barsoap

joined 1 year ago
[–] barsoap@lemm.ee 0 points 6 months ago* (last edited 6 months ago) (6 children)

Propaganda isn’t always fake news, it can also be true stuff presented in a biased way.

It can also be true stuff presented in an unbiased way. There's a disconnect here between the proper definition of the word, which is perfectly neutral, and its connotations because the what secretary for tsunami safety doesn't call their stuff "propaganda" but "public service announcement". Still the same thing, though, the tsunami safety secretary is trying to persuade the audience to not be stupid and get to high ground as soon as the sea recedes. Very much pushing an agenda, they *gasp* want people to survive and *gasp* use communication to achieve it.

[–] barsoap@lemm.ee 1 points 6 months ago

Great, congratulations. Now figure out how to use the tech without increasing per TB costs and find financing for that with large enough ROI so that the capital that be cares. Seagate already ships 24G HDDs at 20 Euro/TB, whether you want to call a step up to 30 "drastic" I'll leave up to you.

Meanwhile, current NAND chips aren't using EUV. The industry can piggyback off the investment and experience of logic chip manufacturers, HDD manufacturers can't. Synergy and all that.

[–] barsoap@lemm.ee 14 points 6 months ago* (last edited 6 months ago) (3 children)

HDDs won't go away any time soon and that's also exactly what he says but the industry is nowhere close to moving as much product as it did in the past, and the number of companies has been reduced to pretty much three. Technology also isn't going to improve drastically any more while NAND storage only needs that much of a price reduction to match HDDs, currently about 15 vs. 50 Euro/TB. Not spitting distance but it's getting darn close.

It says, after all, "boom and bust" not "rise and fall".

Also, no, tapes are more cost-effective. An LTO-8 tape is about 80 Euros, that's less than seven Euros per TB (uncompressed).

[–] barsoap@lemm.ee 1 points 6 months ago

There's exceptions for therapeutic uses (dental, nicotine gum) and you can also import small amounts for personal use. Chewing is perfectly legal, improper disposal isn't.

[–] barsoap@lemm.ee 1 points 6 months ago* (last edited 6 months ago)

The build more thing is in urban areas to accommodate urbanisation. Coming to think of it the 49 Euro ticket might actually reverse some of that because there's tons of smaller towns, sometimes villages, with proper train connection to the next large city. Low prices drive usage which prompts higher train frequencies which, infrastructure permitting, takes even more pressure off the metropolitan housing market.

That said urbanisation isn't in itself a bad thing -- it makes a lot of sense for a lot of reasons to not have a gazillion tiny villages, it'd just be another form of sprawl. Here in SH, if every train station we do have had a small urban core around it surrounded by village structures in cargo bike distance and then long stretches of fields and nothing, that'd actually be quite nice. The state was very good at doing that in the Hamburg metropolitan area, focussing development on a couple of axes radiating out from Hamburg but it should become more of a general pattern.

[–] barsoap@lemm.ee 2 points 6 months ago (2 children)

Governments should simply regulate more so that people vacationing will go to hotels and houses will be available for residents

Berlin did exactly that: You can rent out your apartment for IIRC 30 days a year, or while you're also living there, if you want to rent out more you need a hotel license and tough luck getting one while there's a housing shortage, least of all for a flat in a residential area.

But OTOH that's all only taking the edge off there's been decades of under-investment in social housing in Germany overall, and the little social housing that got built got built via attaching conditions to building permits for private investors, those apartments lose their social housing status after 20 or such years.

And it's not like there aren't companies who want to build, and build plenty -- but they don't because they can't recoup costs, not in this market where every rich fuck who can afford rent already is living somewhere else: Building costs are too high. Some of that is building standards, permitting, etc, but the bulk of it is financing costs, that is, interest on loans for new construction is way too high. Getting at land is not always easy but there's plenty of mechanisms such as municipalities having right of first refusal for any land sale (if they want to, that's another topic). There's really only one way out of this and it's state coffers because the capital market certainly isn't going to get less greedy.

[–] barsoap@lemm.ee 6 points 6 months ago (2 children)

Singapore actually has housing pretty much nailed down. If Singapore had US-style housing policies 95% of Singaporeans would live in Malaysia and commute because only bank execs etc. could afford living in the city.

...also Singapore would have zero green space left. It'd all be single family homes interspersed by parking lots. You can't spit out a chewing gum over there without hitting a public park.

[–] barsoap@lemm.ee 15 points 6 months ago (1 children)

3d assets in videogames become cheaper and quicker to make

You mean there's going to be a brand-new way for people to completely fail optimising game assets.

[–] barsoap@lemm.ee 35 points 6 months ago

There are lots of ways to improve LLMs that aren’t just increasing the parameter size

The paper isn't about parameter size but the need for exponentially more training data to get a mere linear increase in output performance.

[–] barsoap@lemm.ee 7 points 6 months ago

...and a baby doesn't use the same architecture, not even close, as generative AIs. Babies are T3 systems, they aren't simply systems which have rules on how to learn, they are systems which have rules on how to develop learning strategies that they then use to learn.

I'm not doubting, in the slightest, that AI can't get there: It's definitely possible. It's just not possible with the current approaches, and the iterative refinements that "oh OpenAI is constantly coming up with new topologies" refers to is just more of the same. Show me a topology that can come up with topologies, then we'll have a chance to break through the need for exponential amounts of data.

[–] barsoap@lemm.ee 7 points 6 months ago (2 children)

...and aren't making progress on that front: A linear increase in generalisation still requires a more than linear increase in amount of data.

Also it's not btw that we wouldn't know that our current architectures won't lead to proper intelligence, tl:dr: While current architectures can learn, and represent information, they cannot develop learning strategies or decide smartly on how to represent a particular bit of information. All the improvement that are happening are on that "how to learn better" area, we have no idea whatsoever how to make the jump on how to teach an AI to learn how to learn. AlphaZero is able to learn rules of a game, yes, but it can't learn arbitrary information -- once you throw something other than a game at it it has no idea how to make sense of anything.

[–] barsoap@lemm.ee 2 points 6 months ago* (last edited 6 months ago)

Part of the SoC makes a lot of sense but I'd still like to have an expansion option. Or, well, actually, maybe connecting it up via PCIe might be sufficient, latency is going to take a serious hit but if there's gigabytes of HBM on the chip acting essentially as cache it's probably fine for pretty much all practical workloads. Gigantic memory requirements don't tend to come with purely random access patterns.

OTOH that definitely puts the "N" in "NUMA". I doubt any OS but Linux could deal with the thing sanely.

view more: ‹ prev next ›