- 1.39K Posts
- 1.8K Comments
Agreed that land and water are smaller compared to agriculture, but the electricity usage change is significant. The LLM boom is unlike prior datacenter workloads. The electrical demand is far higher due to the more power hungry chips and running them at full utilization. It’s projected to go from 5% to 15% of all US electrical demand in quite a short amount of time
This is delaying the closures of fossil fuel plants (here’s an example of 15 coal plants in the US last year), and starting to rely more heavily on generators to install capacity faster despite solar/wind being far cheaper
1·5 days agoI have worked on open source projects. I cannot fork sheer number of projects going towards LLMs alone. This is a losing proposition. Open source is not an individualistic action. This is a collective action, and we need developers of open source to live the values of open source
someone else can pick up from here
A big point of my comment earlier was that making a project increasingly LLM generated makes it harder for someone to pick up as quickly. A huge amount of complexity can be added insanely fast. In this rsync example, the entire testing system was changed overnight (while generating issues in the process). The projects become harder to work on in general
EDIT: also to add, this still has the issues of not knowing where the un-copyleftable code lies and/or having to rework large portions of the project are if you want to keep that
deleted by creator
That framing is missing a lot. Open source software is way more than about using the code. For me, it is not the bugs and quality that concern me most about things like this (though I do have concerns with that too). It’s about the broader issues with LLMs in terms of cooperate power, environmental impact, etc. Calling it out is less about any one project and more about stopping the whole open source ecosystem from spiraling into an LLM-dependent mess. LLMs themselves can easily become the death of FOSS in a broader sense
LLMs flip the power dynamics of development on their head. For starters, the outputs are likely no longer copyrightable in many jurisdictions, which undermines copyleft licenses (rsync is under GPL for example).
The kind of code that LLMs generate also tend to add complexity rather fast where it becomes more and more difficult for any human to understand it. Becoming dependent on LLMs makes development more of a question of computing power rather than effort. Companies will be able to spend more than you. FOSS will not be able to compete nearly as well. It’s also an inherent dependency on big tech companies who will be happy to exploit that the second they can or cut you off it you start to hit their bottom line. Software cannot be free in terms of freedom if modifying it in a reasonable amount of time starts to almost require a tool controlled by someone else
Using “Open Source” (which has somehow become “public weights” to most) / local LLMs are hardly freedom from this either given that they will always be behind given the massive financial costs to make models, unlike traditional software. If you find any advantage or way to reduce resource usage to make a better model, the bigger tech companies will just quickly scale that up far bigger than you can and meet or exceed what you have. It still just as well makes your ability to modify software dependent on the hardware you have. How free is open source software if it becomes increasingly difficult to modify without an expensive GPU?
No other way to review long patches besides LGTM says only profession that keeps doing that
Corn is about evenly split between ethanol and animal feed in the US. That’s also based on USDA numbers that under count the amount going to animal feed by excluding exports and excluding more indirect ways corn go to animal feed
Today’s corn crop is mainly used for biofuels (roughly 40 percent of U.S. corn is used for ethanol) and as animal feed (roughly 36 percent of U.S. corn. […] Only a tiny fraction of the national corn crop is directly used for food for Americans, much of that for high-fructose corn syrup.
https://www.scientificamerican.com/article/time-to-rethink-corn/
Not who are replying to, but I’ve got some good news, moving to plant-based diets is also better for wild animals too
Livestock farmers often claim that their grazing systems “mimic nature”. If so, the mimicry is a crude caricature. A review of evidence from over 100 studies found that when livestock are removed from the land, the abundance and diversity of almost all groups of wild animals increases
And not only that but fewer crops are needed for animal feed so less disruption overall from that end too
If everyone shifted to a plant-based diet we would reduce global land use for agriculture by 75%. This large reduction of agricultural land use would be possible thanks to a reduction in land used for grazing and a smaller need for land to grow crops.
Almost all of the latest commits on the project now :/
https://github.com/RsyncProject/rsync/commits/master/
Not that datacenters are great, but agriculture should not be glossed over like that. The place the water goes for agriculture is not where we want / need the water to go. I.e into the plants moved elsewhere and into the air carried away. It depletes these waterways
Correspondingly, our hydrologic modelling reveals that cattle-feed irrigation is the leading driver of flow depletion in one-third of all western US sub-watersheds; cattle-feed irrigation accounts for an average of 75% of all consumptive use in these 369 sub-watersheds. During drought years (that is, the driest 10% of years), more than one-quarter of all rivers in the western US are depleted by more than 75% during summer months (Fig. 2 and Supplementary Fig. 2) and cattle-feed irrigation is the largest water use in more than half of these heavily depleted rivers
https://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1064&context=wffdocs
“meanwhile it’s eternally captured by a data”
The closed-water loop systems are that concerning water-usage wise at all. They don’t use as much water. It’s the open-loop ones with evaporation tower that use much more water. Those also go into the air and flow somewhere else, same as it is with agriculture
“if not contaminated with chemicals too”
Runoff with pollution from datacenters is more of an issue from construction from my understanding rather than the cooling (not that this isn’t an issue!)
It’s worth noting that agriculture has continuous problems with runoff. Fertilizer and manure runoff is a massive concern from agriculture, often a massive one for local water quality. For instance, one region in NZ needs a 12x reduction in the dairy industry nearby just to meet safe drinking water standards
And that place is not where we want / need the water to go. I.e into the plants moved elsewhere and into the air carried away. It depletes these waterways
Correspondingly, our hydrologic modelling reveals that cattle-feed irrigation is the leading driver of flow depletion in one-third of all western US sub-watersheds; cattle-feed irrigation accounts for an average of 75% of all consumptive use in these 369 sub-watersheds. During drought years (that is, the driest 10% of years), more than one-quarter of all rivers in the western US are depleted by more than 75% during summer months (Fig. 2 and Supplementary Fig. 2) and cattle-feed irrigation is the largest water use in more than half of these heavily depleted rivers
https://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1064&context=wffdocs
It’s around 15% exported in the American west and 4% nation wide. That’s not nothing, but it’s mostly for domestic consumption, the headlines about it are kind of misleading by making it sound like a majority
Alfalfa is the third largest economic product in the US, but only 4% is exported annually. In the western states, however, which are high producers close to shipping ports to major export markets like China, Saudi Arabia and Japan, about 15% is exported each year
https://www.theguardian.com/us-news/2019/mar/25/california-water-drought-scarce-saudi-arabia
Current datacenters are much more concerning environmentally for their electricity usage. The previous 20 years before the current LLM boom, their electrical usage was more or less flat. In the US, it’s now estimated to go from 5% of the US electrical demand to 15% in the next few years and is delaying fossil fuel plant closures
The water usage is concerning in some local situations (often more so from pollution from poor construction) for various data centers, but agriculture and especially animal agriculture really does dominate water usage in water scarce areas and is enormously wasteful with water. For instance, in the American West, it’s mostly all going to animal feed where plants for human consumption use significantly less
This is not to say it’s good that it’s using this water. Just that we really should actually also be very much concerned about the agricultural impact because it’s horribly inefficient. Producing animal products is massively inefficient
10·7 days agoSpeaking broadly: Plenty of other issues or security vulnerabilities can exist that a good sandbox won’t catch. Like software can insecurely store and transmit passwords, have bad randomness for something security sensitive, secretly be mining crypto behind the scenes and burn through battery/electricity, etc.
There are also a number of companies doing culture meat for cats. Cats very much can eat more than just animal products, so using other ingredients for some portion of it would not be an issue
So if i created my own keyboard from scratch on a open hardware microcontroller, could i implement this?
The USB controller that interrupts the CPU lives on the other side of the connection, so you’d just be building hardware that responses to the polling. If you’re curious what that looks like, Ben Eater has a cool video looking at what that looks like for a USB 2 keyboard https://youtu.be/wdgULBpRoXk
There’s also the case of Bluetooth dongle keyboards not working in UEFI (except that one) but USB always do. Is it this or just the UEFI not having drivers?
I am no USB Keyboard expert, but through the power of looking it up it seems like most of these do not operate as a HID (human interface device, like mouse and keyboard) so need driver support, but some start up with a basic HID proxy which might be you have one that works. From an older thread about BIOSes rather than UEIFs
A keyboard using Bluetooth cannot access the BIOS. Logitech Bluetooth keyboards get around this by having a dongle that pairs with the keyboard in a more basic, non-Bluetooth mode until the driver kicks in and switches modes. Microsoft might be similar mode with their keyboards and dongles, but I cannot confirm that.
https://superuser.com/questions/242457/use-a-bluetooth-keyboard-to-access-edit-the-bios
Technically, interrupts are still often involved… just from the USB controller on the state of the polling instead of the keyboard directly on a keypress
Some keyboards implement the USB Boot Keyboard profile specified in the USB Device Class Definition for Human Interface Devices (HID) v1.11 and are explicitly configured to use the boot protocol. These are limited to 6-key rollover (6KRO) and will interrupt the CPU every time the keyboard is polled (even if there is no state change) unless the USB controller is programmed to tell the keyboard to respond with negative acknowledgments, which the USB controller discards in hardware without interrupting the CPU, when there are no state changes to report
https://en.wikipedia.org/wiki/USB_human_interface_device_class#Keyboards
Any economic system is going to face enormous pressure with the mass demand not being addressed, capitalistic or not. The demand side of the equation is going to be changed if you want to avoid an uprising. We’re not talking about some small level of reduction in production
For instance, if you wanted to move to a grass-fed only beef production, you could only supply at most around a quarter of all current beef production while using 100% of grassland (which would create deforestation pressure). This is while simultaneously increasing methane emissions and number of cattle slaughtered. If you want to avoid a methane emission increase, you’d need to go far lower production
We model a nationwide transition [in the US] from grain- to grass-finishing systems using demographics of present-day beef cattle. In order to produce the same quantity of beef as the present-day system, we find that a nationwide shift to exclusively grass-fed beef would require increasing the national cattle herd from 77 to 100 million cattle, an increase of 30%. We also find that the current pastureland grass resource can support only 27% of the current beef supply (27 million cattle), an amount 30% smaller than prior estimates
Taken together, an exclusively grass-fed beef cattle herd would raise the United States’ total methane emissions by approximately 8%.
https://iopscience.iop.org/article/10.1088/1748-9326/aad401/pdf
As long as animal agriculture exists in any scale it’s the invariable outcome. As long as there’s mass demand, there will continue to be mass production
7·1 month agoStarting this off with: I am not an epidemiologist and most of the epidemiologist I’m seeing online aren’t yet too concerned
That being said, they have not found any rodents on the ship, though that does not mean they didn’t just miss them in their search. The version on the ship has been confirmed to be the Andes Virus (ANDV) which is human-to-human transmissible in a way that most hantavirus are not
It’s hard to say exactly how the virus will behave outside of a cruise ship (which are known for spreading diseases more than other locations), but we can potentially look at a past outbreak in 2018 in a small town for an idea
In this work, we described the isolation of the strain responsible for the largest ANDV PTP transmission outbreak, which occurred in the small town of Epuyén and began on November 2, 2018. This strain, ARG-Epuyén, exhibited a high capacity for PTP transmission, necessitating the implementation of quarantine measures to curtail further spread [8]. The median reproductive number (the mean number of secondary cases caused by an infected person) was 2.12 before control measures were implemented and subsequently dropped to below 1.0 by late January
Not sure I get the joke?