22 Comments
User's avatar
Ben Mathes's avatar

This is great and needs to be paired with the equivalent of "let me google that for you", because some people err on the other side and interrupt and ask you for something that they could have just asked chatgpt for.

Rick Manelius's avatar

Yes! I think the https://dontpastetheai.com/ website is an attempt to do that. It starts off gentle and has a more aggressive version.

I think a lot of people just aren't aware of how it's being perceived/received... so they keep doing it.

Ben Mathes's avatar

I think I mean the inverse? Like this: https://letmechatgptthat.app/

Karen Melchior's avatar

just what I was thinking and commented :D

Jeff Clark, MD's avatar

I love this. Great acronym.

MC's avatar
Aug 22Edited

We have found an elegant way to solve this.

More and more. AI output is NOT for the person but for the person’s agent

So message is split

Short note person to person

*Agent note*

agent output to agent

The note to the person is always from you. Can be a sentence or two. Doesn't need to be long

The agent note that follows is not meant to be read by the person. It's meant to be processed by their agent

Person -> person

Agent -> agent

I think this will become the new norm. We've gone from a two-way conversation to a four-way conversation.

Rick Manelius's avatar

I love that you and your team have already found an acceptable norm to embrace this in a positive way. I think it's critical to have these conversation now, because norms will continue to evolve as AI usage increases in volume and complexity.

poet-warrior's avatar

I think the way to solve this is to refine our holistic standard for what is shareable: minimally as pre-AI it should be necessary, and correct/true. And before AI there was quality gauge that was hard to fake. And part of quality is it whether the content was framed appropriately- because one of the failure modes of these models is correct/necessary but verbose and inaccessible text/output. We need to update our individual notions and social and institutional norms around sharing

Rick Manelius's avatar

Yeah we are nowhere close to having a consensus on social norms on this... but I think it should be established at least at the company team level. I'm seeing some folks sling Claude output at each other in Slack threads, and that's not productive 90% of the time.

Santiago Janse's avatar

Super agree on the approach. For me, I if see a wall of text, I go Ctrl+F and look for em dashes (the — that AI likes to use for everything)

If there are any, this is likely AI generated, and thus I'll use AI to summarize it. If you don't write it, I don't read it.

Rick Manelius's avatar

Yeah it's even more than em-dashes. There are other obvious tells (negative parallelism). I can tell when the human's voice drops off and the AI one comes in.

Doesn't mean it's not valuable, but AI tends to bloviate. And so my interest drops a ton.

Lily Clark's avatar

I appreciate the credit! And great read. My thoughts exactly

Daniel's avatar

People have a website for this: https://dontpastetheai.com

Rick Manelius's avatar

Outstanding. I didn't know that, but will use that from here on out!

Niraj Agarwal, Ph.D.'s avatar

AI;DR is still TL;DR...the wall of text, now pretending to be from me. If we AI;DR we pretty soon won't have much to read. I do miss the thinking-writing loop, but editing-curating is a must, otherwise we cease to be human.

Rick Manelius's avatar

Yes it's similar. Although even in a short paragraph, I'll tune out within a few words if it's clearly AI.

James Thornton's avatar

Thank you for voicing this!

Danny's avatar

Concur. I’ve been noticing a related phenomenon with AI: the gap between what someone can produce and what they actually understand is getting much wider.

Someone can now generate a very sophisticated-looking strategy, technical spec, or analysis in a few minutes. It uses all the right terminology, anticipates a bunch of edge cases, and sounds like it came from an expert. But the person sending it may not really understand the assumptions behind it, know which parts actually matter, or have any sense of how difficult it would be to execute.

I think of it as borrowed competence AI gives you access to the language and output of expertise without necessarily giving you the underlying mental model. And because the output looks smart, it’s easy to feel like the thinking behind it must be equally solid.

It also has a funny way of turning uncertainty into certainty. Someone starts with a vague thought like, “we should probably make sure we’re collecting the right data,” and AI turns that into a highly specific 15-point technical requirement. Suddenly all the hesitation and ambiguity that was present in the original thought is gone.

AI absolutely makes people more capable. But I think we’re going to have to get much better at separating the sophistication of the output from the sophistication of the person’s actual understanding of it.

Casey's avatar

GPTZero AI Detection

Model 4.9b

We are highly confident this text was AI generated

Chance this entire text is...

AI 100% Mixed 0% Human 0%

Everyday Freethought's avatar

"we should expect that everyone is utilizing AI at SOME point in their process".

I stopped reading at that point. AI is garbage. Society is making a big mistake not kicking this nonsense to the curb. I only expect stupid people to utilize AI.

Karen Melchior's avatar

it's the flip side of the page: "Let me google that for you" ? And the AI version :)