows/columns that way.)
Maybe because the problem seem simple, it is that much more fun to dig into. Some good ol' time with a whiteboard.
Cabinette[0], a macOS app for musicians/producers to manage their music catalog. My WIP folder has about 75 songs, ranging from electronic/orchestral/piano, and I wanted something with dedicated filtering, tracking progress over time, and pretty drag/drop. So... I built it!
Currently in TestFlight, but 1.0 is launching soon :) Feel free to give it a spin, and drop thoughts on the feedback board.
(Another idea I'm starting is an easy-learn, hard-to-master, productivity app: split calendar and todo list, with fluid drag/drop and power user features. Targeting the iPad at first, and will hopefully bring to more platforms. Keep an eye on my website[1] for launch!)
I'm reminded of a time that an intern took down us-east1 on AWS, by modifying a configuration file they shouldn't have had access to. Amazon (somehow) did the correct thing and didn't fire them -- instead, they used the experience to fix the security hole. It was a file they shouldn't have had access to in the first place.
If the intern "had no experience with the AI lab", is it the right thing to do to fire them, instead of admitting that there is a security/access fault internally? Can other employees (intentionally, or unintentionally) cause that same amount of "damage"?
I like thinking of CS theory as "math, with more hand-waving". Or, I can't remember where I read it, but something about CS being the mathematics of asymptotes.
The first image on the marketing page is an AI-generated picture of a person staring back. At best it's unsettling, and at worst it makes me wonder how authentic of a user experience the app can offer. Above-the-fold, first landing page content is valuable real estate. Ethically/copyright-ambiguous, inauthentic marketing is a strange thing to value imo.
According to the article/filing, she wasn't given specific reasons for the demotion, treated discriminatorily, and given a PiP that was unattainably difficult.
One would assume that choosing to file litigation against Amazon would be done thoughtfully / with a plausible rate of success; they have an army of lawyers. What informs the opinion that a PIPed employee isn't worth even listening to, esp. from a lawsuit? Amazon is not exactly a shining star for work culture, and this situation doesn't sound unfathomable.
I think it more targets the “apple enthusiast”, or at least the “tech enthusiast”, who is used to paying a $$$ premium (at the cost of polish) to be the First at a new thing. Every company does it but Apple has a strong pattern of “v1 is setting up for v2” — e.g., the original iPhone being expensive af with no apps, but getting 3G and apps the next year.
Also, I’d wager the people buying this can mostly Certainly afford it, given how expensive it is. For those people the money is meh, or justified (business expense, app platform). For others who are Maybe affording it, returns seem like an easy failsafe if it doesn’t meet expectations.
These models aren't actually doing any musical comparison -- they are trained on audio, and from audio, piece apart with a "note" and a "melody" and an "instrument" are from the labelled training data. No intentional theory is being done!
Algorithmic music composition has usually been split into two:
1. Generate notes (re: theory, genre)
2. Generate sound
(i.e., EMI[0], Kulitta[1], MusicNet[2])
Now we are doing both at the same time, and backwards. The model isn't (necessarily) going "write melody, then generate the sound", but rather, "here are 500 songs that are described with X, 500 with Y, and you want XY, so we'll combine these two" :)
(This is my best understanding, so feel free to correct)
They used to have this in Seattle (Car2Go (Merc), and ShareNow (BMW)) but they had a messy back-and-forth acquisition, shutdown, then apparent resurgence?