Quartz sync: Mar 18, 2024, 11:05 AM

content/Essays/ai-infringement-drafts.md

@@ -10,12 +10,16 @@ Artist's will, don't exploit
 ### Detour: plagiarism
 There's also the problem of correctly sourcing information used in forming an opinion.
 
-One proposed "solution" to AI use of copyrighted works is interestingly to attribute that those works were used in the first place.
+One proposed "solution" to AI use of copyrighted works is interestingly to cite those works used in generating an answer. But I actually think an anti-plagiarism argument that I disagree with regarding human work finds footing here. I talked about the *reductio ad absurdum* point in [[Essays/plagiarism#Response to Frye|🅿️ my response to Frye on plagiarism]]...
 ## Economics
 WIP
+
+### What these incentives teach us
+At the end of the day, these policy arguments are here to suggest what direction the law should move in. To solve the economic "half" of the AI problem, what about a different kind of commercial right? Something more trademark than copyright. ==use of expression; remedies too==
 ## The enforcement problem
 WIP
 ## Building universal truth
 WIP
-## Why is piracy ethical, but not AI training?
+## Ethics
+Why is piracy ethical, but not AI training?
 WIP

content/Essays/ai-infringement.md

@@ -52,10 +52,11 @@ One common legal argument against training as infringement is that the AI extrac
 
 <img src="/Attachments/common_crap.svg" alt="Common Crawl logo edited to say 'common crap' instead" style="padding:0% 5%">
 
-Everything AI starts with a dataset. And most AI models will start with the easiest, most freely available resource: the internet. Hundreds of different scrapers exist with the goal of collecting as much of the internet as possible to train modern AI (or previously, machine learners, neural networks, or even just classifiers/cluster models). I think that acquiring data without authorization to train an AI on is copyright infringement standing by itself.
+Everything AI starts with a dataset. And most AI models will start with the easiest, most freely available resource: the internet. Hundreds of different scrapers exist with the goal of collecting as much of the internet as possible to train modern AI (or previously, machine learners, neural networks, or even just classifiers/cluster models). I think that just acquiring data without authorization to train an AI on it is copyright infringement standing by itself.
 
 Acquiring data for training is an unethical mess. **In human terms**, scrapers like Common Crawl will take what they want, without asking (unless you know the magic word to make it go away, or just [[Projects/Obsidian/digital-garden#Block the bot traffic!|block it from the get-go]]), and without providing immediately useful services in return like a search engine. For more information on the ethics of AI datasets, read my take on [[Essays/plagiarism#AI shouldn't disregard the need for attribution|🅿️ the need for AI attribution]], and have a look at the work of [Dr. Damien Williams](https://scholar.google.com/citations?user=riv547sAAAAJ&hl=en) ([Mastodon](https://ourislandgeorgia.net/@Wolven)).
--  Sidebar: and acquiring this data is copyright infringement too, as unlicensed copying. The case is tremendously stupid: [*MAI Systems v. Peak Computer*](https://casetext.com/case/mai-systems-corp-v-peak-computer-inc) holds that RAM copying (ie, moving a file from somewhere to a computer's memory) is an unlicensed copy. As of today, it's still good law, for some reason. Note that every single file you open in Word or a PDF reader; or any webpage in your browser, is moved to your memory before it gets displayed on the screen. Bring it up at trivia night: just using your computer is copyright infringement! It's silly and needs to be overruled going forward, but it is good law.
+
+The reason that it's copyright infringement? [*MAI Systems v. Peak Computer*](https://casetext.com/case/mai-systems-corp-v-peak-computer-inc). It holds that RAM copying (ie, moving a file from somewhere to a computer's memory) is an unlicensed copy. As of today, it's still good law, for some reason. Every single file you open in Word or a PDF reader; or any webpage in your browser, is moved to your memory before it gets displayed on the screen. Bring it up at trivia night: just using your computer is copyright infringement! It's silly and needs to be overruled going forward, but it's what we have right now. And it means that a bot drinking from the firehose is committing infringement on a massive scale.
 
 But then a company actually has to train an AI on that data. What copyright issues does that entail? First, let's talk about The Chinese Room.
 
@@ -81,12 +82,19 @@ The idea and expression being indistinguishable by AI may make one immediately t
 ### Generation
 WIP
 #### Detour: actual harm caused by specific uses of AI models
-My bet for a strong factor when courts start applying fair use tests to AI output is harm, in that the AI use in the instant case causes or does not cause harm (and I actually wrote this before the [[Essays/no-ai-fraud-act|No AI FRAUD Act]] 's negligible-harm provision was published. -ed.). Here's a quick list of uses that probably do cause harm.
-- Election fraud, including even **more** corporate influence on US elections ([not hypothetical](https://www.washingtonpost.com/elections/2024/01/18/ai-tech-biden/) [in the slightest](https://openai.com/careers/elections-program-manager), [and knowingly unethical](https://www.npr.org/2024/01/19/1225573883/politicians-lobbyists-are-banned-from-using-chatgpt-for-official-campaign-busine))
+My bet for a strong factor when courts start applying fair use tests to AI output is harm, in that the AI use in the instant case causes or does not cause harm { *and I actually wrote this before the [[Essays/no-ai-fraud-act|No AI FRAUD Act]] 's negligible-harm provision was published. -ed.* }. Here's a quick list of uses that probably do cause harm.
+- Election fraud and misleading voters, including even **more** corporate influence on US elections ([not hypothetical](https://www.washingtonpost.com/elections/2024/01/18/ai-tech-biden/) [in the slightest](https://openai.com/careers/elections-program-manager), [and knowingly unethical](https://www.npr.org/2024/01/19/1225573883/politicians-lobbyists-are-banned-from-using-chatgpt-for-official-campaign-busine))
+	- [Claiming](https://www.washingtonpost.com/politics/2024/03/13/trump-video-ai-truth-social/) misleading voters? 
 - Other fraud, like telemarketing/robocalls, phishing, etc
 - Competition with actual artists and authors (I am VERY excited to see where trademark law evolves around trademarking one's art or literary style. Currently, the arguments are weak and listed in the mini-argument section). 
 - Obsoletes human online workforces in tech support, translation, etc
 - [[plagiarism##1 Revealing what's behind the curtain|🅿️ Reinforces systemic bias]]
+#### Detour 2: An Alternative Argument
+There's a much more concise argument that generative AI output infringes on its training dataset. I don't plan to engage with it much because I can only see it being used to sue a *user* of a generative AI model, not the corporation that created it. Basically, generative AI output taken right from the model (straight from the horse's mouth) is [not copyrightable according to USCO](https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence). If the model's input is copyrighted, and the output can't be copyrighted, then there's nothing in the AI "black box" that adds to the final product, so it's literally *just* the training data reproduced in another format. Et voila, infringement.
+
+This argument isn't to say that anything uncopyrightable will infringe something else, but it does mean that your likelihood of prevailing on a fair use defense should be minimal.
+
+Note that there are many conclusions in the USCO guidance, so you should definitely read the whole thing if you're looking for a complete understanding of the (very scarce) actual legal coverage of AI issues so far.
 ### Where do we go from here?
 Well, getting to evaluation of the above by courts would be a start. Right now, courts are ducking AI issues left and right on standing and pleading grounds. Once there's more solid (or honestly *any*) coverage of the legal arguments on the merits, the reasons why the law should be enforced that way as a matter of policy will become more important.
 # Policy

content/Essays/on-linux.md

@@ -7,6 +7,7 @@ tags:
   - seedling
   - essay
 date: 2023-08-23
+lastmod: 2024-03-17
 ---
 > [!hint]  This page documents my many adventures with Linux and why I enjoy it.
 > If you're looking to get involved with Linux, feel free to browse the [[Resources/learning-linux|resources for that purpose]] that I've compiled.
@@ -26,7 +27,7 @@ It did need some special setup to run RetroArch, so I created a script and left
 
 The response I got was amazing! Everyone in the organization was extremely grateful, and I'm so happy I undertook that project.
 
-Unfortunately, the cabinet was scrapped earlier this year due to space requirements and a shifting purpose for the room, but it did end up being used actively for a few years, so I don't regret the project at all.
+Unfortunately, the cabinet was scrapped in mid-2023 due to space requirements and a shifting purpose for the room, but it did end up being used actively for a few years, so I don't regret the project at all.
 ## Bare Metal
 This has been my favorite part of my journey. Unrestrained, Linux is...surprisingly good, actually. 
 
@@ -46,4 +47,29 @@ Having to deal with subpar systems after my taste of how convenient Linux made t
 $\downarrow$ **Here's how that's going**: { WIP }
 ### Kernel
 For the love of god, don't ever use the default kernel when daily driving. A custom kernel will squeeze every ounce of performance out of your hardware the way Windows would. I recommend the [CachyOS Kernel](https://github.com/CachyOS/linux-cachyos).
-- Fedora has a copr, and it's available on basically every Arch distro. Sorry debian/ubuntu users.
+- Fedora has a copr, and it's available on basically every Arch distro. Sorry debian/ubuntu users.
+### Animosity towards Windows
+I've had to retain a Windows dual-boot in order to use the exam software required by my law school. I could probably virtualize it since its protections are laughable (it won't boot if you have a second monitor or have a vmware workstation *host* installed, but the fact that Windows is running as a Hyper-V *guest* isn't an issue), but I don't want to risk being falsely flagged as academically dishonest until after my final 3L exams when it literally doesn't matter. Regardless, every time I boot it back up I'm presented with many of the reasons I swapped to Linux in the first place.
+
+First, the startup. Windows has quite a few non-privacy, non-furtive idiosyncrasies, but by far the most infuriating is how the system hitches for 3-4 minutes during and after login from a shutdown or reboot.
+### Jumping Ship to Arch
+I started playing around with Arch on my 1TB expansion card when Fedora announced they were considering dropping X11 a few months ago. Interestingly, I ended up wanting to use Wayland with Arch anyway.
+#### Digital Extremes and Wonky Macros (DEs/WMs)
+I first tried Hyprland with a random sensible config I found on YouTube, and once I stripped out Kitty for Alacritty I quite liked it. The only issue was that toolbars on things like Firefox and Dolphin take up way too much screen real estate.
+
+Then, I added gnome and the gnome apps, was fun to try the newest gnome and see how well integrated with Wayland it was.
+
+And finally, I booted this back up once Plasma 6 dropped. Honestly, it's the first Plasma desktop that's actually looked good to me, so this will probably be what I swap to. Wayland was also great but it was less fault-tolerant than GNOME: I had to **enable kernel mode setting to get Plasma to work with Wayland NVIDIA multi-monitor.** 1.75x scaling on the Framework internal monitor and 1x on the 1080p worked like a charm.
+
+I may also try [Niri](https://github.com/YaLTeR/niri) and [Karousel](https://github.com/peterfajdiga/karousel) soon if I upgrade to an ultrawide monitor.
+#### Other Fun Times
+Having an installed OS that you can throw anything on without regard to breakage has been great for toying with whatever catches my fancy. This is actually where I experimented with (wip) [[Projects/vfio-pci|GPU passthrough to a Windows VM]].
+
+I've also been doing some Rust toolchain witchery on here but I'm not ready to write about it yet.
+#### Progress
+I've figured out what I want for my eventual install, just haven't done anything yet.
+- Desktop: Wayland
+- Greeter: SDDM
+- DE: Plasma 6
+- Theme: Graphite
+- Filesystem: BTRFS

content/Essays/plagiarism.md

@@ -55,15 +55,34 @@ Second and perhaps most importantly, because of the actual issue of AI bias, tra
 - Countless actual examples exist, too many to list. I documented one incident [here](https://social.treehouse.systems/@be_far/111990173625090669).
 ### #2: \[citation needed\] for responses to prompts
 Not to be confused with Molly White's [excellent newsletter](https://citationneeded.news/). This requirement is a more fine-grained mitigation for the transparency issues present in the dataset at large. It also provides evidence for potential copyright infringement lawsuits if the AI has also copied the expression of the paper it sourced. Note that this isn't the be-all, end-all solution to the problem of copyright infringement by AI. Read more of my take on that [[Essays/ai-infringement|🤖 here]].
-## To-be-written
-I want to address piece-by-piece [an argument by Brian Frye](https://www.techdirt.com/2024/01/09/plagiarism-is-fine/) supporting plagiarism in general. He's a prolific IP scholar, so I'll probably look through his academic works as well (*Against Creativity*, 11 N.Y.U. J.L. & Liberty 426 (2017), looks pretty interesting). To be clear, I don't want to get into the absolute witch hunt that inspired the linked article, but in the article he reiterates his greater conclusions about attribution to say that ALL plagiarism accusations are silly, which are what I want to respond to.
-- Planned topics: granularity, necessity, nature of the work/merit, nature of the work/type of content.
 
-I also want to discuss disrespect of creators’ intent for their works and what to label that practice. This applies both to my AI essay and plagiarism talking points. 
+---
+
+## The Anti-Plagiarism Argument: Response to Frye
+Above, I outlined some specific examples that I come across in my daily life as a contributor, digital gardener, and academic writer. But now, I'd like to address piece-by-piece [an argument by Brian Frye](https://www.techdirt.com/2024/01/09/plagiarism-is-fine/) supporting plagiarism in general. I've also structured my complete case that supports good-faith deterrence of plagiarism as a social and academic norm. 
+- Sidebar: The specific event that sparked Frye's article was conducted in anything but good faith.
+
+There are many instances where enforcing an author to either use original content or state whose content they are using can be valuable, many of which demonstrate the key values underlying anti-plagiarism sentiment. The article makes several claims about plagiarism, and I have a different interpretation of quite a few of them. Since it's a summary piece, I also looked at Frye's considerable scholarly work on plagiarism to get a better understanding of the points made.
+### Granularity
+
+### Necessity\[?\]
+
+He's a prolific IP scholar, so I'll probably look through his academic works as well (*Against Creativity*, 11 N.Y.U. J.L. & Liberty 426 (2017), looks pretty interesting). To be clear, I don't want to get into the absolute witch hunt that inspired the linked article, but in the article he reiterates his greater conclusions about attribution to say that ALL plagiarism accusations are silly, which are what I want to respond to.
+### Originality
+
+### Different strokes for different folks
+#### Merit
+
+#### Content-type:
+
+## To-be-written
+I want to discuss disrespect of creators’ intent for their works and what to label that practice. This applies both to my AI essay and plagiarism talking points. 
 
 Anyone who identifies as a "proud plagiarist," this is your notice that I may respond to your opinions, and I will properly attribute you when doing so.
 - Readers: **Don't harass anyone I cite, please**. We disagree on the topic, and since all it really bears on is respect and authoritative nature until it goes into copyright infringement territory, there aren't any high stakes.
 
 I'm dying to dig into enterprise software engineering and attribution/licensing as well. "The StackOverflow problem" is something that the industry has been struggling with for years, and there are some pretty strong counter arguments to my position that come out of critique of softeng and originality. Given the existence of Copilot (and StackOverflow's ai stance), this ties into AI as well.
-## Further Reading
-This paper spells out what we should be thinking about relative to information authority, trust, and societal need when talking about generative AI. **Sections 4 and 5 are very good**; section 6 jumps the shark by immediately forgetting that it's about modern generative AI and ranting about historical Google bugs instead (which the paper would actually classify as a discriminative IA system, good under its arguments). [Bender & Shah (unpublished)](https://faculty.washington.edu/ebender/papers/Envisioning_IAS_preprint.pdf)
+## Further Resources
+This paper spells out what we should be thinking about relative to information authority, trust, and societal need when talking about generative AI. **Sections 4 and 5 are very good**; section 6 jumps the shark by immediately forgetting that it's about modern generative AI and ranting about historical Google bugs instead (which the paper would actually classify as a discriminative IA system, good under its arguments). [Bender & Shah (unpublished)](https://faculty.washington.edu/ebender/papers/Envisioning_IAS_preprint.pdf)
+
+This feels a lot more callout-y and like a public shaming of those who have plagiarized, but massively popular video essayist hbomberguy has a [piece on YouTube content plagiarism](https://www.youtube.com/watch?v=yDp3cB5fHXQ). I think that it's an especially dangerous area to be shouting "go plagiarists go" in, because of the effects. People's entire livelihoods are at stake on YouTube; someone else should not be able to make considerably more money than the original uploader for a reupload just by virtue of an algorithm that cannot be understood. This is not entirely a plagiarism problem (it's a platform and platform inertia problem too, but I haven't written my essay on those yet...sneak peek: [[Essays/content-death|Content Death]]), but this kind of unfair competition is a very significant side effect. See also [ProZD's experience](https://www.youtube.com/watch?v=b9iw6UUMOuw) and [follow-up](https://www.youtube.com/watch?v=Fel4WTp7cTc).

content/Projects/my-computer.md

@@ -43,7 +43,7 @@ I use a [bare git repository](https://www.atlassian.com/git/tutorials/dotfiles)
 #### Config Hell
 - There are a lot of little tweaks I do to software to make it fully useful to me, which is the one argument I’ve ever raised *against* compartmentalizing through Flatpak, Snap, etc. 
     - I have a bunch of Flatpak programs with absolutely no settings sync or remotely near the capability to sync, so what do I do when I want to migrate?
-- Hey kids wanna see a dead husk of a man? Come find me three hours after I update my Neovim install. Dear lord, that thing breaks OFTEN. 
+- Hey kids wanna see a dead husk of a man? Come find me three hours after I update my [[code-editors#Neovim|Neovim]] install. Dear lord, that thing breaks OFTEN. 
 ### Future?
 Upgrades are inevitable with any piece of hardware. Now that my GPU is up to a 3060ti from a 1650s, I'm looking to upgrade my cpu. I'll follow through on that when linux figures out thread scheduling on newer cpu die layouts (p-core e-core is still rough at the moment).
 ## Software

content/Projects/vfio-pci.md

@@ -0,0 +1,30 @@
+---
+title: eGPU Passthrough with VFIO
+tags:
+  - project
+  - seedling
+  - difficulty-advanced
+date: 2024-03-17
+lastmod: 2024-03-18
+draft: false
+---
+I dabbled in running a Windows 11 KVM under an Arch Linux host on my laptop and passing my 3060 Ti eGPU to the vm. Here's how it went.
+
+## Bottom Line
+It works! Here are the **fact takeaways**:
+- The Framework Laptop 13 11th gen's motherboard separates a Thunderbolt 3 eGPU into its own IOMMU group that only contains the card, its audio device, and the Thunderbolt controller.
+	- There are two Thunderbolt controllers on the mobo (one per side), I'm unsure what would happen if there were two TB devices on one controller. All my tests were with one TB device per controller.
+- PCI passthrough to the VM works without a hitch once you find the right IDs.
+- As long as the VM is shut off when you do it, you can hot-plug and -unplug the eGPU since the driver it binds to isn't a video driver.
+
+And **my thoughts**:
+- It's just too much trouble on a laptop.
+- This setup is even more CPU bound than normal eGPU. The windows VM frequently pegs at 100% CPU (partially a Windows issue tbf), and I couldn't even run a game on it without it freezing. I'd need to retest on a higher spec CPU to be certain that this isn't workable for me.
+## Method
+I followed [Archwiki - PCI Passthrough through OVMF](https://wiki.archlinux.org/title/PCI_passthrough_via_OVMF). Be sure to follow the steps that require you to modify the modules included in your initramfs + their dependencies, **as well as** the module configuration with modprobe config files.
+
+Kernel arguments: `intel_iommu=on iommu=pt modprobe.blacklist=noveau,nvidia,nvidia_modeset,nvidia_drm vfio_pci.ids=8086:1576,10de:2489,10de:228b systemd.mask=all-ways-egpu-boot-vga.service,all-ways-egpu-shutdown.service,all-ways-egpu-set-compositor.service`
+- The 3 PCI IDs are the card, its audio device, and the Thunderbolt bridge.
+- If your host isn't already setup to use the NVIDIA card in Linux, you probably only need `intel_iommu=on iommu=pt modprobe.blacklist=noveau vfio_pci.ids=8086:1576,10de:2489,10de:228b`.
+## Next Steps
+Since the performance is subpar, I think messing with Linux hotplug will be more worthwhile. The PCI arguments on egpu.io cause my system to crash so I need to spend like a day troubleshooting that at some point. When I do, I'll make another entry on it.

content/Updates/2024/mar.md

@@ -6,13 +6,17 @@ tags:
 date: 2024-03-01
 ---
 ## Housekeeping
-Howdy, y'all. *Trump v. Anderson* got a decision, and it's about what I expected. I'm not a Supreme Court scholar (I just moonlight as a reactionary haha), so here's someone who **is** to explain the effects. Unfortunately on Substack, [Steve Vladeck's One First: The Shoddy Politics of Trump v. Anderson](https://stevevladeck.substack.com/p/70-the-three-biggest-problems-with)
+Howdy, y'all. I've now been maintaining this garden for about 6 months, and I've definitely found a rhythm!
 
-I'm also trying to improve my writing style, because I struggle with conveying high-tech, informed entertainment for all audiences. Suggestions appreciated, so please consider my more verbose opinions on tech in this garden a continuing work-in-progress until I find a voice I'm happy with.
+*Trump v. Anderson* got a decision, and it's about what I expected. I'm not a Supreme Court scholar (I just moonlight as a reactionary haha), so here's someone who **is** to explain the catastropic effects. Unfortunately on Substack, [Steve Vladeck's One First: The Shoddy Politics of Trump v. Anderson](https://stevevladeck.substack.com/p/70-the-three-biggest-problems-with)
+
+I'm also trying to improve my writing style, because I struggle with conveying high-tech, informed entertainment for all audiences. **Suggestions appreciated**, so please consider my more verbose opinions on tech in this garden a continuing work-in-progress until I find a voice I'm happy with.
 ## Pages
 - Making significant headway on the AI infringement essay. **Status 80%**, I might just publish soon after some heavy edits to curb verbosity.
 - Seedling: [[Resources/law-students|Resources for Law Students]] (extracted from [[Essays/law-school|Law School is Broken]], new section added)
+- Seedling: [[Projects/vfio-pci|eGPU Passthrough]]
 - Content update: [[Projects/rss-foss|Toward RSS]]
+- Content update: [[Essays/on-linux|The Linux Experience]]
 ## Status Updates
 - Fixed the license on the repo, it mistakenly identified my written content as MIT-licensed.
 - Added RSS feeds to the homepage's metadata, which should allow better integration with auto-discovery tools such as [RSS Is Dead](https://rss-is-dead.lol).

content/bookmarks.md

@@ -5,7 +5,7 @@ tags:
   - misc
   - seedling
 date: 2024-01-13
-lastmod: 2024-01-143-07
+lastmod: 2024-03-07
 ---
 One of the core philosophies of digital gardening is that one should document their learning process when trying new things. As such, here's my very disorganized to-dos and to-reads in the form of a public bookmark list. This page will change very often.
 

content/templates/new-note.md

@@ -3,4 +3,5 @@ title: New Note
 tags: 
 date: <% tp.date.now("yyyy-MM-DD") %>
 lastmod: <% tp.date.now("yyyy-MM-DD") %>
+draft: false
 ---