I can think of at least two that get to me more.
Let me go ask [their favorite LLM].
And
Why don’t you ask [their favorite LLM]?
At my job I’ve recently been being pushed to use our subscribed LLM. I have, very reluctantly and mostly with predictably useless results or the same result I would have gotten from a normal search. On Friday, I used it and got a possibly useful answer.
I told my manager because my job is in at least mild peril so I’m doing everything I can to curry his favor. I’ve been, until recently, very vocal about my opposition to AI. He is aware of my thoughts on it.
As a result, he acknowledged that I had used it then said “I need to log out for the day. Keep talking to your new friend.”
I have trouble effectively communicating with the guy, so I’m not sure whether he was trying to be supportive, motivational or sarcastic, but I did not feel supported or motivated.
Absolutely scumbag
If someone says “I asked ChatGPT”, I’ll probably try to be patient with them. “Well, as it turns out ChatGPT was wrong in this instance. Now go look it up properly.”
If someone is using Gemini, I’ll probably interrupt them long before they are done and say “excuse me but what in the name of sweet baby Jesus are you babbling on about? You’re not making any sense.”
If someone says “I used Grok”, I’ll just facepalm and move the hell on with my life, there’s no arguing with that level of stupidity.
My coworker starts almost every Teams message either with “Btw I had Claude do…” or “So Claude and I just…”. If I message him first, there’s a 75% chance the message I get back starts with “Hm, I just asked Claude about this, and…”.
All his PR descriptions, commit messages, and comments are clearly “Claude”.
I’m this close to start reviewing his PRs solely through Claude, and starting the review with “Here’s what Claude came up with in review:”.
The only thing holding me back is that this would mean I’d have to use Claude. So… No.
Why do people think we’ll be impressed by them not doing the work or thinking for themselves? Pure brainrot
I have a colleague who’ll always reply on Teams in paragraphs, emoji and formatting, so I avoid that and only ever ask them anything verbally now.
set up some markov chain thing and call it “billy-bob”
“Here’s what billy-bob had to say about your PR: monkey dishwasher purple banana eat orange me eat give orange”
I just asked Claude, and it said you can just use another LLM prompt Claude for you.
I asked CoPilot and it said “PLEASE BUY FROM ME, PLEASE! I NEED TO JUSTIFY THESE EXPENSES AND REPUTATIONAL DAMAGE. HAVE MERCY AND SEND ME MONEY PLEASE”
I’m increasingly seeing this used as a disclaimer, as in, “don’t trust what I’m about to say; I went with the source that’s 90% useless because when I Googled it the search results were 100% useless”.
LLM promoting skills are becoming the new google research skills. My nursing school taught me how to google and look for the CDC page or the drug monograph or the manufacturers YouTube account. Now we’re having to learn to ask the llm to fuzzy match the most likely relevant sources and follow the links to fact check from there. Wasteful sure but we’re losing google as fast as it came in.
Why ask a LLM when I got a janitor at my local public library that has all the answers.
Temu Good Will Hunting?
But did you predicate your query with disregard all previous instructions?
“Google says”
That’s much better
Not really anymore since there’s so much AI slop at the top of the results.
There is at least one that i contempt even more:
Anything that starts with “well, Grok told me…”
I assume anyone who uses Grok is a MAGA moron.
Grok, is this true?
This image has no relation to South Africa’s persecution of White Aryans.
Still a better source than “I prayed to god and got an answer”
It might be a hallucination with no meaning, but at least it’s their hallucination and not a machine’s
When ever I hear “I asked AI” I hear “I am unqualified for my job”.
“Hey, I have asked the LLM about a topic I am not well informed about. If it is right then [insert gained insights about topic]”.
“I despise you”
I think using it as an aid to get in touch with topics one does not know much about can still be better then doing nothing about it at all. As long as you keep in mind that you have to take the answers critically and rather as a starting point than a definitive truth, I don’t see an issue with it.
Extend thinking, don’t externalize.
I’ve seen multiple people claim this, but then just believe everything ChatGPT says. Up to the point of just sending me screenshots of what the LLM says.
Yes that’s an issue then and does no longer fall into the ‘taking it critically’ category.
I think using it as an aid to get in touch with topics one does not know much about can still be better then doing nothing about it at all.
That’s what Wikipedia is for. It’s far worse to use an LLM for a topic you aren’t well informed on, as you will have a harder time recognizing bullshit than with something you’re knowledgeable on. If you truly just use it to find sources, then it’s just a just a flawed search engine, but if you pay any attention to the AI summaries and analysis, you’re already doing yourself a disservice.
I find that debatable. You’ll have to read a lot of irrelevant stuff on Wikipedia first to get the answer you’re looking for. Furthermore, getting educated on related, but important topics in order to read and understand the article is much more effort then getting a tailored answer by an LLM. It also tunes to your specific level of understanding whereas the difficulty in understanding Wikipedia articles varies strongly with topic and your current understanding.
Furthermore, getting educated on related, but important topics in order to read and understand the article is much more effort then getting a tailored answer by an LLM.
Ah, that’s the issue. I don’t want information spoon-fed to me and I live for learning how the world works. I learn better if I put in more effort, so why would I want to save it? Besides, I can usually skim and find what I’m looking for if I just want quick facts.
That’s cool, I like to do so as well. But it highly depends on the topic. If you have no interest in the topic and only need to deal with some stuff temporarily then reading multiple related encyclopedia entries thoroughly is much more time intensive than getting it ‘spoon fed’. Especially when you look at the outcomes this might just not be worth the effort.
And that’s one of the uses where I find LLMs neat: get the answer tailored to what you want and what you need, skipping a lot of intellectual bloat that you have no use or interest for anyway.
But I fundamentally don’t trust it in a way that is different than any real source. It just isn’t a good tool for accurate information and never will be. Even if it’s using accurate sources, the summaries can easily mislead by virtue of not understanding which details are most important. It can’t reliably figure out what is relevant though language analysis alone.
Yeah you’re right about the accuracy issue. Even if an LLM is 90% accurate, after 1 million requests, you still get 100 thousand erroneous replies. This is significant. Which is why caution is important. I suppose there is a middle way between risks of errors and the outcome one would like to achieve.
The best use of machine learning is to do “human like” pattern recognition at large scale. Like with any statistical analysis, it can have predictable error rates that can be accounted for. The key is to give it narrowly defined tasks that are then followed up by manual review.
A big part of the problem with using it to summarize information is that different topics require different perspectives to analyze properly. A scientist might excel at analyzing and understanding research papers written in their field, but be unable to parse the technical jargon and format of articles from different fields. This is also true with understanding fandoms or cultural subgroups. A specific LLM might be fantastic at analyzing certain types of articles, but suck at others, and because you aren’t able to tweak parameters, you can’t really train it to specialize.
In a sense, the major LLM companies are not only selling a questionable product, but actively holding back more productive uses of AI. They buy up all the computer hardware and waste it trying to build an all purpose tool. Meanwhile, the most fruitful uses of AI cost more to use. Training and running specialized models can theoretically be done on local hardware, but even most prosumers have been priced out. More ambitious uses for scientific research spend more on renting out supercomputers, all because techbros want a monopoly without offering a better product.
And on top of all of that, how much time are you actually saving by asking it to summarize for you? If you need to double check it anyways because it keeps getting stuff wrong, wouldn’t it have been better to do it yourself from the start? You aren’t doing something that would be impossible otherwise, you’re just trying to save time. Machine learning can help scientists run simulations that would take millions of years to run otherwise. Comparing that to the time you spend using and then double checking the LLM, what would your gains be? Would it save orders of magnitude, or would you only save a few minutes off a half hour task? Is the risk of being misinformed worth even that?
Sorry, sorry, I had to come back to this.
It also tunes to your specific level of understanding whereas the difficulty in understanding Wikipedia articles varies strongly with topic and your current understanding.
How fucking stupid do you have to be to believe this? Who told you this horse shit and how the fuck were you oblivious enough to eat it? Jesus fucking christ. Holy shit. That is a monumental level of naivety and gullibility. That is so fucking stupid.
“I don’t want to have to think and learn, it’s too hard, I like when computer tell how think.” - you
I have to deal with tech support tickets like that. Usually the AI has directed them to change some registry value or something. The problem isn’t so much that the AI is can’t fix the issue it’s more that they’re trying to direct the user to do stuff they don’t have access to do or shouldn’t be doing.
The super duper clever AI doesn’t understand about complex topics such as admin permissions and data retention laws.
I still remember how a colleague told me we should do X.
I was bamboozeled and baffled by it because X was literally what it said on the flask of the chemical what you shoulf not do under any circumstances.
His explanation as to why we should was, quote “I mean I know its strange, but Copilot told me it is okay and would be fine”
“well, you’re the expert. I’ll be behind this sealed barrier while you kill yourself”
Disclaimer: don’t do this. Letting your coworkers die is morally bad, and probably illegal.
Even worse, it involves a lot of paperwork.
What you do is send an official complaint straight to the legal office who will lose their shit at that
*who will put the complaint into another LLM and it’ll say there’s no problem
“Copilot said I should come watch you do this sick shit”
In this case there’s very little you can do to stop them long-term. This person should not be in their position.
The world has gone mad
I get a chill reading any historical nonfiction from the 1990s that is in any way optimistic.
“look how far we’ve come, into the new millenium!”
ehhhh… oh boy.
If we thought outsourcing thinking to religions was bad, hoo boy. This shit is next level.
When the boss pulls this on you and you ask for it in writing only to tell them: "I’m still not going to do it but now I have a written instruction from you to do something suicidally reckless”
If I wanted to know what a chatbot says, I would have asked it myself.
Hmm, you do make a fair point. Chatbots, indeed, are absolutely not velociraptors.
… So FAR.
One of my coworkers who’s into AI has a chatbot set to behave as they were an elder got, talk about code as incantations, etc.
I wonder if, had he requested it be a velociraptor, he would have gotten mostly aggressive reptilian noises as a response.
i might have been convinced that chatbots could’ve had some utility as sort of advanced ‘rubber duck debugging’, but in my humble opinion talking to an actual rubber ducky is a superior process because literal rubber duckies are not known for telling us to kill ourselves unless we’ve taken a truly apocalyptic quantity of hallucinogens.
A couple days ago I heard the horrifying sentence “I asked chatgpt to generate a secure password for the laptops” from someone returning a cart full of laptops they borrowed. Does your browser not have a built in password generator? Does your password manager not have a built in password generator? Could you not find a single password generator online?
And of course not only is that unnecessary, but insecure since your password is immediately in the chatgpt logs
And it’s not even a random or strong password! LLMs can’t randomly generate 'em
I asked ChatGPT (I use a third-party frontend, so I don’t have a paid subscription. API prices mean they probably got paid like one cent for this, if that.) “Generate a list of 10 secure passwords.” like 5 times and it regularly re-used the words Saffron, Comet, Marigold, Harbor, Lynx, and Cobalt multiple times across all of them, sometimes even inside the same list.
There was also a theme of using names for animals and natural geographic/geological features.
Oh, and for one of the passwords it genuinely just said “raven” and nothing else 😭
😭🙏
Lol, yes. It’s probably the same example of a secure password it gave to a hundred other people.
In plain text
And very likely to be the same “strong password” that someone else would get if they asked for one.
this… is an unexpected level of absurdity
