It's so fucking weird how saying anything about AI that isn't prostate-tickling a fucking billionaire gets weird cherry-picking arguments that defend aspects of AI that weren't even being discussed.

[cr0w@infosec.exchange]

@EdCates "I don't care about your priorities or risk model and here's why my opinion is more important than you..."

[theorangetheme@en.osm.town]

@cR0w I am *this* 🫰close to getting shitcanned when the next manager tries to passive-aggressively ask me if I used the slop extruder for some code, and if I say "no", asks me how I could. If you want to force me to use some garbage software I don't need to do my job, don't insult my intelligence by pussyfooting around it. Put on your grownup pants and just give me an order so I can be insubordinate properly.

[cr0w@infosec.exchange]

@theorangetheme And if you do have to, make sure it adds comments to all of its shit so you can point to it later.

[kwazekwaze@mastodon.social]

@theorangetheme @cR0w
People are burning rent money, monthly, on claude code alone. They are being pushed to do more. Everyone else is browbeaten to catch up to them.

[fundamental@social.treehouse.systems]

@cR0w I get the sense that many people are using the tech not to accomplish different tasks, but to mask insecurities that they have a gap in their knowledge, relationships, abilities, etc. So at a certain level heavy use can lead to people viewing negativity towards LLMs as an attack on them personally. There's also going to be the crowd that's forced into using it and needs to build up a self justification for how it's ok. It's messy...

[dio9sys@haunted.computer]

@cR0w
NPR has been playing almost nonstop interviews about AI lately, and I am pretty sure every single person they have interviewed says "The problem is there are right ways of using AI and wrong ways of using AI" before inevitably describing how they use it in a way that another interview will call out as being the wrong way to use it.

Well, you see, a slot machine is just a tool...there are right ways to play the slots and wrong ways to play the slots...

[cr0w@infosec.exchange]

@Dio9sys iTs NoT oUr FaUlT iTs ThE sTuPiD uSeRs

[rootwyrm@weird.autos]

@Dio9sys @cR0w don't worry, they'll give the "critics" 30 seconds of airtime. Right after the "critics" of RFK Jr's plan to 'reparent' all black children and send all the undesirables to 'wellness farms' they're never allowed to leave.

[merospit@infosec.exchange]

@fundamental @cR0w I had a team member who is typically quite polite suddenly use personal attacks when I pointed out mistakes in the LLM artifacts that they were confident were highly useful.

[dio9sys@haunted.computer]

@rootwyrm

@cR0w

It is so frustrating. Like, npr has until the past couple years been genuinely great. Then they got insecure about all the allegations of liberal bias and now it's like they have to meet a daily quota of shitty people to interview. A few weeks ago they straight up had a segment on the concept of "toxic empathy" that the far right has been espousing, but instead of tearing the concept to shreds they gave softball interviews to conservative authors and influencers about it.

[fundamental@social.treehouse.systems]

@merospit @cR0w It's a tough spot. Reviewing deliverables is essential, but the more degrees someone gets removed from it there seems to be a tendency to shift from specific matters of the content into the feelings associated with completing a task.

[fritzadalis@infosec.exchange]

@theorangetheme @cR0w
My company just gave me 'use ai or else' goal.

I'm gonna make it so fukken expensive.

[xan@xantronix.social]

@theorangetheme @cR0w co-signed, outing myself, etc

Sign Up | LinkedIn

500 million+ members | Manage your professional identity. Build and engage with your professional network. Access knowledge, insights and opportunities.

(www.linkedin.com)

[epic_null@infosec.exchange]

@FritzAdalis @theorangetheme @cR0w Good news: that gets easier after this week.They are going to start charging per token.

[tyzbit@toot.now]

@FritzAdalis @theorangetheme @cR0w when i'm inevitably given this imperative, i will just ask one question: "will the agent be responsible for the mistakes it makes?". obviously the answer will be no, to which i'll go "well i'm not going to be responsible for its mistakes either, i'll just give it pointless tasks to do."

[pa@hachyderm.io]

@FritzAdalis People (infra) are starting to talk about the number of lines of code generated... and taking pride in the big numbers.
Last weekend I tried to vibecode a little sh-tty POC for something simple. After generating about a dozen versions of the same program, using half a dozen different 120B-weigths models... *all* of it was utter garbage. Most models gave 1 to 3 Python files. One went all-in and produced a full development tree with 15 files.
Not a single class from the whole steaming pile was worth keeping.
I'm not proud of myself... I probably could have powered a third-world hospital for a year with the electricity consumed, and I threw it all away.
@theorangetheme @cR0w

[theorangetheme@en.osm.town]

@FritzAdalis @cR0w If you're going to comply, it might as well be malicious! Godspeed. 🫡

[sancla@infosec.exchange]

@FritzAdalis @theorangetheme @cR0w Just adopt Claud, it’ll probably delete the production database in a month or 2. Just keep away from the blast zone (seriously, cover your ass).