It’s probably an assumption baked into the training pipeline from that ancient time six months ago when we mostly were manually following each edit tool call.
Data center construction is consuming all of the tradespeople who usually do work in homes and causing price spikes for contracting similar to what we are seeing for hardware prices. I’m guessing those workers are doing at least OK for themselves ;)
Residential contractors are typically not the same as commercial contractors. Simply for the fact that commercial charges 2x what residential does is one of those reasons. Not saying there is zero overlap, but the pools are not totally shared.
Yea, commercial has far stricter inspection and material requirements then residential. You're playing with a whole lot more angry pixies in a pipe.
Most of the overlap I see is when there is a commercial building collapse (economically, not literally) and you end up seeing master electricians working on residential. Much more rare to see the opposite movement unless it's a permanent change.
> commercial has far stricter inspection and material requirements then residential. You're playing with a whole lot more angry pixies in a pipe.
Not only that, but the trades stick to their trades.
You don't get, for example, datacentre electricians rocking up pretenting they can install network cable to spec and running away having (if you are lucky) only done a continuity test on the cable they just installed.
Meanwhile every man and his dog, or even just the dog, will very happily turn up at your house and tell you they can install your CAT6 to spec for you when they've never seen a Fluke in their entire life.
In a bunch of recent residential building I've seen the electricians "attempt" to install networking cables as it's a value add they can install for very little while the walls are still open.
Needless to say I've re-pulled a lot of those cables due to so many failures.
> You don't get, for example, datacentre electricians rocking up pretenting they can install network cable to spec and running away having (if you are lucky) only done a continuity test on the cable they just installed.
I frequently have my electricians pull and terminate network cable if it’s only one or a few (daisy-chained VAV controllers or a network drop for a BAS panel), if the customer needs a formal TIA-568 test then I’ll sub that out. My electricians are already pulling and terminating building automation cable, CAT 6 isn’t gonna stop em :)
I think some of this comes from everyone "moving up a notch". If the "top" electricians are going to data centers, the second rate are covering their old jobs, and on down until residential electricians are doing light industrial and commercial work (can probably charge more, and open up a new avenue of business).
A residential electrician would be completely lost on a data center site. Commercial electricians travel around to the various data center projects. Most of the trades are similar.
> Data center construction is consuming all of the tradespeople who usually do work in homes
I want to know what country you live in that you can find domestic tradespeople who will do work to the same standard as datacentres, and do it at domestic prices. ;)
Not really because queries are batched in production. Your query alongside a bunch of other peoples’ queries are running the GPU at full speed so yours only uses a tiny fraction of the total energy
Yes if you search anything in Google, you typically get sites that have a cookie panel, ask to allow notifications, show "Login With Google" popover taking the keyboard focus, and lot of ads before you can even read the content. No wonder people do not want to use Google anymore.
It's symbiotic. Google has been aggregating data from these websites for years to serve as summaries in search, reducing traffic to the sites. More pernicious is AMP, where the site is basically copied to the Google server. This force websites to use AdSense and structure the content in certain ways. If a site doesn't want to use AdSense, or allow Google to summarise the site, it gets down-ranked in search. A death sentence. Further, Google's algorithms routinely promote spam because this is most profitable for Google. Hiding the important information 3/4 of the way down a 10,000 word article with a hundred different kinds of ads on the page. Since the "good" websites have to compete with these spam websites, they either adapt or give up monetisation.
One of the more horrifying stories to read was about [Code Yellow back in 2019.](https://finance.yahoo.com/news/google-2019-code-yellow-blurr...) Google's algorithms were getting too good. Users were finding what they wanted too fast. They'd type in their query, the first result was good, then the user would close the tab. For Google, this meant less ad revenue. They wanted users to stay on Google for longer. *So they chose to degrade the algorithm quality.* Users would have to click multiple links to find the information they wanted, and inside every link was ads served by Google. This is one of the pivotal moments in one of the worst cases of enshitification in software history.
I use Kagi Assistant. It doesn't just give a summary, it is more like a personal "Googler"/web surfer. I feel like I'm involved in the process.
My coworkers use Google and often quote back the AI summary me instead of checking the source and referring to that. It's scary seeing even those with higher education behave like this.
The other day I found that one old game of which I originally caught just a glimpse and remembered only a fragment of what one voice actor was saying is in fact Lords of Magic.
I attempted to google this a few years ago to no avail.
> The tradeoff to using this kind of high-security operating system is that it’s less convenient, so few people do it, said David Lie, a professor of computer and electrical engineering at the University of Toronto. But Prof. Lie said there’s no legal prohibition against using a privacy-focused operating system, and in some cases, it may be essential for political dissidents, activists and journalists.
[…]
> Citizen Lab at the University of Toronto called the bill’s data-retention requirements, which would mandate electronic service providers keep detailed data of nearly every Canadian for a year, revealing their movements and who they contacted, “almost certainly unconstitutional.”
> Major tech companies such as Signal, which provides secure messaging, and DuckDuckGo, a privacy-focused search engine, have said they would withdraw from Canada if they were forced to reveal or retain user data that compromises users’ privacy.
The article quotes extensively from the GrapheneOS lawyer, privacy advocates and security researchers. It doesn't owe you an unquestioning hagiography of GrapheneOS.
I was surprised how well-written it was for a newspaper article on something technical. They really took the effort to talk to both sides. I strongly disagree with the anti-privacy side, but getting both takes is what a good journalist should do. (Taking a stance belongs should be in an editorial or opinion piece.)
Stupid question, but why didn’t the DOD mandate IP compatibility for literally all comms like 20+ years ago? Is it pure corruption/incompetence or is there a good reason?
A lot of military comms are point-to-point or broadcast between radios where TCP/IP doesn’t really make sense. On top of that, you need to understand how much legacy military networking stuff is out there. This stuff long predates the internet. We had tactical data links and target tracking where a ground station could receive a target info from a remote radar array, and automatically direct fighters on autopilot on an intercept course with enemy contacts in like 1960 (see SAGE as an example). These systems have been around for decades and been continually improved on. There’s a lot of inertia to maintain compatibility. Because you really don’t want someone to die because your ally tried to warn you but their tactical data link isn’t compatible with yours.
The DOD effectively can't "just mandate" things and it's because of politics (which is a superset of corruption and incompetence).
There's a very small number of major contractors and if you pick one of their APIs as the new standard, then you disadvantage all the others.
If you pick a greenfield project, then it will take a long time, and once you are done everybody will claim that they can't use it (see e.g. Ada; quick version they started working on a mandate for a programming language in 1975, mandated Ada in 1991 and by 1996 removed the Ada mandate).
lolwut? This is Hacker News of all places do people not realize how much memory, and more importantly bandwidth, these systems need to work? The idea of a distributed botnet of AI using the compute of its victims to continue its inference is pure science fiction given how LLMs actually work.
> The idea of a distributed botnet of AI using the compute of its victims to continue its inference is pure science fiction given how LLMs actually work.
An attack like this doesn't really need the exploited computers to run inference, do they? If I was an LLM bent on destruction of the internet, I'd be writing programs to run on each computer, not turning each computer into an LLM itself.
A few programs to break in, install themselves and remain asleep until they are needed, another few to spread through grabbing every OpenAI, GLM, whatever key, another one to remain asleep on computers (whether hosted or desktops) that have adequate GPU, etc.
The GP was referring to the AI 'living off the land' so to speak by using its victims compute to avoid being shut down which is clearly laughable.
More to the point, so many people in this thread are making completely contrived and outlandish stories up about how AI might try go ruin our lives without any evidence backing them up in any way. It is hysterical. This is the most important technology in our lifetimes and people want to freak out and turn it into the next nuclear power, with progress banned in all but name.
It is really mysterious how awestruck folks are at the HF attack. I mean just watch one modern agent (qwen 3.8, deepseek 4, glm 5.3) rip apart a coding problem and nearly destroy your computer in the process, it's a wonder it took "months" and "millions" in the first place.
OpenAI has too much money, a common post-growth-stage issue that leads to pursuing a million stupid things with no clear plan. Like running a bunch of agents for months without any idea how to keep track of progress.
I agree with the idea that it's not probable but I do think it's important to point out-- we only need these elements to run with the bandwidth they have and the token rate because we want to see things in human-scale time. But slow things down to a 1tok/sec doesn't matter to this hypothetical anti-aligned LLM. Time is, after all, relative, and it's not like LLMs give a shit how long something takes. They don't have squishy stupid organs that fail after a certain amount of time, or those pesky glands that emit impatience hormones.
But yeah you're not going to be able to shard out the terabytes of Fable weights that are needed to run inference without addressing some fundamental physics problems.
There is an automatic declassification system with a 25 year clock but it’s a bit more restrictive than what you propose. That said we have a cornucopia of classified info from the Cold War era available today and that has had significant academic value.
reply