I am old enough, that personal relations was key in getting information and business reviews, trust. So these days needed a service (business legal related) done. Via AI called some links they had, they could not do it. So last, used old fashion way (in JP/TW)of asking people in our building (big office tower), and yes got what we needed. So sometimes old neighborhood connections helps a lot in info and trust, personal connection. I will give back small gift!, but they did some hard searching just on relations.
It can not do anything competently.
Some industries deliberately made the entire process opaque and difficult for AI to help the consumers. Most services that require a middleman acting as agents probably fit the description. That doesn’t mean AI can’t help the agents themselves who have most of the data available to them.
Still cannot give you a hug. Or a surreptitious fondle in the elevator…
Depends on what is needed, how it is prompted, and who is judging the product
For simple things with a decent prompt and a non expert observer, it must be magical and terrifying
For someone with a range of expert knowledges including computers, it looks like a good start
Mathematicians and data entry clerks alike are looking at a future where the computer does it better in most ways, except of course for creativity
New Brunswick just released an educational document that hallucinates citations. I have also asked a paid AI to look for references for me, but when I checked I already had them or they were fake. Waste of time to verify that real sounding papers from known journals were not real. I can’t trust it to write anything. It can help with advice for cutting word length, but the advice often includes cutting key passages from an argument which is because it doesn’t actually understand what it sees
I’m going to submit a rejected paper to a better journal (scope is better fit, wasn’t my first choice because of a 6 month average turnaround). This means I need to completely change references from APA to Harvard. A mechanical task that a computer should be good at, but I won’t waste a moment trying that one with AI. Safe to assume failure.
TLDR; still sucks at formatting
that sounds like a job for EndNote. Pretty good at citation formatting and changes…
I’m running AI policy for my university, and it does indeed suck (GenAI, not my policy, thank you!). Plus, so many ways that AI will completely change the university experience: even the need for a university, at least for undergraduate degrees. What will the value of a university education be when AI can teach you pretty much anything?
already people can see that coding as a career is basically dead and so are most Comp Sci degrees… (but then who is around to check the AI slop that Github Copilot spits out?). And MIT just debated the value of their entire mathematics faculty: OpenAI just dumped out hundreds of proofs of open questions. (Not that we are certain that they are correct, so there is still SOME work for math people to do, I guess.)
AI can make passable music, novels, artworks. Even poetry is not immune (but the product is usually inane and monotonous, although the same can be said for many actual human poets). But that’s the point: passable, not creative, and never brilliant.
and No, I definitely do not want Copilot to summarise a document that I AM CURRENTLY WRITING MYSELF VERY WELL THANK YOU! Fuck off out of my life, you dumb bag of logic gate transactions.
For the business and industry I’m in it’s fairly useless so far, because we deal with complex expensive solutions that have almost no margin for error .
The AI fucked up is not an answer we can provide. Internally or externally. We are also in a regulated environment, and we aren’t even permitted to use most AI providers.
I really want AI to work for me and make things more efficient because despite appearances on here I have a busy job 
.
I use it about once a day at most at present, that could go up obviously sooner than later.
https://x.com/iamchrizten/status/2110122233133748282/video/1?s=49
Could just be a fake tikkytocky
Still funny .
I do use it weekly, but mostly I chuck a bunch of papers into an agent and tell it to do this and that to it, and then crosscheck the output against the facts. Kinda like a RAG (resource augmented GenAI). There’s a hell of a prompt for that, kept in my system prompts in Claude or Copilot (we have an internal Copilot implementation so that’s used for sensitive data to stay protected) and i do check the output myself (and almost all is rewritten by me) but it does a reasonable job of frameworking big presentations for me. None of that is life threatening or commercially sensitive/contract defining, so I’m cool with that.
… but i do get tired of reading countless emails written by generic AI for no frikken reason (one good reason I am happy with is from ESL writers who struggle, and most of our staff at junior levels are like that). But for me, it’s way faster to write the response myself than to get a bot to do it in most cases.
I tried Mendelay when it was new in like 2010 or 2011, was very unreliable citations and I didn’t like the way it crawled through my folders and organized because my stuff was already well organized. In education and linguistics APA is pretty standard so it is a rare problem. I don’t know if the current year EndNote can take a finished paper and just fix all the references or if I need to install an app and go through a bunch of nonsense which might be better for a new paper but in this case will take more time than doing it myself (as I expect an AI would)
Can AI come up with the questions yet?
Edit: a thing about proofs, I had to learn to write math proofs as part of my computer science degree, it was a basic thing leading to truth tables and then logic gates. Some of us learned logic with exams before AI was a thing, but math proof logic is something that it makes sense an AI would be able to do. There are fixed rules, and it can brute force thousands of possibilities quickly to see where they lead.
It’s interesting, because when I see people online rely on AI for basic logic stuff I learned in my first year, I see that as a sign that they aren’t particularly intelligent. But I tried using AI to cut a paper down by thousands of words, and it would add conclusions to sections that not only weren’t in the original prose but didn’t logically follow from what was. A first year student or an older adult over their head I would expect it from. Not an advanced intelligence.