@vwbusguy@mastodon.online
@vwbusguy@mastodon.online avatar

vwbusguy

@vwbusguy@mastodon.online

Do you use your powers for good or for awesome?

Senior #DevOps Engineer at a major research institution (views/opinions my own) and a severe #Linux and #opensource enthusiast. Also into #retrocomputing, #fountainpens, #Volkswagen, #music, and #cooking. #fedi22

Hoosier currently living in California.

Pixelfed: https://pixelfed.social/vwbusguy
Firefish: https://easymode.im/@vwbusguy

This profile is from a federated server and may be incomplete. Browse more on the original instance.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

The state of LLM right now is more or less that of a corporate-sanctioned Napster that hasn't (yet) been gutted by litigation. I'm curious to see what kind of more mainstream torrent/Spotify type thing emerges from this after the hype bubble inevitably pops. In the meantime, I suggest not betting the bank on any one major vendor. Copilot, ChatGPT, Llama, etc could all be one good lawsuit away from shutting down. Factor that into the business risk of adoption.

vwbusguy,
@vwbusguy@mastodon.online avatar

Truly open source models trained on public domain or fully owned data require more up front investment but also carry less foreseeable risk than other options. To that end, ARM hardware is momentarily cheap at a time where very little of anything is cheap, so buying some Ampere boxes and having it scrape your own website and data is a pretty good proposition. Whatever LLM service, don't expect it to reliably think or make decisions for you.

vwbusguy,
@vwbusguy@mastodon.online avatar

@1dalm Your assertion doesn't make much sense to be. Microsoft Copilot, in its various flavors, is based on OpenAI GPT. Github Copilot also sourced from copy-left sources into a fully proprietary service. It's a matter of time before that eventually plays out in court.

Meta actively markets Llama as open source when it categorically isn't at all by its license terms. It unambiguously doesn't meet OSD.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

With all the AI talks, debates, and such this week, I think I'm overdue for a watch of the original Ghost in the Shell.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

The problem with the is that there are a whole lot of people who aren't very discerning, critical thinkers.

vwbusguy, to linux
@vwbusguy@mastodon.online avatar

Audio on was long the butt of many joke. My goodness, has delivered. I love how well my A/V IO works and behaves out of box on these days.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

Current state of internet forums:

Spammers using AI to get around the spam filter AIs that forums delegated moderation to.

vwbusguy, to opensource
@vwbusguy@mastodon.online avatar

FYI - is NOT . The license is categorically not open source. Among other things, the llama 2 and 3 licenses explicitly violate Field of Endeavor.

I see all sorts of blogs and marketing materials claiming things are "open source" because they used llama somewhere. Please do not take these claims at face value.

vwbusguy,
@vwbusguy@mastodon.online avatar

This isn't specific to #Llama either, but other AI models listed on the #Meta page as "open-source".

For example, seamless carries a non-commerce clause.

https://github.com/facebookresearch/seamless_communication/blob/main/LICENSE

#opensource

vwbusguy, to RedHat
@vwbusguy@mastodon.online avatar
vwbusguy,
@vwbusguy@mastodon.online avatar

@maxamillion @rfc2549 Considering the default license for galaxy stuff is Apache, an opt-out was more than they legally needed to offer.

vwbusguy,
@vwbusguy@mastodon.online avatar

@maxamillion @rfc2549 This is a significant distinction. Copilot is expressly proprietary. If all of these code generating models were GPLv3 and included clear license attribution, there would be much less ambiguity and controversy around them.

vwbusguy,
@vwbusguy@mastodon.online avatar

@maxamillion @rfc2549 And there is a valid point that some prominent companies in the space are using data inappropriately to do inappropriate things and almost certainly in reckless violation of copyrights.

The tech definitely got ahead of the courts on this one.

That said, there are legit ways to train and use AI. I've been in attending a symposium on this stuff all week and am about to go watch an ethics panel discussion on this right now, FWIW.

vwbusguy,
@vwbusguy@mastodon.online avatar

@rfc2549 @maxamillion It is being recorded, but I don't know the extent to what will be made public, since it is a UC Symposium.

https://ccs.ucsb.edu/events/ai-community-practice-cop-spring-symposium-2024

vwbusguy, to linux
@vwbusguy@mastodon.online avatar

#SUSECON is scheduled during #Juneteenth this year, which is unfortunate. I know it's not a German holiday, but I imagine it will impact participation from Americans.

#SUSE #Linux

vwbusguy, to random
@vwbusguy@mastodon.online avatar

Overheard recently:

"Do you ever feel guilty about plagiarism for using generative AI?"

"Not at all. The truth is we've been plagiarizing all along. Every JD we've ever posted started as a copy and paste from a similar job posting elsewhere. Every policy was based on another, likewise."

🤔

YvanDaSilva, to random
@YvanDaSilva@hachyderm.io avatar

How many of you wanted to start a company just to try to do better than what you are experiencing?

vwbusguy,
@vwbusguy@mastodon.online avatar

@vathpela @YvanDaSilva I don't mind being a boss, but otherwise I'm with you - I've done it once before (I owned a coffee shop) and the stress and overhead of owning a business quickly and utterly consumes your entire life.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

I'm offended that I should be asked to verify my 2FA credentials before purchasing HomeStarRunner merchandise. Have the data scrapers not learned anything about me to not know that this is entirely on brand?

vwbusguy, to random
@vwbusguy@mastodon.online avatar

Sometimes I still can't believe that I actually own a PRS. My goodness this guitar is amazing!

vwbusguy, to linux
@vwbusguy@mastodon.online avatar

I wonder if has some kind of minimal and/or way to run x86 applications on an host, similar to Rosetta for MacOS or the similar WIP Microsoft compatibility.

vwbusguy, to random
@vwbusguy@mastodon.online avatar

I just had an AI help me write some necessarily regex heavy stuff and I'm not complaining about that.

vwbusguy,
@vwbusguy@mastodon.online avatar

There was some stuff I wasn't sure how to sanely do with and thought I might have to resort to a script and the AI helped me get out of the weeds with the playbook.

vwbusguy,
@vwbusguy@mastodon.online avatar

On this note - if you're thinking about implementing some novel, innovative new config file syntax for your fancy enterprise program, please don't. 😭

qlp, to random
@qlp@linh.social avatar

I keep forgetting that my Prometheus and Grafana server at home is running on a Raspberry Pi and not on a Proxmox VM.

I almost always end up downloading the amd64 version instead of the arm64 version. It also doesn't help that amd64 and arm64 look a little too similar when not looking at things carefully.

🤦🏻🤦🏻🤦🏻

vwbusguy,
@vwbusguy@mastodon.online avatar

@qlp Yeah, I remain annoyed at Microsoft and Canonical for this (and everyone who followed from there). x86_64 and aarch64 are far less ambiguous.

bagder, to random
@bagder@mastodon.social avatar

Remember to take the user survey 2024 - if you can spare a few minutes.

https://daniel.haxx.se/blog/2024/05/14/curl-user-survey-2024/

vwbusguy,
@vwbusguy@mastodon.online avatar

@bagder Got mine in! Thanks again for the excellent toolset!

  • All
  • Subscribed
  • Moderated
  • Favorites
  • megavids
  • GTA5RPClips
  • thenastyranch
  • ethstaker
  • everett
  • Durango
  • rosin
  • InstantRegret
  • DreamBathrooms
  • magazineikmin
  • Youngstown
  • mdbf
  • slotface
  • tacticalgear
  • JUstTest
  • kavyap
  • modclub
  • cisconetworking
  • cubers
  • ngwrru68w68
  • khanakhh
  • tester
  • anitta
  • normalnudes
  • Leos
  • osvaldo12
  • provamag3
  • lostlight
  • All magazines