ChatGPT - Interesting Things You’ve Used It For

[andy_s]

jagnet said:

andy_s said:

I bought my son a years subscription yesterday, today he has done someone's 6 blog entries she couldn't catch up on, various company email templates and 2 topic posts, 'just the tip of the iceberg' he said, now he knows how I felt in the late nineties when I first double-tapped the IE button thing.

So long as he does some heavy human editing and fact check like a crazy person.

The more ChatGPT generated content there is, the easier it becomes to recognise the traits that it's AI generated. I can't help but feel that there'll be a backlash given how wordy but ultimately superficial much of its blog post generating is.

Maybe it's just me, but I usually perma block any authors on Medium that churn it out by the AI prompt full.

Oh yeah - it's played within the envelope and checked and tweaked, as for bland, repetitive, word-frequency, have you seen corp comms the past ten years...?

[jagnet]

andy_s said:

Oh yeah - it's played within the envelope and checked and tweaked, as for bland, repetitive, word-frequency, have you seen corp comms the past ten years...?

I think those must have made up a large amount of the training data.

But it's great, because now we can use ChatGPT to create even more wordy but ultimately meaningless missives, then everyone receiving it can use ChatGPT to summarise it into bullet points, before creating their own wordy responses, which in turn get summarised by AI.

Corp comms will eventually just end up as LLMs talking to themselves. Would anyone even notice?

[andy_s]

jagnet said:

andy_s said:

Oh yeah - it's played within the envelope and checked and tweaked, as for bland, repetitive, word-frequency, have you seen corp comms the past ten years...?

I think those must have made up a large amount of the training data.

But it's great, because now we can use ChatGPT to create even more wordy but ultimately meaningless missives, then everyone receiving it can use ChatGPT to summarise it into bullet points, before creating their own wordy responses, which in turn get summarised by AI.

Corp comms will eventually just end up as LLMs talking to themselves. Would anyone even notice?

[Hoofy]

jagnet said:

More of a grumpy coder that's had far too much exposure to marketing companies pretending to all the world to be website developers, but they can't code for st.

Having worked for some, strong my intolerance for bull has grown.

On the other hand... balance argument.

So in summary....

If you've enjoyed reading our deep dive into cynicism, do leave a comment.

Success Secrets: Why I'm Following Your Lead

Fair enough, I can see your point and agree. After all, you have a lot of experience in this area, and I respect your opinion. You've worked on similar projects in the past, and you've always been successful. You also have a deep understanding of the industry, and you're always up-to-date on the latest trends. I'm confident that your advice is sound, and I'm happy to follow it. I think it will help us to achieve our goals and objectives more efficiently and effectively.

[andy_s]

[jagnet]

[jagnet]

Had another common ChatGPT tell pointed out to me this morning.

It does love to start marketing style articles with variations on "in today's fast paced digital world..."

I think I auto tune-out opening lines like that already. That's the trouble with today's fast paced digital world, I guess.

[Hoofy]

[bigandclever]

I suspect this might be the wrong thread but I'll ask anyway ... I'm trying to upload some data to GPT-4 via the Noteable plugin and then run an Apriori algorithm on the data to do a bit of basket analysis. I'm hitting an immediate wall:

chatgpt said:

Use the dataset uploaded on Noteable with the path: basket.xlsx
Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

What might I be doing wrong? I strongly suspect it's because I'm at work but is there a way to get a more meaningful error? For instance, I don't know how 'it' knows where to go for the .xlsx file, could it be that? I've also tried the full project address, didn't change the result.

I'm using this video as reference (it has some ads, sorry in advance) .. https://youtu.be/zjpOzR_QFA4?si=7k3V8j5IJs4UKg-H

Thanks in advance.

Edit to add ... I can run it all successfully from home on my own PC so it's definitely something to do with the work laptop. I'd still like to understand where I can get more meaningful error information

Edited by bigandclever on Monday 9th October 18:16

[272BHP]

In the last week or so the quality of response from GPT-4 has been very disappointing and unusable on many occasions.

Now that they have a decent amount of subscribers have they turned down the dial on the servers?

[CharlieCrocodile]

I've got it to write a few "all staff" emails in the past few weeks, I got a few compliments from the C-level team for great comms.

[lizardbrain]

272BHP said:

In the last week or so the quality of response from GPT-4 has been very disappointing and unusable on many occasions.

Now that they have a decent amount of subscribers have they turned down the dial on the servers?

Im sure they have sufficient servers, I understand active users have slipped, there must be other financial motive.

Perhaps there is more value per user in selling the service B2B.

Or they kept getting sued so had to restrict huge parts of the learning?

Just guessing.

I much prefer anthropic for softer use cases anyhow

[andy_s]

Bit O/T but we mentioned training dataset earlier, [corp comms], here's the sources/share/key of 'the pile': https://arxiv.org/pdf/2101.00027.pdf And yeah, all the old ENRON emails was a surprise [a common dataset?] as was YouTube subtitles, clever. It also goes into why/how it's blandified against being un-PC. An interesting one to flick through if you fancy a break from stabbing pins in your eyes.

[outnumbered]

CharlieCrocodile said:

I've got it to write a few "all staff" emails in the past few weeks, I got a few compliments from the C-level team for great comms.

Says something about the quality of your C-level team

Honestly if you can't be arsed to actually write an "all staff" email, why should anyone bother read it ?

[James6112]

outnumbered said:

Says something about the quality of your C-level team

Honestly if you can't be arsed to actually write an "all staff" email, why should anyone bother read it ?

Nobody does anyway!
A quick scan (to see if i’m getting the boot, hopefully replaced by a robot & a payout ) then delete

[dhutch]

I used it to produce the first cut draft of my flexible working request appeal letter!

Basically re-wrote the whole thing before it was finished, but it got my out of the doldrums when I didn't know how to start.

[nodavhere]

Taguht Chat Gpt to talk like my Girl friend.. Single forever

[Hoofy]

dhutch said:

Basically re-wrote the whole thing before it was finished, but it got my out of the doldrums when I didn't know how to start.

This is exactly how people can use it.

It's harder to get starting but much easier to edit someone else's work. Change that, delete this, expand on the second paragraph, change this part into bullet points.

I've probably already posted before but I love it and use it to create a lot of online content for SM.

[turbobloke]

It's still getting basics wrong, but remains polite.

ChatGPT
You are correct. I apologize for the oversight. The cyclopentadienyl cation (C5H5+) is indeed considered antiaromatic.

An antiaromatic compound must meet certain criteria:
It must have a cyclic, planar structure.
It must have a continuous loop of p orbitals.
It must have 4n π-electrons, where n is an integer.

In the case of the cyclopentadienyl cation, it has a cyclic, planar structure with a continuous loop of p orbitals. However, it only has 4 ?-electrons due to the positive charge and a vacant p orbital on carbon, and 4 is not a (4n+2) number as required for aromaticity, making it antiaromatic.

Thank you for pointing out the error.

[Mr Penguin]

turbobloke said:

It's still getting basics wrong, but remains polite.

ChatGPT
You are correct. I apologize for the oversight. The cyclopentadienyl cation (C5H5+) is indeed considered antiaromatic.

An antiaromatic compound must meet certain criteria:
It must have a cyclic, planar structure.
It must have a continuous loop of p orbitals.
It must have 4n ?-electrons, where n is an integer.

In the case of the cyclopentadienyl cation, it has a cyclic, planar structure with a continuous loop of p orbitals. However, it only has 4 ?-electrons due to the positive charge and a vacant p orbital on carbon, and 4 is not a (4n+2) number as required for aromaticity, making it antiaromatic.

Thank you for pointing out the error.

What was your initial prompt?