Things AI can do (but there aren't many)

[Madman Starryteller]

Has AI reached that level of plausibility?

Yes.

A few signs that you're dealing with a chatbot:
- They're overly friendly.
- They're yes men.
- They lie around their own hallucinations.
- They are the spawn of Satan.

 

[Madman Starryteller]

Yeah, I didn't get that "automated checks" thing on either of my submissions. Both came back with similarly worded first paragraphs that the stories couldn't be accepted as presented but would be considered if edited in line with the suggestions. The thing is, if the publication had indicated that its first reader was a machine, I'd not have made the submissions. The feedback reads comprehensive and sensible, while still praising the writing and encouraging further engagement. I still doubt there was a human involved. Has AI reached that level of plausibility?

That said... your writing probably is good. But I bet you'd rather want to hear such things from a human than a chatbot or automated system.

 

[Homer Potvin]

Has AI reached that level of plausibility?

Give it another year or so, and we won't even remember what non-AI, digital correspondence sounded like. One of the many reasons I left my job was the realization that I couldn't keep up with emails, agendas, schedules, and shit with turning it over to AI. Everyone I dealt with was using it, and they weren't waiting around for the humans to find the time to respond. And that we were done hiring humans for jobs that AI could do. It was one of those walk out the door before you get carried out moments.

 

[Naomasa298]

Yeah, I didn't get that "automated checks" thing on either of my submissions. Both came back with similarly worded first paragraphs that the stories couldn't be accepted as presented but would be considered if edited in line with the suggestions. The thing is, if the publication had indicated that its first reader was a machine, I'd not have made the submissions. The feedback reads comprehensive and sensible, while still praising the writing and encouraging further engagement. I still doubt there was a human involved. Has AI reached that level of plausibility?

I dunno, but both the responses I got sound like a mix of AI and human reading. It *could* be an AI grammar checker or AI rewriting the email, but that automated checks lines suggests a significant amount of work done by AI.

 

[Rigor Mortis]

I dunno, but both the responses I got sound like a mix of AI and human reading. It *could* be an AI grammar checker or AI rewriting the email, but that automated checks lines suggests a significant amount of work done by AI.

Mileage may vary. I don't think a human could have read, never mind respond so cohesively, to the longer piece I sent within the timeframe. In my view, the human involvement most likely went no further than a quick check that each of the four rules of robotics were applied:

- They're overly friendly.
- They're yes men.
- They lie around their own hallucinations.
- They are the spawn of Satan.

 

[Homer Potvin]

I don't think a human could have read, never mind respond so cohesively, to the longer piece I sent within the timeframe.

INT. OFFICE BUILDING - Afternoon

A MAGAZINE EDITOR snoozes behind his desk. The wrapper of a half eaten sandwich sways to the HVAC on his desk. Bits of lettuce cling to his shirt. His tie knot is terrible. His email DINGS and his eyes fly open.

SMASH CUT TO:

COMPUTER SCREEEN - "You got mail, bruh" flashes in emergency red on the screen.

CUT BACK TO ECU of the Magazine Editor as his sits up in a hurry, knocking the sandwich and a few feet of papers onto the floor.

MAGAZINE EDITOR
Holy shit... I need to read this submission now before we all die!
(grabs his mouse and wipes his lips... sweat forms at his brow)
Wordcount, wordcount... Sweet Baby Jesus, what's the wordcount? 12K. Okay... okay... I can do this. Hail Mary full of grace....​

 

[Rigor Mortis]

INT. OFFICE BUILDING - Afternoon

A MAGAZINE EDITOR snoozes behind his desk. The wrapper of a half eaten sandwich sways to the HVAC on his desk. Bits of lettuce cling to his shirt. His tie knot is terrible. His email DINGS and his eyes fly open.

SMASH CUT TO:

COMPUTER SCREEEN - "You got mail, bruh" flashes in emergency red on the screen.

CUT BACK TO ECU of the Magazine Editor as his sits up in a hurry, knocking the sandwich and a few feet of papers onto the floor.

MAGAZINE EDITOR
Holy shit... I need to read this submission now before we all die!
(grabs his mouse and wipes his lips... sweat forms at his brow)
Wordcount, wordcount... Sweet Baby Jesus, what's the wordcount? 12K. Okay... okay... I can do this. Hail Mary full of grace....​

The effrontery! I may report you to the owner.

Ok, maybe an experienced editor might read 12k words in an hour and a half, but I was also thinking of the frequent intervals when the dude would be muttering things like "wtf?" "ah now" "get on with it" "does anything happen in this piece of shit?" and "is this what an honours degree in Princeton gets you?" That stuff takes up a fair bit of time too, you know.

It was really just the most measured, perfectly phrased feedback that I've ever received that leads me to conclude that it was a machine responding.

 

[Stuart Dren]

I wouldn't be shocked if agents and publishing houses start doing this with novel submissions too. They can just instruct the LLM to flag what has trending content while being first filter for "poor" writing.

The sudden deference to LLMs that I see with friends, colleagues, and businesses makes me think of the bystander effect. We all just stand around with our hands in our pockets, quietly consenting. "Yeah seems to know what it's doing eh." I think its confident tone plays a surprisingly large part in that.

Bonus: Ms. Dren forwarded an important email to her manager. He told her the email chain was too long (a whole 3 emails) so he was going to get GPT to summarize it.

 

[TrevorD]

If you were an agent and got dozens of submissions per month, wouldn't you utilize technology to cull the herd a tad? A lot of money's on the line for them. Reputation too. Are you consistent with their other offerings? I'd skip the purist talk and be all over it.

 

[Rigor Mortis]

If you were an agent and got dozens of submissions per month, wouldn't you utilize technology to cull the herd a tad? A lot of money's on the line for them. Reputation too. Are you consistent with their other offerings? I'd skip the purist talk and be all over it.

Do that if you want. Just be upfront that that's what you're doing and the rest of us can make informed choices.

 

[Trish]

If you were an agent and got dozens of submissions per month, wouldn't you utilize technology to cull the herd a tad? A lot of money's on the line for them. Reputation too. Are you consistent with their other offerings? I'd skip the purist talk and be all over it.

Consider this: If you are "consistent" with their offerings, why do they need you? If there are already ten of you on their list, do they actually want more?

Trending content: The "trending" content wasn't a trend until one thing created that trend. If they only flag things that are already trending the stagnation is going to be real.

LLMs aren't infallible, they recognize patterns, but they also get overwhelmed quite easily, particularly if you are a compact writer, and overcompensate when called on it. It's quite difficult to get them to level back out again and stop being stupid.

Do any of us really want that as the deciding factor for whether or not a human reads? I don't. I don't write for a machine, I write for humans.

 

[Rzero]

I'd prefer a human reader, but agents aren't reading manuscripts half the time. They or their interns are reading half the query letter and deciding not to bother with the manuscript. I'd rather someone or something take a look at my actual writing, whether I wrote a good pitch or not. Marketing and creative writing do have some overlap, but they're two different skillsets. What if my book is great but I'm just not pitching it right?