Genuinely think using something like chatgpt in your job should be a sackable offence for many reasons but for one its so untrustworthy that it will always be a dereliction of responsibility and kind of deliberate sabotage
Comments
Log in with your Bluesky account to leave a comment
Depends how you use it though. Using it without checking would be like sending off marketing material to print and assuming every spell check suggestion was perfect
Sounds like inserting an extra step before doing the investigation you needed to do anyway. After all, if you're needing to look it up in the first place you probably don't know specifically what it hallucinated, right?
Yeah I wouldn't use it for research or investigation. I give it straightforward menial tasks to do so I don't have to, and I check what it gives me back.
Long time DevSecOps and Coding engineer, too many devs use generative AI to generate code, thinking that they can just "audit" and then publish it into codespaces, and from my perspective...its NOT right. Too many of us NEVER had that leverage when we were younger, we just had to figure shit out.
I use it in a limited and responsible way. I'm a teacher and I ask it to number gaps in texts or find songs with certain words or phrases. Or to scramble sentences etc. That is not an irresponsible use, it's just timesaving.
It’s not much of a time saver if you have to rigorously check the results because it can’t be trusted to do its job without making mistakes. And if you aren’t rigorously checking the results of the hallucination machine, what part of that is responsible exactly?
Lmao that’s barely generative is it? Counting words? Searching words? Scramble words? Come the fuck onnnn no need for chatgpt weve been able to do that for twenty years with ease thats your defence lmaooo
you are right however a lot of older people are so technologically inept they believe AI is a good thing and are often encouraged to use it by dubious higher-ups. the amount of older people I see praising AI in complete blissful ignorance frustrates me to no end
Understandable. I should have worded it better. I was merely stating how ignorant people are to its insidious nature and would rather use it as a crutch than actually develop a skill
When I was in college one professor constantly drilled into us, again and again, not to use stack overflow, that companies would sack us if we just asked someone online to answer a technical problem, that we needed to understand what we were doing and develop our critical thinking skills. So...welp.
It should also be a red flag of incompetence. When somebody tells me that ChatGPT improved their writing, I don't think better of ChatGPT, I think worse of their writing. Everything it generates is at best barely mediocre.
It's like autocorrect on steroids and occasionally LSD.
So many jobs in my profession (content marketing) are now demanding that we use gen ai for “efficiency”. It’s good at summarizing other existing content (sort of. If you only care about vibes) it cannot write for you. It’s infuriating.
Many tech companies are actively encouraging the use of AI by every employee, providing dedicated tools and built in products like MS Copilot, but with rules around the kind of data that can be shared. It's untrustworthy, but that's not stopping the rush toward the cliff. :(
Unless you're doing creative writing or something where the hallucinations might be helpful 😂
But yeah the lawyers who keep getting admonished for citing fictional cases is just awful. 🤦♀️
Using it for creative writing is NOT on, either. It's basically handing over imagination to a machine that is full of garbage and that steals from creatives to have a dataset from which to draw. You can't copyright stuff you don't write, either. It's basically cheating. So Nope. Just so much nope.
I was mostly trying to think outside the box for something that might technically work, but yeah it's not a good idea. Just better than contempt of court because you're relying on it entirely like a dummy 🤦♀️
I find it very helpful, and it speeds some work up a lot, but to get the benefit I have to be really clear on what I need from it, and what good looks like. Then relentlessly and iteratively understand and evaluate what it does. Not a replacement for thinking.
When is it appropriate to use the mistake machine? You don’t actually have to add a level of “may or may not be wrong” ever its always a dereliction of responsibility and a choice to care less about the outcome
When genai makes a mistake i think the person using it should be fully held responsible cause they chose to use an unreliable machine but so far seems like this isnt how things are being treated in the world there’s just excuse making for the uwu smol idiot robot
My company bought a version of chatgpt that's supposed to hold a bunch of our company info, so we can ask it stuff, but it is notoriously wrong. Rather than answer questions, our bosses tell us to ask gpt so now when they do that I have a number of saved screenshots I copy-paste in response.
A favorite is "how many R's are in strawberry?" There is obviously a coding reason why AI gets that wrong, but my bosses do not know the coding reason so to them it's just like 🤯. But also you can ask it to make ascii images and it is humorous to show a "drawing" of a dog that is definitely a snake.
Abdication of responsibility is the most pathetic and shameful aspect of humanity, and the entire basis of business is built on it. It was very clear that part of the hype cycle was driven by the idea that blame for any shortcomings or failures could be scapegoated to an entity that would-
Not receive any consequences for doing wrong. It was never that it would do a job as good as a human, but that it would be the perfect outsourcing machine.
My workplace recently gave clearance for hundreds of software developers to start using GitHub Copilot for code assist, with the caveat that it doesn't reach production without full code review.
I still don't use it. It can never give me right, or even close to right solutions.
If it was for anything remotely important, shouldn't you check that you know *why* it works, so that you know it will actually cover all use cases rather than just the ones you check? In which case, if you're a coder, just... code.
It depends. If it’s for a throwaway tool in a language you don’t care about (looking at you Powershell) then no. If you’re learning a language then yes, absolutely, same as other types of coding autocomplete.
My experience as an experienced coder was that GitHub copilot was useful, but not transformative. It’s just a tool. Dunno whether the cost of it is worth it at a corporate level, that’s above my pay grade.
If you were told to use a word processing programme that changed every w to an m, you'd be right to say it's not fit for use, and we can't trust its output. That's more accurate than I'd trust GenAI to be.
Definitely. Cutting my comment down to the character limit rather lost that sense, but it was intended to be there.
The advantage people find is that it can direct their attention quickly to an efficient language technique that they were previously not sufficiently aware of. A fancy search tool.
Well, what ChatGPT is doing is basically indexing StackExchange, as you suggest, and then returning results in a more usable format. As I said, I prefer myself to just use a search engine. It disrupts my learning process less, and I'm deeply uncomfortable with the moral aspects of using it.
Comments
It's like autocorrect on steroids and occasionally LSD.
But yeah the lawyers who keep getting admonished for citing fictional cases is just awful. 🤦♀️
I find it very helpful, and it speeds some work up a lot, but to get the benefit I have to be really clear on what I need from it, and what good looks like. Then relentlessly and iteratively understand and evaluate what it does. Not a replacement for thinking.
This is the problem.
Probably some training is in order.
- IBM, 1979.
the *other* thing IBM was doing in 1979
I still don't use it. It can never give me right, or even close to right solutions.
You don't trust the ChatGPT response, but rigorously double-check that it works.
Me, I'd rather search. But sensible people find sensible uses for it.
The advantage people find is that it can direct their attention quickly to an efficient language technique that they were previously not sufficiently aware of. A fancy search tool.