I think I screwed up by using ChatGPT too vigorously at my freelance work

But they may decide that since it can be done much faster than they thought, they are paying way too much.

A lot of people have - and will have for a long time - a vague feeling that AI-assisted work isn’t “real” work. It shouldn’t matter to the client how the work is getting done so long as it’s done correctly, but it will matter.

But I don’t think that’s necessarily what’s happening here.

If a transcriber isn’t actually listening to the material but is instead relying on a series of mostly-but-not-always reliable AI-assisted tools, I think there’s a good case to make that they’re not doing the job they’ve been paid for.

If I hire you to transcribe something of mine, I do indeed expect you to listen to the whole thing. That’s the point of a transcription. Filtering it through computer programs is unlikely to give me an actual word-for-word transcription, right? And if it’s not word-for-word, it’s not a transcription. It’s notes.

And these are sermons meant to deliver specific messages, so I can understand why they’d want to make sure it’s verbatim.

If OP is listening to the sermons afterwards and reading along with the finished work to ensure accuracy, then we’re back to the bit at the top of my post.

Blessed are the cheesemakers?

It’s a Christian group, so maybe they are concerned you are using methods that some might consider…unnatural?

That’s my assumption for why they’re asking. The suspiciously fast job rate makes them concerned that they’re not getting the verbatim transcripts, properly formated for grammar and readability, they’re paying for. Short of them combing through all of the returned documents (which defeats the point of outsourcing it) asking about how they’re coming back so fast seems reasonable.

Just tell them that there are computerized tools to make the process more efficient, and that due to your experience, you know how to use these tools effectively. You don’t need to tell them the details. And it’s true: Even if you plunked me down in front of a terminal with ChatGPT and a YouTube channel, I don’t think I’d know what to do to get a transcript out of it. I could probably figure it out, but I haven’t, yet. You have, and that’s worth something.

If you feel you have to be more specific, tell them you trained as a court reporter/stenographer - they easily type 200 wpm. Many go much faster than that - 300 wpm. The steno machines are computerized nowadays. I did that training myself, and so did another SDMB member who worked as a professional court reporter (I did not continue with it due to arthritis).

Before answering any of their questions, and especally before answering with made up bullshit lies as some have suggested, you should ask them what their actual concerns are.

Do you like my quality? if not, how can I improve it? What specifically is wrong with any example you’re unhappy with?
Do you like my responsiveness to your needs? If not, what can I do differently?
Do you like my price vs results? If not, what can I/we change to make you happy?

Notice in each case you’re NOT questioning the validity of their tentative conclusion that something is wrong. You’re asking for feedback to address their concerns directly with changes you can (theoretically) implement. That’s good conversational and negotiating tactics. And will make them feel better about you and your work. I suspect they’ve got some vague feeling something is wrong. Not an actual logical problem, just a feeling. You being forthcoming and open will assuage that feeling.

If at all possible, have this conversation over the phone via live voice or something like Zoom. Not via text, chat, email, etc. You can much more easily suss out what’s bugging them, without wigging yourself out. Or worse yet, by answering questions they haven’t thought to ask. Which is almost always a way to torpedo yourself.

Who owns the rights to these sermons? The OP seems awful blase about others people IP or data.

…I’m not sure your intent here: but if the OP hasn’t trained as a court reporter/stenographer, then they should tell anyone they trained as a court reporter/stenographer. That would be a bad thing to do.

“Since it’s work done for the Lord, I made it my number one priority, and told my other clients…and my family and friends… that I’d see them when I’m done with all of your work.”

(You can also secularize this by just telling them how important this is to you, so it’s a priority)

But I would also consider finishing a transcription, then going out for lunch, then telling them you’re done…

eta: Also be praying that no one at the client is a fan of this message board!

I think this is the thread where it starts:

Basically, these sermons are on Youtube, and Youtube offers an automated transcript that is in a format that needs to be winnowed down to paragraphs, and Chat GPT is helping with getting the formatting right and, I would assume, a sanity check where the OP watches the video while reading along to the transcript. I don’t know about the latter part–just what I would do. At least that’s what I think is going on here, but we’ll have to wait.

I meant it in the generic sense of machine stenography, which, while it’s what court reporters use, it’s also closed captioners, real-time event captioners, captioner/translators for the deaf, etc.

Formatting text into paragraphs should be done via regular expressions and a language model, and one would have to be nuts to do it any other way. To proofread the result, on the other hand, one needs to read all of it carefully.

…I’m not getting the sense that the OP is trained on any of that.

I think the OP simply needs to tell the truth. If the transcripts are accurate, then I don’t think it will be a problem. If the transcripts aren’t accurate, then the OP has problems. My first suggestion would be to audit/review the work that has been submitted for accuracy, so you can prepare yourself for what comes next.

Like I said, I don’t know if that’s for sure what it is. The reason to do it that way is you don’t want to deal with regular expressions and are not a coder. That said, ChatGPT can help you with that, too! :slight_smile:

I would expect that as soon as they know he’s using AI to get the work done in a fraction of the time they based the page/pay rate on, they’ll lower the pay scale.

So very much this. Having spent most of my working years in a technical writing career, I can say that we are about to see pay for this kind of work drop drastically. While the OP may be making hay while the sun shines; the sun isn’t going to shine much longer on this kind of work if a human isn’t laboring over all of it. Why, right now, I bet small businesses are about to go ker-flop as the knowedge of how to do this spreads.

Time to polish up some additional skills so you can remain relevent in the marketplace.

I’m thinking of a Reddit thread I saw recently. The comments were…not sympathetic.

https://www.reddit.com/r/ChatGPT/comments/138clv9/spent_5_years_building_up_my_craft_and_ai_will/

I had assumed, reading the OP, that these were sermons that had been posted by the organization that was hiring him.