I’ve been running half-a-billion parameter models comfortably in a web browser, especially with WebGPU, and you can definitely run billion parameter LLMs in the browser. It becomes a heavyweight browser app, but if the main costs are running ML models you can pretty easily serve static files from a directory and let clients’ browsers do the heavy lifting. Feel free to reach out if you have questions, happy to help, I’ve been working on language web apps as well
This looked incredibly non-obvious until I looked at the page in a much smaller window and the numbers became very easy to spot. Cool idea. Not how I imagined it from the description.
Thanks for this, it's the morning here and I was squinting way too hard to see where the hell would the analog clock of sorts be. "Maybe it's the rocks position?"
To GP: really cool idea, it'd make for a nice screensaver in an always on display!
>Transit should be free, like sidewalks and parks.
And most similarly, roads. The theoretical 'farebox recovery ratio' of most roads is 0, but this is never part of the conversation. Maybe if it were, transit would look much better in comparison.
NYC has 470 stations, by far the most. Who cares about long-distance stops with nothing in between. NYC has train stations in every single neighborhood in the city.
It's more complicated than that, because some stations with multiple lines get counted multiple times for that. Though even if NYC does have 20 more stations than Shanghai, it's a bit of a stretch to call that "by far". When, again, other cities have twice the tunnel length and nearly twice the ridership.
There are neighborhoods in Queens and Brooklyn where the closest subway station can be miles away, and there is an entire borough in NYC that has no general subway access at all, so not buying that one either.
The difference is that when a sidewalk or park is ruined and un-walkable then it's cheap and easy to just put up some orange cones. So the cost of failure is basically zero.
Not the case with the 4 train, or the L, or the 2-3. Those trains can't fail. So it's very expensive to make sure they don't. Nobody cares if Madison Square Park is blocked off for a day. (Maybe except the vagrants).
But the city will literally grind to a halt if the green or red lines aren't running.
Conversely, there’s nothing wrong with respecting someone’s choice.
I was going to write “No one’s stopping you from using Lenna,” but admittedly if you tried to submit a paper with it, you’d face a stiff headwind. I don’t think that’s her fault though.
I think commenting short summaries like this is not beneficial on HN. It destroys all nuance, squashes depth out of the discussion, and invites people to comment solely on the subtitles of a full article. That's not the kind of discussion I would like HN to degrade into — if I wanted that, I'd go to Buzzfeed. Instead I hope everyone takes the time to read the article, or at least some of the article, before commenting. Short tldrs don't facilitate that.
As much as I agree with this, I must admit that it did trigger me to read the actual article :)
I assume that in a not-so-distant future, we get AI
powered summaries of the target page for free, similar to how Wikipedia shows a preview of the target page when hovering over a link.
While i agree with all the potential downsides you mentioned, i still lean heavily on the side of short summaries being extremely helpful.
This was an interesting title, but having seen the summary and discussion, im not particularñy keen to read it. In fact i would never have commented on this post except to reputation yours.
Looks like lsb was the one who submitted the article, and this comment appears to be submitted at the same time (based on same timestamp and consecutive id in the URL), possibly to encourage people to read the article in case if the title sounded like clickbait.
This isn't “default permit” or “enumerating badness”. Its kind of the opposite.
The idea is that you don't want to store text in your database in a form that is safe when rendered as HTML, JS, JSON, SQL, etc. That would be "enumerating badness". Instead, at the moment you render the text as HTML, you encode the text into an HTML-friendly form (via escape characters). If you want to embed the text into a SQL query, have your SQL library add sql-specific escape characters where needed in the text. Same for your JSON library, and so on.
Its the responsibility of an encoding library to encode and decode text in the appropriate way. A JSON or SQL library should be able to encode then decode any arbitrary unicode string, even one which contains quote characters. Just like how any arbitrary unicode string should be able to be used on a webpage, in a text field without being able to interact with the rest of the page in any way.
Most libraries already do this if used properly. SQL libraries (using parameters) will escape text where needed. React will embed text in an html-safe way. JSON libraries escape quotes in strings. And so on.
‘Quite high’ is still not what you’d think of as ‘quite high’ if you’d never experienced earlier versions of eink though.
LTT on YouTube has some video reviews of Dasung’s displays. I know tastes vary when it comes to LTT - and they focus on some ridiculous things in the reviews (Can you game on them? Obviously, no.) but if you get past that I do think the videos do a a good job of showing how the monitors feel.
