skipper's musings – Telegram
skipper's musings
73 subscribers
104 photos
14 videos
2 files
47 links
i only write in lower case unless im serious.
i hope we can be friends on the modem.
blogs where I post longer stuff sometimes:
https://blog.skipper.blue/
https://highguard.net/
Download Telegram
lunduke used to be a moderate. he didnt really talk much about politics in computing, just funny stuff or drama that happened.
eventually he started commenting lightly on politics but instead of taking a side would just say "yeah i really dont think this behavior is good for the community."
🕊1👨‍💻1
of his last 12 videos, 3 are politics adjacent (4chan ofcom drama), 2 are not political (ai), 7 are political.
lunduke didnt change. he didnt "move right". he didnt watch too much youtube and get recommended racist alt-right joe rogan videos that turned him into a racist dudebro, or whatever.
he just got tired of idiots who got in because we didnt gatekeep hard enough.
i think this is a really good example of most western white guys who werent political 10 years ago. they started getting fatigued.
👨‍💻2
im pretty sure every single IoT company has burned their customers one way or another
>need an internet connection to dispense water
>need an internet connection to change your bed position, or to heat/cool it
>literally just shutting down and turning your device into a brick (see: jucero)
>does work slower and at a lower quality than you could do manually (see:jucero again, no wonder they shut down)
>locks you into their service/eco system and then jacks up prices, or starts charging for things that used to be free
skipper's musings
im pretty sure every single IoT company has burned their customers one way or another >need an internet connection to dispense water >need an internet connection to change your bed position, or to heat/cool it >literally just shutting down and turning your…
recently philips, maker of hue lights, changed their eula/tos to require an account. i dont have an account and the only way i use their services at all is by using their app (local only) to send commands to the hub that controls my lights, and connecting to their update server. I put my hub on a segregated vlan that cannot connect to the internet. the only way the hub gets updates is by staging updates on my phone and then pushing them at the absolute most inopportune times to the hub. This is not even something that i want to do. im forced to do this.
i dont consent to the new terms of service but the app locks all features unless i consent to the tos. i had to download a much less functional app to be able to connect to the bridge.
I ended up contacting support and had a circular argument with them:
"oh, you cant connect to the bridge until you consent? just hit okay on that notice then!"
"but i dont consent."
"then you wont be able to use the bridge."
"the one that i paid for? that is in my house? that i own? that's the issue im trying to fix. i cant connect to the bridge. my bridge. the one i own. the one that is inside my house."
"you just need to click consent and then you can use it again. :)"
we ended up having a conversation about consent. i asked things like "is it still consent if someone is coerced into it?" the support rep genuinely did not seem to understand what consent actually means.
the good news is i found an open source bridge server that can run on an rpi which controls more than just hue lights, so i can buy less expensive lights now.
this is what you get when you let indians and ai write your software
the first time i searched for ":lightning" it didnt come up with the ⚡️ emote (not pictured), but it did find it the second time. i hate software that doesnt reliably act the same each time and even more than that i hate software that isnt reliable the first time.
2👨‍💻1
the worst thing about being a high functioning autistic person is that you're stupid enough to say things to service employees to deserve getting the customer service voice and just barely emotionally intelligent enough to know they're doing it to you.
ai generated image with perfect text
🤯21
😁1
so i finally got around to reading the privacy policy for ms edge and uhhhh its not good
when you use edge for the first time (and never again afterwards) youll get a prompt asking you to "make your edge experience more useful" without actually describing what that means at all. its on by default and if you dont uncheck it, MS will slurp up your entire browsing history.
i've banned other browsers from corporate environments for less egregious privacy violations.
1
/g/‘s Tech Memes
Photo
im really starting to believe this meme
heres the rundown:
as ai gets better you can run it on cheaper hardware. it also increase in fidelity, meaning easier and better deepfakes or whatever youre doing with it.
people in charge dont like the idea of giving that power to the common man.
as a result they have to make it impossible to run AI on local hardware. the only way they want you to use AI is via a service that they control, so that you cant translate hitler speeches or AI generate a video of trump fellating bill clinton. (you dont need to generate that. it happened :3c )
so they decided to restrict the hardware. the biggest manufacturer of consumer RAM, Micron, has exited the consumer market. they owned a lot of brands including crucial. i think gskill was also micron.
between AI companies buying up all the ram and micron exiting, both regular computer ram and videocard vram have exploded in price, upwards of 500%. this makes it very expensive and difficult to run large AI models at home.
this is already horrifying, but what makes this truly apocalyptic is that modern devs absolutely suck at writing software that is not ram intensive.
writing an application for multiple platforms is hard. If you want to support windows, mac, android, and ios you have to write four separate applications.
the way around this problem is to write webapps- basically websites that are stored locally on your computer and run as a program on your computer. this way you just need one codebase for every operating system, and its easier to build a website than it is to write a native application.
the downside: webapps are VERY ram hungry. discord is a webapp. it easily uses half a gig of ram. whatsapp used to be a native application. it would use maybe 250mb of ram. meta recently scrapped the whatsapp native application and have switched to maintaining a webapp. the new webapp version now uses upwards of a gigabyte of ram (4 times more ram than before.)
what this means is that users are getting squeezed. you cant upgrade your ram because the upgrade would cost like 500$, but also your computer sucks because you have like half as much ram as you need
skipper's musings
im really starting to believe this meme heres the rundown: as ai gets better you can run it on cheaper hardware. it also increase in fidelity, meaning easier and better deepfakes or whatever youre doing with it. people in charge dont like the idea of giving…
how i see this playing out:
1. butlerian jihad: everything is done to kill AI. the bubble bursts. openAI dies in a massive fire. oracle takes a huge hit, is bailed out, and gets bought out by MS or palantir. MS takes a big hit and is bailed out by the government. any interest in local models dies because git repos dont want to get legal heat for hosting code that can be used for crime or something. they'll probably write laws making it illegal to gen images of children or deepfakes and that'll kill stablediffusion. services like midjourney and novelai die because payment processors dont want to touch that.

2. hardware prices stay sky high until "local" computers die. the only pc you can buy is a thin client that connects to a server that does all of your computing for you. you have zero privacy. windows becomes a truly online only operating system as a service. you pay monthly to "connect your computer to the windows ecosystem" and if you dont pay your computer becomes a useless brick.

3. prices drop and AI is made illegal in some weird way. distributing local models that can be used to do something "bad" becomes a crime. all models are capable of telling you how to build a bomb or generating obscene images and as a result generative ai dies.