> This notion of machine bad, human good just is not realistic
Glad I found this quote. It is quite helpful for an AI to search the web on behaolf of me... even if it was finding where I can buy particular/similar peanuts locally I got from abroad.
Content providers will not agree with this decision, because machine browsing = no ads. Until that gets resolved, I don’t see incentives to align, since any free search requires ads for continuous business.
It could be serving ads if they could persuade the machines to do the purchase.
In fact, even ads ingested by the training data set at this very moment could be useful. Go to Gemini and tell it you want to buy a jacket or whatever and it will recommend some products it ingested from the training data.
This notion isn't just unrealistic, but extremely dangerous. If we accept "machine bad, human good" line of thinking, the only logical conclusion is that we'll have to verify our biometric every time we'd like to access the internet. Like the UK age verification but 100x worse.
As much as I dislike gatekeeping measures like UK's age verification, you can't deny the genuine problem that exists in this case. But it isn't 'machine bad'. There is no good technology or bad technology. It's the intention of those who wield it, that is good or bad. In other words, it's good people vs bad people with technology.
The issue in this particular case is that those content and their web servers are set up for human traffic. In the worst case, a human consumes a few megabytes of data from the server and then leaves. A few of those visits will convert into a job or business opportunity - a fair bargain. LLM scrapers are not like that. They're greedy resource hogs. They not only want everything you have, a whole bunch of them do it repeatedly and endlessly to your server. There's no possible way to justify the cost of such massive bandwidth consumption for a bunch of parasites that never give anything in return. And what do we get? A crappy user experience from all those sites putting up protection measures. This is the tragedy of the commons.
So who is the culprit? The greedy bunch who created the technology that behaves like this and then benefits immensely from it. Are those bad people? Absolutely! Naturally, we need them and their ill intentioned creations off our shared spaces. This isn't anything new. This game has been playing out in different forms since eternity.
And 7kW-10kW water heater right INSIDE of your shower, right?
I was amazed that a socket couldn't be installed for the purpose of LED mirror that is a meter away from shower, but they seem to be fine at running water heater inside shower in UK.
My shower is on the hot water system, not electric.
But for fixed stuff like the light, yes, it meets IP standards. If you meet the right standards you can of course have electrics in splash domes. A standard socket is not going to meet those ingress protection ratings.
A socket 1m away from a shower is likely to have wet hands plugging things in.
If I was the maintainer of a terminal emulator, I would see a quite obvious way to improve the situation for my users: change the default colors so that dark blue is brighter.
There's no obvious way to unilaterally improve the situation across the whole ecosystem, that's true. But I don't understand why individual terminal emulator maintainers don't fix it for their users.
Because it means making choices, breaking assumptions, etc.. They have made it user-customizable so they don't have to go through all that.
FWIW, the current de-facto standard is set by xterm. Here is a relevant excerpt of its source code:
! Disclaimer: there are no standard colors used in terminal emulation.
!
! The choice for color4 and color12 is a tradeoff between contrast, depending
! on whether they are used for text or backgrounds. Note that either color4 or
! color12 would be used for text, while only color4 would be used for a
! background. These are treated specially, since the luminosity of blue is
! only about half that of red/green, and is typically not accounted for in the
! RGB scheme.
!
! Blue text on a black background should be readable.
! Blue backgrounds should not be "too" bright.
!
! Originally color4/color12 were set to the names blue3/blue
!*VT100*color4: blue3
!*VT100*color12: blue
!
! They are from rgb.txt respectively:
! 0 0 205 blue3
! 0 0 255 blue
! However, blue3 is not readable on a black background.
!
! Another choice was from the Debian settings:
!*VT100*color4: DodgerBlue1
!*VT100*color12: SteelBlue1
!
! From rgb.txt:
! 30 144 255 DodgerBlue1
! 99 184 255 SteelBlue1
!
! Some users object to this choice because the background (color4) is brighter
! than they are accustomed. Others point out that the different weights for
! the red/green components make it appear to be not really blue. Finally, it
! provides poor contrast against color13 and color14.
!
! The current choice uses equal weights for red/green (effectively adding a
! gray to the result). It is brighter than the original choice, and provides
! more contrast between color12 and color13, color14 than SteelBlue1 did.
! Contrast of color4 against black is slightly improved over the original.
!
! Some refinement is certainly possible (you are welcome to try) -TD
There are fewer blue cones in the fovea centralis than there are in the surrounding parts of the macula, so humans can't resolve details as well in blue light.
The sensitivity of S cones is also simply much lower than that of the M cones. It's clear that pure (0, 0, 1) blue is perceived as vastly darker than pure (0, 1, 0) green. Blue light must be about 10x brighter (in linear intensity) than green light to be perceived as equally bright; the brightest full-saturation blue in sRGB looks about as bright as the very dark green (0, 0.1, 0). The contrast on black background is incredibly poor.
Which is why people who understand color tend to add a bit of green in to make a color which still looks deep blue but is much brighter than what #00f looks like
Speaking of shortcuts... yesterday I noticed that excel online opens in a localized version.. and with grey letters it tells what shortcut to use to jump to search bar: Alt+Ē
Meh... non-ascii chars as hotkeys maybe work in some languages, but not mine where I have to press silent letter (Apostrophe) to make e -> ē, but it doesn't work within shortcuts: ALT + (' + E)
> 5. Snow is easiest to shovel when it’s just fallen. The more time passes, the more freeze-thaw cycles – even gentle ones – build up and make the fallen snow denser and tougher. (This might be less true in very cold places where it never gets above freezing during the day? I don’t know, honestly.)
If it is very cold and no freeze-thaw cycle, the snow is very... Dry and grainy and still OK for shoveling.
But yes, the puffy stuff just fallen from sky is very nice for shoveling.
As I am someone from EU, please explain me what can you do with this SSN number?
I mean is it like a unique database row id which happens to be a non-changeable-lifetime password which is stored in multiple places in plain-text and you can use it to... "unlock some doors"? Make legally binding agreements remotely... ? Or what?
Or it is PII - privately identifying information which is more of a privacy issue here?
It's used for all sorts of "prove you are who you are" situations. It's most commonly associated with applying for credit/loans, and taxes, but definitely not limited to those things. It's ridiculous that an immutable 8-digit number + name is used for authentication in the USA. It even says on the card "FOR SOCIAL SECURITY AND TAX PURPOSES - NOT FOR IDENTIFICATION" but apparently we've all lost our minds and ignore that. It can be very difficult to go through business processes if you refuse to give your SSN - some healthcare providers will just refuse to serve you.
With it, people can take out loans in your name, get into your accounts, file fake tax returns and get tax refunds in your name, and generally act as if they're you. Things are getting a little better nowadays (with additional information required) but we still don't have a secure method of identification online / over the phone.
Over here we use a PKI cert for that. A smartcard providing the root of that trust is provided by the government after verifying your identity using the typical stuff used for identity documents (any biometric data on file, birth certificate, etc.). That still doesn't mean that it's impossible to steal an identity, or acquire a made up one, but it does make it a whole lot harder.
The thing about social security is that it was supposed to be used for a fairly narrow system, and the physical cards even have text like "not to be used as identification" on them. And then we used it for that anyway
Wouldn't matter. No one is interested in enforcing it, and there is too much value in the datapoint to credit rating agencies to tear the entire edifice down. Hell, back in 2011, I was part of a group migrating away from SSN usage at the Federal level. The biggest delay? Waiting for another semantically compatible I'd to manifest. TIN (taxpayer identification number) could be synthetically combined with a couple other ID's in the dataset such that they could finally retire the SSN's we weren't supposed to be using in the first place.
Basically in the EU, you usually have an ID card (or a passport/driving license/visa card, they're recorded on all of those too) that has a combination of a citizen ID and a document ID. Both of these details are combined considered to be "you" for the purposes of anything to do with the government. The government has a registration of every citizen ID+document ID combination and knows as a result what documents are in circulation. They're technically not required in most of Europe, although you must be able to procure one at request for legal reasons (ie. getting your employment properly sorted, opening a bank account, or a law enforcement official asking for your identity). Revoking a combination is as easy as getting a new ID card/passport since the combination is what counts. ID documents also usually expire eventually, so there's also an inherent time limit to what a leaked combination can cause issues with.
They're also as I understand it, used to handle things like sending everyone voter IDs for elections in advance; this is how the government knows who to send the voting cards to.
Bafflingly, the US does NOT have a national identification method that works like this. There's no country-wide identity document that provides the same assurances. As a result, most US entities (government branches & corporations) have settled on a "closest possible"... which is the social security number. A number that's used to identify every person with attachment to the US in some form since social security is something every US citizen has to interact with. (It also includes a ton of non-citizens since as I understand it, social security is something foreign workers also have to interact with, but that's besides the point.) It's a 9 character long numeric string that identifies you as a person... and has almost no revocation mechanism, even if it ends up in a data breach.
Yet in spite of this, it's still used as a country-wide ID mechanism for a lot of different things and replacing it with a proper ID mechanism has as I understand it (not American) very poor support as it's a culture war issue.
It's often used as a way to verify identity. Historically it's been one of the more secret pieces of information about someone, so while name and birthday are not very secret, if someone wanted to steal an identity, it's generally the SSN that is hardest to figure out. As a result though, I think a lot of places treat it as "If you know the SSN, then you are who you say you are."
As an example, if you call your bank to report a lost credit card, and that you'd like it shipped to a different address than the one you registered with them, they'll ask you for the last 4 digits of your SSN.
So yeah, someone who knows (name, SSN) or especially (name, address, phone, SSN) can do a lot of harm.
I'v seen quite a few Tesla Ys that needed repairs and... they seem to improve the car year to year or even months to months. Car interface suddenly changes to RJ45, some metal parts changed to aluminium (if I'm not mistaken), various things that become easyer to fix and so on. Low Voltage battery getting Li-Ion. Front under body changes: https://service.tesla.com/docs/BodyRepair/Body_Repair_Proced...
And then the airbag controller gets newer and newer.
Not something to market about, but you see steady incremental improvements.
What I want to say, the serviceability is very good for the cars. You get open documentation, you can access toolbox for a price, but it's there for the simple DIYer. Need to change pyro fuse? No problem, pop up docs, order part, change it. The parts are cheap.
Sorry I pressed downvote and cannot revert my press...
I had to set up CF for a small local business in a very small country that has ecommerce presence targeted mainly at local population. It just gets non-stop ongoing traffic a hosting provider cannot handle.
reply