numpad0

And it's not a tiny kernel of truth, even iPhone(2007) had Wi-Fi/BT combo chips with FM radio. Apple was partly responsible for death of FM radio, by not implementing the software feature. That's also literally 20 years ago.

Jack Dorsey (@jack) said in an announcement on X

...

That used to be the case, but dudes through a voice changer are quite rare now.

Maybe it's LLM and human actor splitting work, but she looks too natural, too fast and too error-free to be LLM based since too long ago. Just their dictation -> thinking -> TTS loop is too fast to be LLM. Context awareness and character adherence is too perfect as well.

Does anyone actually believe the premise of Neuro-sama that she is an AI? The original premise was that she is a lightweight AI model livecoded on streams, but that is obviously not the case anymore.

The VTuber phenomenon is often framed as escape hatch for streamers who do not prefer to show faces; IMO, it couldn't have worked if audiences found it disappointing - the audiences must prefer the presenter hiding their face than prominently displaying it. I think that's more noteworthy than the person behind it opting to hide their faces.

That's supposed to make skin appear lively. Human skins in AI images tend to look clouded, opaque, and overall un-alive, so to speak.

known-good prompt-response pairs are more useful than random semi-coherent texts presumably

The whole thing feels like that the US collectively isn't feeling confident about itself in the coming creativity-limited world that AI/AGIs supposedly bring, and that it's not happy about that pessimistic rational projections. I don't know exactly why wouldn't it be or why nothing seem to be done about it, though. Both look to be fundamentally solvable problems to me.

... is this a reverse Dropbox comment? Everyone up to COVID was treating companies not moving some infra to AWS like how hardware companies would be treated without a mockup humanoid to show around.

It feels to me more like, East Asian nations tend to have robust domestic sets of expectations that tend to be more stringent, reinforced by geological and cultural barriers, and comparatively lower international castes. Anyone born in EA have no choice but to constantly beat the average and to make contributions to the street cred jar for the nation. And that produce such things as TSMC chips and DJI drones and 7-11 egg salad sandwiches.

Ubuntu snap

People say they think it's faster than top of the line RTX 3070[1] and hands down the fastest GPU ever for local LLM purposes, and OTOH they're like Intel integrated graphics with RAM shared VRAM for ultra thin laptops, so the truth is either of that, or somewhere between those two. Probably wherever its TDP suggests.

1: (Laptop)

That's not the elephant in the room. The elephant in the room for gaming on Mac is Apple's policy non-commitment to traditional gaming. Technical support for non-gambling gaming is just a checkbox item to them for mandatory 45-second segments in product launches. That transcends into their customers and develop into various gaming and Mac related chicken and egg problems.

It seems absurd to suggest that the internet isn't global

Internet from L1 to L4 is global but WWW at L5 and above was always sort of fragmented. Look at chat apps: Messenger, WhatsApp, WeChat, LINE, KakaoTalk, Telegram, etc. The fault lines for userbases of these apps roughly align regional borders.

DisplayLink creates virtual displays on CPU and stream compressed feed over USB. That takes non-negligible amount of CPU time and RAM bandwidth, and it makes sense to give it a max concurrent stream count(set aside which of 4 or 6 are more reasonable).

I thought this discussion is more about somehow aligning pipes over multiple lines in existing console emulators(impossible), than about implementing a complete custom graphical text rendering system specific to your app that butcher font files to put glyphs wherever you want?

That feels like cheating since you're not rendering provided font at that point. Besides you might as well just use SVG for diagrams than pretending to be text only.

ok... I disagree, and I mean no offense, but there's going to be too much contexts and nuances to be taken into account that it's probably not worth trying for both of us. All I can say is that the problem is not that the pipe characters are given inconsistent width but that non-ASCII characters has all different random widths, and if you need texts with double-width characters like ▶, 漢, ก, etc., to align in a grid, you have to pick a fixed-width(not "monospace") font for the specific language and exclusively use that font for everything within that contiguous text area. Or you can try to fix Unicode so that symbols become variable width so to align to grid or something, but that's going to take a lot of effort.

This isn't a renderer bug, it's font. Proportional fonts aren't designed with alignment in mind at all, and you can't just expect monospace fonts for all languages outside of ASCII range to be present on random systems, or a single font or font family to support multiple different languages consistently.

You can't really control alignment of deeply Unicode characters like Thai or "→" against monospace characters without serving your own monospace fonts that are guaranteed to work for the characters you'll be sending out, assuming you can always have one in hand.

Yup, 2ch.net/5ch.net/5ch.io uses the MS P Gothic and there were app features and fonts like Mona Font to emulate that. They were not imageboards, though.

I just thought that MS Gothic(non-P) should be kind of widely supported, have all the symbols you need, while also being a monospace, unlike most monospace fonts that only support ASCII symbols.

Aren't those non-ASCII "rich" symbols from Japanese fonts around PC-98/Win95 domains anyway? For me with my background, it was always obvious that mixing full-width character in ASCII text never go well for various reasons. For ASCII arts, it is obvious that vertical lines never line up, and there are going to be tons of wasted spaces and different kinds of whitespaces needed to compensate for those. I wonder if specifying MS Gothic and retuning widths for it could help, at least for Windows/Linux.

But a lot of car "hacking" is like installing RGB fans back to back so they whine louder. Which is totally a made up strawman argument that nobody is making, but also not hacking.

It feels odd that extremely few in Anglosphere is pointing that out. Tesla chassis design and handling tuning is apparently atrocious enough that many total amateur dads on car review sites have to mention it, in the Japanese sphere.

Theoretical max speed of 4G/5G are like 5Gbps down/1Gbps up, without going into mmWave. Actual link speeds of Starlink terminals don't seem to be published in the open, but it's at least throttled to up to 310Mbps down/44Mbps up for Priority plans. Looks like it's supposed to reach 1Gbps down with V3 sats, but then again, cellular is like up to 5Gbps yesterday, at least in the spec.

It's just that Starlink being a new thing and with zero users tended to be less congested, hence it tended to do better in somewhat of an unfair comparison in a remote cabin side by side against an LTE equipment. It was NEVER faster in theory compared to terrestrial cellular.

$40B is also less than a single year's worth of defense budget, of Japan, so... it's just about what's expected of a vaguely forward looking government funded program. It's not really a giant high risk make-or-break bet.

Yeah. I don't see the problem with Chinese prompt stealing proxies, if it's just pure free choice and discount for explicitly insecure use cases, especially when the frontier providers they route to are soft-assumed to be doing something similar.

I think the point GP's trying to make is that Japanese market require everything to be specifically designed for it, despite that the opposite rarely being true. One related concept I've seen mentioned is "Japan takes executive summaries of things and follow its own (weird/bad) interpretation of it".

BYD so far is doing much worse in Japan compared to Tesla, despite them having standard lineup of an SUV, two sedans, and a compact. A part of the reason would be that there's no particular motivation for their customers to pick specifically a Chinese car, but it's also true that a lot of Kei cars are sold in Japan, and producing a ground-up Japan-specific product was the solution that BYD took, which, I suppose, seemed to be a bizarre decision to GP.

Hydrogen stations basically don't exist in Japan either. There are less than 100 total throughout Japan and whopping ten in Tokyo. California probably have more per capita of those.

I don't understand how "Japan is betting on Hydrogen" meme survived this long. Toyota sold something like up to 20k total of passenger FCEV in past 10-20 years. That's less than a quarter worth in Prius sales numbers. It always has been a joke and a futurism porn.

Is it coming from some internal Slack channel at Tesla or something? Whoever spreading it don't know what they're talking about.

Maybe they just don't work? Otherwise someone's leaving tons of money on the table. Which implies nobody is.