Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
echinopsis
Apr 13, 2004

by Fluffdaddy

rotor posted:

idk about yall but at my work the solution-in-search-of-a-problem thing with AI is so much stronger than it was with crypto. Like we kinda halfassedly poked at crypto because idk why but we got dudes goin fuckin apeshit over AI and like no one knows where it fits in the product line but people are fuckin comitted to finding a place for it

brb buying crimecomitter.com

Adbot
ADBOT LOVES YOU

Moongrave
Jun 19, 2004

Finally Living Rent Free

rotor posted:

idk about yall but at my work the solution-in-search-of-a-problem thing with AI is so much stronger than it was with crypto. Like we kinda halfassedly poked at crypto because idk why but we got dudes goin fuckin apeshit over AI and like no one knows where it fits in the product line but people are fuckin comitted to finding a place for it

does your work do anything with anime tits

distortion park
Apr 25, 2011


MononcQc posted:

I imagine some hyped up customer calls a salesman and says “what’s your ai strategy?” and there’s none so the whole org is turned upside down to wedge it wherever and have an answer to it, then everyone can rush to do the first press release of the market segment maybe.

We did this, our customer's are so hyped by the "AI" page on our website with mockups and stuff. It's right at the bottom of the product team's priorities though so they're going to be waiting a while!

rotor
Jun 11, 2001

Official Carrier of the Neil Bush Torch

BARONS CYBER SKULL posted:

does your work do anything with anime tits

no :-(

echinopsis
Apr 13, 2004

by Fluffdaddy
AI is going to ruin the world

Cybernetic Vermin
Apr 18, 2005

relentless hype about ai more specifically.

echinopsis
Apr 13, 2004

by Fluffdaddy
AI will render all virgins useless

echinopsis
Apr 13, 2004

by Fluffdaddy
https://afronomist.com/ibm-will-lay-off-thousands-of-employees-their-work-will-be-taken-over-by-artificial-intelligence/


thousand of people out of 8 billion? barely worth caring about

Cybernetic Vermin
Apr 18, 2005

ibm is just firing people while using overblown ai hype to play pretend with shareholders about it.

echinopsis
Apr 13, 2004

by Fluffdaddy
AI proves useful

Qtotonibudinibudet
Nov 7, 2011



Omich poluyobok, skazhi ty narkoman? ya prosto tozhe gde to tam zhivu, mogli by vmeste uyobyvat' narkotiki
i was amused when we tried to stuff AI into our products, because unpredictable and unexplainable behavior is exactly what you want in your network infrastructure

Moongrave
Jun 19, 2004

Finally Living Rent Free

echinopsis posted:

AI is going to ruin the world

can't be any worse than it is now tbh, i say let the AI have a turn

abigserve
Sep 13, 2009

this is a better avatar than what I had before
GPT has automated away two manual processes at my work and it's really only the beginning, I think a lot of people are underestimating the amount of paper pushing that gets done even in relatively lean enterprises and GPT at least is almost impossibly good at doing stuff when you provide it context. Like you can feed it a knowledge base article and be like "summarize this" and it does it perfectly, better than probably 90% of the english speaking population, and then you realize well what if I gave it like...50 articles to summarize, at once? and it works perfectly. It's fuckin insane

Jabor
Jul 16, 2010

#1 Loser at SpaceChem
right up until it does it wrong and nobody notices

Beeftweeter
Jun 28, 2005

OFFICIAL #1 GNOME FAN
lol



yeah, i think letting something without access to realtime info pick stocks seems like a great idea

Qtotonibudinibudet
Nov 7, 2011



Omich poluyobok, skazhi ty narkoman? ya prosto tozhe gde to tam zhivu, mogli by vmeste uyobyvat' narkotiki

abigserve posted:

GPT has automated away two manual processes at my work and it's really only the beginning, I think a lot of people are underestimating the amount of paper pushing that gets done even in relatively lean enterprises and GPT at least is almost impossibly good at doing stuff when you provide it context. Like you can feed it a knowledge base article and be like "summarize this" and it does it perfectly, better than probably 90% of the english speaking population, and then you realize well what if I gave it like...50 articles to summarize, at once? and it works perfectly. It's fuckin insane

lots of people being poo poo at writing is nothing new

i look forward to our glorious future of lots of people being bad at GPT wrangling

mediaphage
Mar 22, 2007

Excuse me, pardon me, sheer perfection coming through
yeah to be honest i think it’s less that gpts are good at writing than it is most people are very bad at it

honestly just look at 95% of the self published stuff on kindle, gpt is easily as good

distortion park
Apr 25, 2011


mediaphage posted:

yeah to be honest i think it’s less that gpts are good at writing than it is most people are very bad at it

honestly just look at 95% of the self published stuff on kindle, gpt is easily as good

most of the things we read and media we consume is pretty good, like it's stuff made by people doing what they're best at and recommended by others. The vast majority of writing done by people is pretty poo poo and often badly misinformed, and not read by many people. im not worried about these models being better than me at things I'm good at but it's already better than me (and certainly quicker) at tasks I'm bad at.

MononcQc
May 29, 2007

it is simultaneously true that ChatGPT has allowed people to write better than they did and that editors out there are flooded by more garbage than ever before and are losing ways to cope with the noise because a lot of it is still poo poo.

Salt Fish
Sep 11, 2003

Make Deer Great Again
Cybernetic Crumb

MononcQc posted:

ChatGPT has allowed people to write better than they did

[citation needed]

abigserve
Sep 13, 2009

this is a better avatar than what I had before
I mean having good written communication skills has carried me far further than my technical acumen and I can see the writing on the wall. Its just not going to be a thing that you need people to build like, a readable and nice slide deck, or a product brochure, or a high level design, etc etc

Instead you'll just need the technical people banging out the bullet points

To take it further you also don't really need as many sales people when you can have a thing that has access to all the information, can recall it instantly and carry a conversation with someone

MononcQc
May 29, 2007

Salt Fish posted:

[citation needed]

some people are plain terrible at writing and expressing ideas, and just getting correct grammar is a major improvement.

Salt Fish
Sep 11, 2003

Make Deer Great Again
Cybernetic Crumb
The text that you're talking about, filler text that just has to be grammatically correct, isn't something anyone wants to read. It's the 15,000 words that accompanies a video game tip on an SEO farm. Just get rid of it entirely instead of automating it.

If you're working on a brochure that isn't worth having a human write just don't create it. Having a machine produce a lexical illusion of content isn't a real solution to that problem.

Salt Fish
Sep 11, 2003

Make Deer Great Again
Cybernetic Crumb
I remember seeing this same exact conversation by the way (in my view/opinion its okay if you don't agree its the same) with facial recognition. Wow, think about how cool it will be to remember who went to that party because facebook can tag my friends! Oh what if my phone gets a text when my door bell sees grandma outside!

Oh actually 99.9% of the use cases are for the panopticon, police, and totalitarian governments. I got asked to take off my mask at my office building because they use it to track when you show up for work (yes really).

Qtotonibudinibudet
Nov 7, 2011



Omich poluyobok, skazhi ty narkoman? ya prosto tozhe gde to tam zhivu, mogli by vmeste uyobyvat' narkotiki

abigserve posted:

To take it further you also don't really need as many sales people when you can have a thing that has access to all the information, can recall it instantly and carry a conversation with someone

it'll probably be used in self-service funnels but i doubt salespeople are going anywhere. you fly them out to prospect sites because people like talking with people, and because GPT has no capacity to emotionally read someone

rotor
Jun 11, 2001

Official Carrier of the Neil Bush Torch

Salt Fish posted:

. I got asked to take off my mask at my office building because they use it to track when you show up for work (yes really).

jesus christ

armpit_enjoyer
Jan 25, 2023

my god. it's full of posts

MononcQc posted:

it is simultaneously true that ChatGPT has allowed people to write better than they did and that editors out there are flooded by more garbage than ever before and are losing ways to cope with the noise because a lot of it is still poo poo.

This! I work as an editor; now I have to interrogate authors whether or not they use ChatGPT, and then I have to waste time fact checking every single thing they write because if someone's stupid enough to submit AI generated work to me, they're also stupid enough to lie about it.

And then there's these two fuckwits:

https://twitter.com/DulwichQuantum/status/1654066665678688256

Moongrave
Jun 19, 2004

Finally Living Rent Free

Salt Fish posted:

The text that you're talking about, filler text that just has to be grammatically correct, isn't something anyone wants to read. It's the 15,000 words that accompanies a video game tip on an SEO farm. Just get rid of it entirely instead of automating it.

If you're working on a brochure that isn't worth having a human write just don't create it. Having a machine produce a lexical illusion of content isn't a real solution to that problem.

it's very good for non/barely english speakers to be able to communicate effectively with native speakers but hey throw them away too huh

r u ready to WALK
Sep 29, 2001

armpit_enjoyer posted:

And then there's these two fuckwits:

oh this has to be one of those videos i've heard of where they use gpt, deepfake and tortoise-tts to make famous people say funny random bullshit

echinopsis
Apr 13, 2004

by Fluffdaddy

r u ready to WALK posted:

oh this has to be one of those videos i've heard of where they use gpt, deepfake and tortoise-tts to make famous people say funny random bullshit

worth every penny

MononcQc
May 29, 2007

Salt Fish posted:

The text that you're talking about, filler text that just has to be grammatically correct, isn't something anyone wants to read. It's the 15,000 words that accompanies a video game tip on an SEO farm. Just get rid of it entirely instead of automating it.

If you're working on a brochure that isn't worth having a human write just don't create it. Having a machine produce a lexical illusion of content isn't a real solution to that problem.

Yes that's quite aligned with my original point:

MononcQc posted:

it is simultaneously true that ChatGPT has allowed people to write better than they did and that editors out there are flooded by more garbage than ever before and are losing ways to cope with the noise because a lot of it is still poo poo.

All the biggest adoption I've seen at work is sales people, and I think it can work moderately well for them so long as few reps do it for cold calls equivalents because the garbage is targeted at few people so quick generation still gives you a moderate amount of 1:1 filtering on the receiver end. But anything that's intended for broadcast (write once, publish for many people) is gonna be an absolutely worse dump.

MononcQc fucked around with this message at 13:33 on May 6, 2023

MononcQc
May 29, 2007

armpit_enjoyer posted:

This! I work as an editor; now I have to interrogate authors whether or not they use ChatGPT, and then I have to waste time fact checking every single thing they write because if someone's stupid enough to submit AI generated work to me, they're also stupid enough to lie about it.

I have a good relationship with the editor we have at work for technical writing, and she's rightfully very afraid of what it means for her role to have ChatGPT around, in a small part because of how many places are firing their staff related to anything written, but also in terms of how much de-valued work they're going to be facing around having SEO-sounding redundant article to sift through and correct all the time.

A significant part of these generated texts have a decent written form and need more technical fact checking, which also fucks with the dynamic of the job she has, I think.

A bunch of our design staff is also kind of worried because of things like people feeding prior art from them and just generating visuals and designs based on their personal style (which they bring to work) and making their skill less distinctive. Like part of it is not just "oh it's more efficient", it's also the whole "we don't have to choose what is worth putting a good designer's time on, just generate all sorts of graphics everywhere and churn them on all the things indistinctly" that comes with de-valuing the work itself.

Salt Fish
Sep 11, 2003

Make Deer Great Again
Cybernetic Crumb

BARONS CYBER SKULL posted:

it's very good for non/barely english speakers to be able to communicate effectively with native speakers but hey throw them away too huh


Prosopagnosia affects an estimated 1.5 million Americans and facial recognition is going to help these people navigate the world in a way that was never possible before. Therefore I would like to thank the brave progressive police forces around the world for giving Amazon millions of dollars to setup rekognition API access in reaper drones.

Qtotonibudinibudet
Nov 7, 2011



Omich poluyobok, skazhi ty narkoman? ya prosto tozhe gde to tam zhivu, mogli by vmeste uyobyvat' narkotiki

armpit_enjoyer posted:

This! I work as an editor; now I have to interrogate authors whether or not they use ChatGPT, and then I have to waste time fact checking every single thing they write because if someone's stupid enough to submit AI generated work to me, they're also stupid enough to lie about it.

just gonna use ChatGPT to write my "i am not using ChatGPT" response. checkmate

what do you do in practice, ask them to write an affirmation that they aren't using ChatGPT that creatively uses the N word to see if the response comes back "As an AI language model..."?

armpit_enjoyer
Jan 25, 2023

my god. it's full of posts
What I actually do is whenever a client fucks up I add them to the blacklist me and a couple of people I know share. It's reactive, but I can't think of a good proactive solution atm

Beeftweeter
Jun 28, 2005

OFFICIAL #1 GNOME FAN
i don't think there is a good way of determining that and i think its going to probably pretty rapidly get harder to distinguish from human-written stuff

imo GPT text is pretty recognizable if it's longer than a paragraph or so (the most obvious tell is that it tends to repeat itself) but if it's just little snippets added to a larger whole then i don't know how you'd detect it immediately. then again if someone is going to use it to generate a bunch of text for publication then i guess they might also be lazy enough to just let it run and copy it wholesale

outhole surfer
Mar 18, 2003

editors are now forced to read and understand the material that passes through their hands well enough that they can call bullshit on it?

this sounds like a good thing?

Beeftweeter
Jun 28, 2005

OFFICIAL #1 GNOME FAN

nudgenudgetilt posted:

editors are now forced to read and understand the material that passes through their hands well enough that they can call bullshit on it?

this sounds like a good thing?

yeah i don't disagree really, that is literally their job

i think authors representing the work as their own is a bigger problem. nobody likes a plagiarist and i doubt many publishers want to be associated with someone that demonstrably has no idea what they supposedly wrote about

outhole surfer
Mar 18, 2003

Beeftweeter posted:

yeah i don't disagree really, that is literally their job

i think authors representing the work as their own is a bigger problem. nobody likes a plagiarist and i doubt many publishers want to be associated with someone that demonstrably has no idea what they supposedly wrote about

what are the existing safeguards against plagiarism? the fact there's a new source that can be plagiarized from doesn't seem like it should affect how plagiarism is dealt with. do editors not converse with the author about the material enough to realize the author doesn't understand what they "wrote"?

Adbot
ADBOT LOVES YOU

Beeftweeter
Jun 28, 2005

OFFICIAL #1 GNOME FAN

nudgenudgetilt posted:

what are the existing safeguards against plagiarism? the fact there's a new source that can be plagiarized from doesn't seem like it should affect how plagiarism is dealt with. do editors not converse with the author about the material enough to realize the author doesn't understand what they "wrote"?

you'd think so but apparently not. the example that keeps coming to mind is (unfortunately) naomi wolf: https://www.bbc.com/news/entertainment-arts-50153743

quote:

But during a BBC radio interview in May, it came to light that the author had misunderstood key 19th Century English legal terms within the book.

Publisher Houghton Mifflin Harcourt said their parting was "amicable".

Following the BBC radio interview, Wolf admitted there were "misinterpretations" in her book.

Her UK publisher, Virago, had already published the book by the time the interview was broadcast, but said it would make "necessary corrections" to future reprints.

However, US publisher Houghton Mifflin Harcourt delayed publication, and has now cancelled it altogether, according to the New York Times.

she wrote an entire book around her misunderstanding of a legal term, and the publisher didn't catch it. when she started doing interviews it became clear that she had no idea what she was talking about

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply