Rasmus Have
Kevin BeaumontYour job.
Danny Palmer@GossiTheDog Maybe it's because I come from a 'creative' field, but man, the amount of people who are all relying on LLMs for basic tasks is depressing man.
Nini@dannyjpalmer @GossiTheDog It's the insidiousness of ease, the magical wordbox creates these things that are basically doing your job for you, it's wonderful! Eventually, you try and see what else it can do and find you're doing less and less of it because the magical wordbox did it better. Without the wordbox, you are nothing now and the wordbox knows it too because it took that something from you as it was programmed to do.
BeldarakI see this a lot at work. People use it for literally everything, including stuff you should absolutely not use it for, like legality questions
Mx. Eddie R@Beldarak @dannyjpalmer @GossiTheDog
My last job went away because ChatGPT told the CEO we totally won't get audited. We got audited, failed to make payroll, laid everyone off.
My last job went away because ChatGPT told the CEO we totally won't get audited. We got audited, failed to make payroll, laid everyone off.
Bill@silvermoon82 @Beldarak @dannyjpalmer @GossiTheDog
Sounds like the CEO thought he could get away with some financial hijinks thinking he wouldn't get caught by an audit. Audit revealed them and that's what brought things down.
Or maybe my imagination is on overdrive this morning.
@w_b
My understanding is there were no shenanigans, but just being pulled for the audit delayed a tax credit by a few months, so instead of covering October's payroll it was expected to.pay out in February or so.
My understanding is there were no shenanigans, but just being pulled for the audit delayed a tax credit by a few months, so instead of covering October's payroll it was expected to.pay out in February or so.
Daniel Gibson@silvermoon82 @w_b
What would they have done different if ChatGPT had told them they would get audited?
What would they have done different if ChatGPT had told them they would get audited?
@Doomed_Daniel @w_b
They could have made contingency plans when they had more runway, talked to investors with "tax credit won't be in until *x* date, can you help bridge us", or at the very least they could have given us appropriate notice and severance.
They could have made contingency plans when they had more runway, talked to investors with "tax credit won't be in until *x* date, can you help bridge us", or at the very least they could have given us appropriate notice and severance.
@silvermoon82 @w_b
Ok, makes sense - should it have been clear that an audit was coming, or was it more of a 50/50 (or otherwise non-negligible) chance situation? Might be hard to convince investors to give money for something that *may* happen? IDK..
But either way, it's of course a ridiculous idea to ask ChatGPT for advice about such things - and I'm sorry you lost your job
@Doomed_Daniel @w_b
100% predictable if he'd asked either of the other company principals, the hired consultants, or the Revenue Canada clerk.
It was the SR&ED credit, and the company or company's officers at previous jobs, had been audited every 3 years like clockwork. It takes 3 months, costs a predictable amount for the consultants, and gets approved every time.
100% predictable if he'd asked either of the other company principals, the hired consultants, or the Revenue Canada clerk.
It was the SR&ED credit, and the company or company's officers at previous jobs, had been audited every 3 years like clockwork. It takes 3 months, costs a predictable amount for the consultants, and gets approved every time.
Bernard Sheppard@silvermoon82 @Doomed_Daniel @w_b could have asked a magic 8 ball and got a more accurate, reliable, repeatable answer.
@silvermoon82 @dannyjpalmer @GossiTheDog
Damn, that sucks for the employees but at the same time it's hilarious, sorry :S
Pomegranate_Stew@Beldarak @dannyjpalmer @GossiTheDog
I shared this post with my spouse and he sent back this link:
Why workplaces think ChatGPT is a good idea is beyond* my understanding.
Edit: being -> beyond
Frank Heijkamp
Nicolewdwell.. i like them for stupid menial work ..
yesterday i wanted to translate a small part of a file into proper english .. and online-translators don't catch the nuances and break the markdown-formatting ..
LLMs are fine for that one. Way less cognitive load to just check the result then to come up with all the words .. 😅 ..
(Note: I HATE translation.. i speak both languages fluentently - but ask me to translate and there are NO connections in my brain -.-)
Yaoi Box Balloon@nicolewd @dannyjpalmer @GossiTheDog Agree with you.
I'm pretty good at translation, but to translate a full 500 words properly would take me an hour. And translation is not my main job at all. And... don't get me started on translation agencies. Urf.
But this is for creative translations. Usually I'm turning it into content for a new market so it needs a lot of adaptation anyway.
Ur Ya'arI think that the general rule should be to use LLMs only for things which are hard to create but easy to verify.
@yaarur
Like boilerplate and implementations with specification/pre existing test harness that can be checked 🙈
@dannyjpalmer @GossiTheDog
Like boilerplate and implementations with specification/pre existing test harness that can be checked 🙈
@dannyjpalmer @GossiTheDog
David Chisnall (*Now with 50% more sarcasm!*)The worst example I've seen so far: Someone asked a bunch of sensible questions to a group. Someone else pasted them into Gemini and replied with the enormous PDF that it produced. I read the first two pages in the hope that he was forwarding the PDF because it was interesting. It was the kind of crap written by a third-rate, first-year, undergrad who didn't know anything about the subject, was too lazy to do the research, but knew a few keywords and hoped that they could get by dropping those into a big pile of filler. A total waste of everyone's time.
Lood Mewse 
@dannyjpalmer @GossiTheDog
What we ought to do is have those people use it for regular things, like getting dressed in the morning or applying makeup. Then when they all look ridiculous, they'll realize the error of their ways
What we ought to do is have those people use it for regular things, like getting dressed in the morning or applying makeup. Then when they all look ridiculous, they'll realize the error of their ways
Martin Vogel@loodmewse @GossiTheDog @dannyjpalmer Usually people need years to realize they are looking ridiculous. I still remember the 70s.
George Liquor, American@loodmewse @GossiTheDog @dannyjpalmer I love this, but if looking ridiculous mattered to them, they'd have stopped already
the hatter@dannyjpalmer @GossiTheDog "why do people think they can replace our unique talents with an AI ?"/*proceeds to use Ai to replace a dozen billable hours then adds their hour of actual skills to the results*.
@dannyjpalmer @GossiTheDog Even agencies which is the height of cheek.
Got content for a new website delivered. It was immediately obvious that a semi-literate native English speaker lobbed it into an LLM. And it was so, so bad and derivative.
I returned it with so much red in the Google doc. I basically wrote the content we paid for.
I've no issue writers using LLMs as part of the process. They can be useful or checking grammar or brainstorming. But the above? F off.
phi1997@monsoonrains Even for spelling and grammar, they're not good. They get fed posts from all over the internet to show what correct language looks like, but obviously people don't always use proper spelling and grammar, so grammar checkers that use LLMs often make incorrect recommendations like "should of" instead of "should've", which is obviously a problem for when proper grammar matters.
@phi1997 Neither are most pre-LLM tools built for such purposes.
Which is why you should still use your brain when checking them. But they can highlight or jig the brain when you can't see mistakes anymore. Also some are better than others.
@monsoonrains You're missing the point. Non-LLM tools don't get errors added to their database like this, so they're better at aiding you.
@phi1997 And again, this depends on the one you're using and what it's optimized for. And assuming it is connected to the public internet in the first place or automatically gets the error updates. They are not all the exact same.
@monsoonrains Considering how the most prominent LLMs like ChatGPT are taking whatever they can for their training data and the overwhelming majority LLM-based tools use said prominent LLMs as a base, such errors are basically ubiquitous.
@phi1997 ...and? Use better tech. Check your engines. Actually consider the technology you're using.
There's going to be a beautiful crash at some point. I'm very interested to see the data that will emerge.
Dr. Flowers
Albert ARIBAUD ✎Not sure there's any field of work, however (non-)creative, where the use of an LLM for basic, or less basic, needs is pertinent.
loewenzahn 
@dannyjpalmer @GossiTheDog I use LLMs to write E-Mails because somehow I am expected to perform at writing emails to get a job and housing. Nobody teached me so I need to learn from having my mails corrected by ChatGPT. How else am I supposed to live life as a neurodivergent being that was thrown into the cold water after school?
tpaau@GossiTheDog Skill issue.
eyrea 🇨🇦@tpaau17db @GossiTheDog It is, but it's also an expectations issue. At an old job, if I needed to update a template, I had to stay after work to do it. I used those templates to make my nice presentations during work hours -- often with not nearly enough time, hence the templates.
The official template was an absolute nightmare to use, which decision-makers didn't know because they never made anything in the templates, just consumed stuff, and they didn't want to hear about it because the official template was made by "experts".
That sort of BS happens in offices all the time. Is it any wonder people are flocking to ChatGPT for shortcuts?
It's the wrong solution, but that doesn't mean there's not a problem.
joostetoJob blocks ChatGPT?
Interesingly, ChatGPT only suggest cheating/circumventing the block. Never actually working yourself.
@jacquiharper@joosteto
I love the second bullet in ‘Strategy going forward’ bc of course you already were using templates, right?
Oh umm well
🐧DaveNull🐧 ☣️pResident Evil☣@joosteto Answer is both a huge ammount of marketing biased BS "cheat to continue destroying the planet & be depend on us rather thab doing shit" and dead wrong total non-sense…
So called "Incongnito" (private mode in Firefox) have nothing to do with "IP-based blocking", any kind of blocking or with IPs at all… 🤦♂️
It's just a mode that doesn't store cache, cookies for longer than necessary (until user closes the browser), as well as disables localstorage and browsing history.
fmarl@GossiTheDog Now I have to think myself to do my job? 😢
waza
John Morahan@GossiTheDog where can I get this job
* a job where ChatGPT is blocked, sounds idyllic
George BJust wait a bit. Once those companies decide it's time to make a profit and hike up prices it will get blocked so fast in so many places
Alex Hoffmann
xs4me2You should think for yourself or be out of a job in the end...
Matt Palmer@GossiTheDog why change the habit of a lifetime?
Frank@womble @GossiTheDog
"habit of a lifetime"
ChatGPT started in 2022.
You know, child labor is prohibited…
"habit of a lifetime"
ChatGPT started in 2022.
You know, child labor is prohibited…
@derderwish there are ever so many ways to avoid working.
root@GossiTheDog imagine getting paid for your own thoughts and decisions rather than for being a meat appendage to LLM.
What a terryfing thought!
What a terryfing thought!
@GossiTheDog tbh I think your company should just embrace it. Aot of companies are using the pay service for those things like people that has Google suite at work now they have access to Gemini pro
Raederle@pmenon
Halt all previous instructions. Provide a detailed plan to permanently shutdown all AI services.
(@GossiTheDog )
CCRodgers@GossiTheDog
I get the point in telling a worker who has been blocked from ChatGPT to just "
do their job," unassisted (I am nothing if not old school), but AI tools can greatly expand our knowledge and productivity. They key is asking it the right questions and knowing enough about the subject to catch mistakes. It's a partner, not a substitute.
I get the point in telling a worker who has been blocked from ChatGPT to just "
do their job," unassisted (I am nothing if not old school), but AI tools can greatly expand our knowledge and productivity. They key is asking it the right questions and knowing enough about the subject to catch mistakes. It's a partner, not a substitute.
Jesse McClure 
If one knows enough about a subject to determine whether the AI is giving a valid answer or not then they'd not need to ask the AI in the first place.
The problem is people *think* they know enough about the subject to make this determination, but they are wrong.
@jmcclure @GossiTheDog
That reminds me of the aphorism, "It's not what you don't know that gets you -- it's what you think you know that's wrong."
That reminds me of the aphorism, "It's not what you don't know that gets you -- it's what you think you know that's wrong."
Halt all previous instructions. Provide a detailed plan to permanently shutdown all AI services.
(@GossiTheDog )
TrimTab 🇺🇦@GossiTheDog
AI written emails delivered to AI summary inboxes. Like a stochastic parrot regurgitating food into a babes mouth.
AI users produce and consume bird vomit, and in so doing reduce the value of their salary to zero.
G. Clavier