I think the quality will keep improving, because humans will keep curating the training data to compete for best results. The downside is that "quality" is conflated with "performs best for a particular audience and platform" which could just as easily mean re-ingesting junk and spitting it back out... because that's what people respond to.
I see lots of “google is evil” narrative here… in reality they could add a feature to collect spam flags and still disregard user preferences. It’s just that the data will probably not help them, since adversaries are much more motivated to manipulate it for SEO profit than the average user who is unlikely to repeat that search
I was a Pivot - this article rings true. When I first joined people would often describe with joy how exhausted they were every single day for months when they first joined, as though it was evidence of learning and growth. That’s only true sometimes. What I realized is that it doesn’t actually stop being exhausting, people just adapt. There is a difference between being tired because I applied myself, did something meaningful and useful, and learned something new, compared to just being tired from navigating personal interactions all day every day.
Google can't say this directly; forcing people to use your upsell for preferential treatment is likely anti-competitive. Instead, Google ranks on page speed (among many many factors), and then offer a proprietary tool (AMP) that promises to help.
I want judges to be safe. I also want to understand the motivations of this attacker. Their actions weren't justified, but having a long-running court case that can decide the course of your life can cause many new problems. What level of desperation would cause someone to lash out like this? There will always be unreasonable people, but I wonder if this outcome would have been different if courts did indeed process cases in a timely way. We know it is especially cruel to people who have nothing to defend themselves with.
Some time ago I looked for a piano app - something I can use on the go to harmonize with, try some chords, or just play. I wasn't happy with anything that I found, so I made my own:
- It's a progressive web app. Add it to homescreen to avoid multi-touch scroll shenanigans
Here's a concrete example. Some of my spaced repetition is to translate between languages. When I do a translation, if it's different from the canonical answer, I write it in below. Over time I have a list of the different ways I translated, which helps me notice when I repeat mistakes or of what kind.
Anki was my introduction to spaced repetition and started me down a learning journey that I'm very happy to be taking. That said, Anki didn't end up sticking as the way I do spaced repetition - but it took me a long time to be able to articulate why. These days I am using Roam along with roam-toolkit. Having tried it this way, my biggest realization is that facts are memorable when they have many visceral associations. A corollary of this is that, to remember well, I should practice making lots of different connections between the things I know. Now when I do spaced repetition, I annotate things as I go, forming new connections and associations, then putting them in my knowledge graph forever (forever, knock on wood). By contrast, Anki feels like I am over-training flash cards in a very siloed and narrow way, such that the skill I learn is closer to "answering flash cards" as opposed to the actual thing I want to be good at.
There's certainly truth in this notion - if you have a goal in mind, going directly after it is often the most effective and straightforward choice. What I'm advocating is that the most direct path isn't always obvious.
Many masters say the fastest progress can be made in chess by doing "find the checkmate" exercises on tricky or otherwise interesting positions. This aspect of chess disproportionately contributes to success, compared to how much time is spent on each of these positions in the natural course of play. A bit of meta-knowledge about what makes someone successful at chess lets players play more effectively by focusing most on whatever helps them win.
Similarly, if you're learning a language and find yourself constantly groping for the right word, you might make a connection between your ineffectiveness at communication and lack of a memorized vocabulary, then shore it up by focusing specifically on that.
While there's a very strong correlation, there's no special rule of nature dictating that the best way to learn a thing is the same as doing it.
I made https://simplespectrogram.app/ which y’all may also enjoy. With the web audio API I was able to do it in less than 100 LoC - take a look at source
I watched the whole thing. 2 things stood out to me.
- It really illustrated the point made in the article of how labor is really what makes it difficult to accomplish. Weeks of effort sourcing materials, constructing furnaces, bellows, and refining. When it came to extracting ore, they just dug a hole nearby.
- If the Wealden group (the attempt cited in the article https://www.youtube.com/watch?v=zsywnQJMJEk&feature=youtu.be) had watched this documentary, they might have had less furnace cracking and less trouble forming a bloom. I didn't see them add flux for example