I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
-
@iris @MCDuncanLab @sundogplanets
Does anybody have links for this? I'd love to know more
@hosford42 @MCDuncanLab @sundogplanets not sure if this is the one I read, but I think it covers the case I mentioned: https://thecybersecguru.com/news/openai-navier-stokes-proof/
-
@ravenbait @sundogplanets
Copied. The LLMs do not learn nor are they trained. It's an abuse of English by PR/Marketing.@raymaccarthy @ravenbait @sundogplanets That's a semantic argument. Her content was given to an "AI" bot that is more than just an LLM, and that content has been assimilated by that bot into its source material. It now knows what Prof. Lawler wrote. It can then freely use that without her permission.
-
I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
Guess I won't be asking that person for help or advice on anything anymore. Enshittification happens in people's writing skills too, I guess.
I also have absolutely no idea how to respond to them. If you're that far down the genAI black hole I don't think pointing out the ethical/environmental/cognitive issues will help. Fuck.
@sundogplanets @LJ How about telling them that if you wanted AI slop, you would've asked ChatGPT yourself?
-
@sundogplanets @ravenbait super cliché but "I statements" work wonders. They don't have to understand your reasoning, but they need to be able to respect your strong feelings in order to have a relationship with you.
@iris @sundogplanets @ravenbait
I love that specific set of feedback rules: I see that you do/did so-and-so. It made me feel such-and-such. I do not like that. I would like you to stop doing that.
-
I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
Guess I won't be asking that person for help or advice on anything anymore. Enshittification happens in people's writing skills too, I guess.
I also have absolutely no idea how to respond to them. If you're that far down the genAI black hole I don't think pointing out the ethical/environmental/cognitive issues will help. Fuck.
i would be shocked if your department had on retainer an IP lawyer, but if they do, give them the heads up that you will be referring the colleague to them for a TL;DR on how that’s an infringement on your team’s copyright and that they will be liable for any plagiarism that results from it.
which, of course, means that now y’all need to create a boilerplate for all emails and comminications about reserving all rights and not giving permission for LLM imput etc
-
I really appreciate all the great advice and suggestions. I'm just... so disappointed and know I have to approach all my interactions with this person differently in the future. I've written up a long email, but I'm realizing that won't work as well as having a conversation with the person. Especially because I'm likely to start crying during the conversation. So. That's going to be FUN.
But maybe it will help one person be less addicted to the forest-burning plagiarism machine? Maybe?
-
I really appreciate all the great advice and suggestions. I'm just... so disappointed and know I have to approach all my interactions with this person differently in the future. I've written up a long email, but I'm realizing that won't work as well as having a conversation with the person. Especially because I'm likely to start crying during the conversation. So. That's going to be FUN.
But maybe it will help one person be less addicted to the forest-burning plagiarism machine? Maybe?
@sundogplanets Do they know you even the slightest?
Passive agression has an undeserved bad rep. Either anti-AI stickers on tour laptop or videoconf background whereever you meet this person
For a bit more serious answer, flattery may work: “I was hoping for an answer based on your insights, blah blah”. Frankly, I don’t understand people who are so readily willing to replace themselves with a machine and probably wouldn’t bother
-
@hosford42 @MCDuncanLab @sundogplanets not sure if this is the one I read, but I think it covers the case I mentioned: https://thecybersecguru.com/news/openai-navier-stokes-proof/
@iris @MCDuncanLab @sundogplanets
Thanks!
I was hoping someone might have links about Anthropic specifically. OpenAI is already known to be a bad actor in this respect, but Anthropic's reputation is generally better. Someone I know insists it's safe to use Anthropic tools for research because of their ZDR policy. If there was documented misbehavior on this front, it would be a counterexample to point to.
-
@iris @sundogplanets @ravenbait
I love that specific set of feedback rules: I see that you do/did so-and-so. It made me feel such-and-such. I do not like that. I would like you to stop doing that.
@Marrekoo @sundogplanets @ravenbait it can work (if the other person cares!) in situations where you disagree on fundamentals: I realize this isn't something that bothers you, but this bothers me quite a bit, and I find it very hurtful that you would continue to do this [involving me somehow], knowing how I feel about it. It's the agree to disagree that still comes with action.
Which can make it a difficult approach to want to take: clearly I am right and you should be convinced, you should realize why I am right, I should not have to resort to couching it as you tiptoeing around my feelings. It can feel degrading to work out how to phrase it.
But sometimes it's the only thing I can find that works.
-
@iris @MCDuncanLab @sundogplanets
Thanks!
I was hoping someone might have links about Anthropic specifically. OpenAI is already known to be a bad actor in this respect, but Anthropic's reputation is generally better. Someone I know insists it's safe to use Anthropic tools for research because of their ZDR policy. If there was documented misbehavior on this front, it would be a counterexample to point to.
@hosford42 @MCDuncanLab @sundogplanets ah, my bad, I misremembered the roles of the players. I don't have a specific link for that.
-
@Marrekoo @sundogplanets @ravenbait it can work (if the other person cares!) in situations where you disagree on fundamentals: I realize this isn't something that bothers you, but this bothers me quite a bit, and I find it very hurtful that you would continue to do this [involving me somehow], knowing how I feel about it. It's the agree to disagree that still comes with action.
Which can make it a difficult approach to want to take: clearly I am right and you should be convinced, you should realize why I am right, I should not have to resort to couching it as you tiptoeing around my feelings. It can feel degrading to work out how to phrase it.
But sometimes it's the only thing I can find that works.
@iris @sundogplanets @ravenbait well said. On the other hand, it does offload stuff to the other party. The hardest thing to do is receiving feedback well. They might take it in to think about it, they might concede and try to change their behaviour, or they might shrug it off or refute it. That's on the other side to decide. And it makes it clear what to expect/invest in the future.
-
@ravenbait Yeah I will DEFINITELY point out that part. I have to think about how to say it politely so it doesn't just bounce off their brain.
I have to think about how to say it politely so it doesn't just bounce off their brain.
A non-threatening allegory can be effective, since there is almost always a consent framing that can connect with a person's life.
Since they're a scientist, maybe "if you asked me to review some numbers from (your project) and I forwarded them to (competing lab) to check, how would you feel?"
Alternately some folks connect with concrete, real examples, so you could cite the mathematicians whose work was visible to OpenAI's training, and then subsequently OpenAI beat them to publishing, using suspiciously identical phrasings.
-
@Marrekoo @sundogplanets @ravenbait it can work (if the other person cares!) in situations where you disagree on fundamentals: I realize this isn't something that bothers you, but this bothers me quite a bit, and I find it very hurtful that you would continue to do this [involving me somehow], knowing how I feel about it. It's the agree to disagree that still comes with action.
Which can make it a difficult approach to want to take: clearly I am right and you should be convinced, you should realize why I am right, I should not have to resort to couching it as you tiptoeing around my feelings. It can feel degrading to work out how to phrase it.
But sometimes it's the only thing I can find that works.
@iris @Marrekoo @sundogplanets @ravenbait
I would try to avoid sending anything to him like the plague.
But if this can't be avoided, your answer seems very sensible.
If he even bothers to read the message, of course.
-
@hosford42 @MCDuncanLab @sundogplanets not sure if this is the one I read, but I think it covers the case I mentioned: https://thecybersecguru.com/news/openai-navier-stokes-proof/
@iris @hosford42 There's even a wikipedia page on it now https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_priority_controversy
-
@kim @frincon @sundogplanets I don't have that kind of access, unfortunately.
@ravenbait @kim @frincon @sundogplanets
It is very likely that kind of people is feeding all their email into an AI and only reading the briefs the AI generates for them, and skimming the answers. So the effect of that kind of warning is probably very limited.
But it can't do no harm.
-
I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
Guess I won't be asking that person for help or advice on anything anymore. Enshittification happens in people's writing skills too, I guess.
I also have absolutely no idea how to respond to them. If you're that far down the genAI black hole I don't think pointing out the ethical/environmental/cognitive issues will help. Fuck.
Crap. Unfortunately I have had to do the same with my friends and editors. People I once trusted use AI now, and I can't even send it to entire domains (I.e. gmail) because I know that it all gets scanned into the AI. I wish I had answers but I don't...
-
I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
Guess I won't be asking that person for help or advice on anything anymore. Enshittification happens in people's writing skills too, I guess.
I also have absolutely no idea how to respond to them. If you're that far down the genAI black hole I don't think pointing out the ethical/environmental/cognitive issues will help. Fuck.
@sundogplanets Just be honest with them. Say something like “I asked you for YOUR personal opinion and critique, not a llm. If I really wanted this response I could have done it myself (but there are multiple reasons why I WILL NOT). Do you not have the time or respect for me and my work or myself to not read what I wrote and respond with your own words and feedback?”
Sorry, I couldn’t help but make unnecessary suggestions
.
I’m also too blunt with people when they cross me. -
i would be shocked if your department had on retainer an IP lawyer, but if they do, give them the heads up that you will be referring the colleague to them for a TL;DR on how that’s an infringement on your team’s copyright and that they will be liable for any plagiarism that results from it.
which, of course, means that now y’all need to create a boilerplate for all emails and comminications about reserving all rights and not giving permission for LLM imput etc
@blogdiva @sundogplanets this is brilliant. "I'm sorry, I really really hate to do this but I had to ask the lawyers and here's what they said" is a power move. Tenure doesn't shield against a legal advisor's recommendations.
-
I asked a trusted colleague for comments on a written document and they fed it in to chatGPT and sent me back that output and I just want to cry.
Guess I won't be asking that person for help or advice on anything anymore. Enshittification happens in people's writing skills too, I guess.
I also have absolutely no idea how to respond to them. If you're that far down the genAI black hole I don't think pointing out the ethical/environmental/cognitive issues will help. Fuck.
@sundogplanets "I asked for YOUR input, not ChatGPT's" would be an appropriate response. Not sure if it's worth it though.
So sorry. That really has to hurt. I'd feel betrayed.
-
@sundogplanets @LJ How about telling them that if you wanted AI slop, you would've asked ChatGPT yourself?
He probably feed it to a specifically tailored GPT, in order to make sure that GPT reflects his knowledge and points of view, and they reviewed it later, yada, yada, yada.
He probably is very proud of his achievements configuring that GPT, and maybe even get offended by anybody implying that the answer is the same as asking a bare ChatGPT prompt for it.
Crazy people says crazy things.
️
