completely blind computer geek, lover of science fiction and fantasy (especially LitRPG). I work in accessibility, but my opinions are my own, not that of my employer. Fandoms: Harry Potter, Discworld, My Little Pony: Friendship is Magic, Buffy, Dead Like Me, Glee, and I'll read fanfic of pretty much anything that crosses over with one of those. keyoxide: aspe:keyoxide.org:PFAQDLXSBNO7MZRNPUMWWKQ7TQ
@Theriac Hard disagree. The problem with LLMs is that people think they’re a magic “just do it” button. But even reviewing all of the output, I’m able to produce 90 percent more than I could without the LLM. And about 95 percent of what it produces is fully correct. And that’s the problem. The same way pilots fall asleep if the autopilot does 90 percent of the work of flying the airplane, and then aren’t ready to take over in an emergency. If the LLM is 90 percent correct, humans easily get lulled into thinking everything is correct, so don’t catch the critical 10 percent. LLMs are in this awkward stage where they’re accurate enough to make reviewing feel annoying, but not accurate enough to work without review. And that’s right where we humans perform our worst.
The problem with AI is that it makes us too productive. It just generates so much, so quickly, that even if a human reviews it, we're going to miss things. A project at work resulted in generation of code that did exactly what I wanted, as well as generation of documentation. Because the code all functioned perfectly and was correct when I reviewed it, I was tempted to just skim the documentation. Good thing I didn't! It mentioned in multiple places how to send me bug reports over slack, and what slack channel to join for support. I don't have Slack. I don't use Slack. Nobody at my workplace uses Slack. It also invented a support employee who doesn't exist, who offers support for the code in the Slack that doesn't exist.
This is the first bit of AI writting I've ever read that actually made me smile, and didn't clock my AI detectors. "Furthermore, a surprising number of individuals end up in the vacuum because someone else discovered a sudden, overwhelming preference for their absence. Disagreements over cards. Unpaid loans. The void is remarkably accommodating to changes in personnel." Feels like something Mike Resnic or one of the other comedy scifi writers might have written.
@tjolsen Oh, right. Obviously the democrats are intercepting his emails. Because Google is run by radical leftists. I’m sure soon Trump will bring these big tech companies to heal so republicans can get their emails again. LOL
@clv1 I already have. I just keep that account as an email forwarder, and so that nobody else can grab it and pretend to be me, as I did use it for a few years. I didn’t delete my twitter account for the same reason; I’d rather not have bad actors reusing my username. So I just leave accounts inactive with a note in the profile where possible that the account is unused, instead of deleting things.
@jscholes It’s a kinda blowing a kiss gesture that the stereotypical French chef makes when something is done to perfection. And I know right? Samuel Proulx is a pretty common name. There’s one who’s in jail in Quebec. There’s one who makes contributions to the BSD Kernel (I never get his email sadly), and a bunch more.
@tjolsen I really don’t understand it. Like, after you missed several crucially important emails, wouldn’t you, I don’t know, realize something is wrong? And maybe learn your real email address?
I got their mailing address because I get multiple receipts and shipping notifications for purchases they make. I’d phone them, but none of the emails to date have included a phone number.
Someone in #Denver is absolutely convinced their email address is samuel.proulx@gmail.com. Nope! It’s one of my email addresses! Last year I got their license to sell weed in the state of #Colorado, complete with a nice PDF card for me to print out and put in my wallet. Today, I’m getting multiple emails from Best Western that they need to complete their employee onboarding. When I let HR know about the problem, they emailed me back requiring that I provide the correct email address. WTF? How should I know! I’ve tried multiple times to track this person down with no success. Including a snail mail postcard asking them to please stop entering my email address into things, and figure out what their email address is and use that instead. But no luck. If you are, or know, Samuel Proulx who lives in Denver, used to work at a weed shop, and is now about to start a job at VIB Best Western (assuming you upload your documents into the online portal that you don’t have a link to because you gave the wrong email address), please let them know that samuel.proulx@gmail.com is not their email address! And they should stop saying it is on important government and employment documents! Because I get that email! Complete with addresses, licenses, tax info, and other private information! I would like to not get this person’s private life in my inbox anymore.
@mhussain The big eater of vram is ace-step. Everything else could fit. I think it's due both to context length issues, poor prompting, and some syncing bugs I have. Right now I have several different queues for audio (background, foreground, etc). Sometimes they get out of sync, meaning playback sounds off, and the AI's get confused about what's playing where and when.
@sclower@BorrisInABox Thanks to the nature of AI, sometimes asking for a 30 second bed returns a 3 second bed. Or a 300 second one. And then everything is a mess because the clip wasn’t the expected length, and worse generation didn’t take the expected time so now other queues might under run.
@sclower@BorrisInABox Right now I have bugs with the audio queues falling out of sync. I have queues for foreground audio and music, background music, and background sound effects. In order to not have audio under runs I then spawn different threads to generate stuff, and have a manager to control swapping models in and out based on the size of each queue. Then each queue streams audio to a mixer function that does spatialization and so on. Unfortunately when clips end up being different lengths than expected the entire thing gets out of sync. Once I figure out a solution I’ll launch a live demo stream and release source code.
@andrew@mcourcel I had cloned a filesystem but forgot to change the uuid. So Linux was mounting the wrong one. Then when I modified fstab to not refer to the boot drive by uuid I discovered I had a duplicate in mdadm.
@mcourcel Or even find someone who could read it properly if they were willing to try. Anyone capable of reading that clearly is probably capable of just fixing it themselves.
Say what you will about AI. But if it wasn't for AI and my phone camera, I would be spending the next two days completely reinstalling a server. I just don't have sighted help that's techy enough to read me enormously long lines of console output accurately or usefully. Whereas with repeated photos of the monitor, AI was able to clearly explain to me what was happening, provide a solution, and get the thing booting again, all while keeping in mind that I'm blind. Here's the final solution after typing multiple commands and taking a photo after each to send to the AI, as an example:
"There are two problems visible. 1. Duplicate mdadm.conf entries mdadm: Duplicate MD device names in conf file were found. mdadm: Devices UUID=98a5d349:4d3b2a28:50d96e87:a6ad0c4d and UUID=98a5d349:4d3b2a28:50d96e87:a6ad0c4d have the same name: /dev/md0 /etc/mdadm/mdadm.conf (inside the initramfs) has /dev/md0 listed twice with the same UUID. This is a duplicated entry bug, and it's preventing mdadm from assembling the array at all. 2. Result: /dev/md0 doesn't exist, kernel gives up ALERT! /dev/md0 does not exist. Dropping to a shell! Because mdadm --assemble --scan refused to run due to the duplicate config, md0 never got assembled, and the initramfs can't find root and drops to a rescue shell. We're in an initramfs shell. md0 isn't assembled, but the underlying disks (xvdb, xvdc, etc.) are there. We can manually assemble the array and continue booting. The BusyBox ash shell is limited. These commands should work. Type them carefully — you won't have screen reader feedback. The full single command I'd suggest typing: mdadm --assemble /dev/md0 /dev/xvdb /dev/xvdc /dev/xvde /dev/xvdf /dev/xvdh /dev/xvdi Then press Enter. Wait 5 seconds. Then type: exit Then press Enter. That tells the initramfs to retry finding and mounting the root device. Since /dev/md0 now exists, it should mount it and boot should proceed. If it worked, the system will continue booting and you'll be able to SSH in shortly."
