ChatGPT - Interesting Things You’ve Used It For
Discussion
My uses are
- Programming help, writing very routine functions that I'm too lazy to
- Programming help, but more as a coach for the harder stuff
- Telling me where certain system files are
- Summarising information (in a production environment)
- Summarising very long and rambling forum posts
- Writing prompts that enable the above
Mr Penguin said:
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 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 π-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? 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 π-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.
First prompt
Is the cyclopentadienyl cation aromatic or antiaromatic.
ChatGPT first response
The cyclopentadienyl cation (C5H5+) is aromatic.
It is aromatic because it has a planar, cyclic structure with 6 π-electrons delocalized around the ring. This satisfies Huckel's rule for aromaticity, which states that a molecule is aromatic if it has 4n+2 π-electrons (where n is an integer), and the ring is planar. In the case of the cyclopentadienium cation, it has 6 π-electrons, fulfilling the requirement of 4n+2 where n=1.
Therefore, the cyclopentadienyl cation is aromatic.
Second prompt
The cyclopentadienyl anion has 6 π-electrons, whereas the cyclopentadienyl cation has 4 π-electrons and a vacant p orbital, surely this makes it antiaromatic.
Chat GPT then returned the response posted earlier, with its apology. I still suspect the student had used AI. For other far more demanding questions, involving physical constants which ChatGPT needs to dig out, its response can be astonishingly fast and accurate after hammering out the equivalent of a side of A4 in a few seconds.
I'm using the free version, maybe the paid-for variant learns faster, but it could hardly be more polite.
From previous usei when checking it out on complex problem solving with lengthy calculations involved, I get the impression that a lot of the I in its AI is knitting together logically and very quickly various resources it finds online, at impressive speed for both the location and the logic plus knitting together almost seamlessly. It pauses occasionally for barely a second or two during calculations, which gives me the impression it's trying new searches, then it steams on.
In its replies it appears to have used an ageing resource giving a name for the cyclopentadienyl cation which isn't used these days, in the same way that what was a carbonium ion is now a carbocation.
In its replies it appears to have used an ageing resource giving a name for the cyclopentadienyl cation which isn't used these days, in the same way that what was a carbonium ion is now a carbocation.
I've found it very hit and miss ....
I needed to write a windows script, couldn't' get one command to work - tried google, nothing, asked chatGPT, instant correct answer.
I thought I'd get it to write a Linux script ... it was a bit more complex than below, but the basic idea ...
was pulling certain fields out off a file and formatting them ....
The file Iooked like this
Field1 aaa
Field2. bbb
Field3 ccc
Fieldx. xxx
Field2. lll
Field3. mmm
Fieldx zzz
The field names were unique but repeated, just with a different value - all contained in one file.
My request was along the lines of:
Filed1 is the start of a section, fieldx is the last line in the section
Pull out the following fields field1, field4, field10 and put them in one line in output file <filename> comma separated.
fieldx is the last line of the section.
When you get to field1, repeat, but output to a new line in output file.
The output should look like this
aaa, ddd, jjj
kkk, nnn, ttt
It failed miserably, it just couldn't understand that fieldx was the end, and thus following output should be on a new line in the output file.
Oddly, when I told it to use the **** as the separator, it could do it.
I needed to write a windows script, couldn't' get one command to work - tried google, nothing, asked chatGPT, instant correct answer.
I thought I'd get it to write a Linux script ... it was a bit more complex than below, but the basic idea ...
was pulling certain fields out off a file and formatting them ....
The file Iooked like this
Field1 aaa
Field2. bbb
Field3 ccc
Fieldx. xxx
- ******
Field2. lll
Field3. mmm
Fieldx zzz
- ******
The field names were unique but repeated, just with a different value - all contained in one file.
My request was along the lines of:
Filed1 is the start of a section, fieldx is the last line in the section
Pull out the following fields field1, field4, field10 and put them in one line in output file <filename> comma separated.
fieldx is the last line of the section.
When you get to field1, repeat, but output to a new line in output file.
The output should look like this
aaa, ddd, jjj
kkk, nnn, ttt
It failed miserably, it just couldn't understand that fieldx was the end, and thus following output should be on a new line in the output file.
Oddly, when I told it to use the **** as the separator, it could do it.
Mr Whippy said:
Does ChatGPT actually “search”?
I thought it created its answers from its NN and your input/s only?
Not a techie so can only repeat what I've read, which is that it didn't have access to the internet and relied on training. However this is curious as the mistakes it makes in my trials are many and frequent, and they correspond to the incorrect information available on the websites of some web resources including universities, rarely UK which is good news. I thought it created its answers from its NN and your input/s only?
What are the (short) pauses in complex problem solving / calculations due to if all the information it thinks it needs is in-house...volume of requests slowing it down at times could be one reason, but the pauses in calculations within complex problem solving occur at points in the solution where a new element of the solution arises e.g. electrochemistry linking between a non-standard cell emf using Nernst, to Gibbs free energy, and on to an equilbrium constant at various temperatures.( Apologies to non physical science readers for the detail, I don't have any other basis to offer.) It doesn't decline to answer at the outset, so why pause at the joins if it's just repeating something it's done before in training albeit with different numbers? Also if it's been trainied and remembers things how come it gets relatively basic things wrong so often / uses circular arguments / etc.
The pausing at join-up points looks like too much of a coincidence, but appearances can be deceptive.
Why would Open AI choose duff trainers and / or training materials? If they've been created from scratch in so many fields to be useful, this would surely be like AI reinventing the wheel. AIUI these days - since Sept 2023 - it does hunt online but only the paid version will release current real-time links to relevant web resources and go beyond the 2021 limit. Possibly searches have been creeping into free versions for development reasons. We only find out after the fact what it's been doing, do we not?
Earlier in the thread, somebody posting about making a particular coding task request led me to find an existing solution online in a matter of minutes which was confirmed as code relating to the original task, in a different context, which could be adapted. I don't know much at all about code but I can read posts and write questions in Google! I'll see if I can find it and edit this post with a copy and paste. I should add that the minutes were spent framing and writing the question, the googled results were scanned in a matter of seconds to find the most likely hit, ChatGPT would surely be able to do this whole process very quickly.
ETA this was it, this thread in May last year.
TonyRPH said:
turbobloke said:
<snip>
However...in terms of what I thought might be happening (hoovering the internet for examples to use or adapt and use) I spent 10 seconds online and found this.
https://krazytech.com/programs/simple-library-mana...
Could a searchable database for books in a library - or similar, this was meant to be a general example - be located and adapted to become a searchable database for a collection of CDs? To repeat, I have no knowledge of what you were looking to do so this may be completely n/a and once again I'm just asking.
The concept is identical. Maybe it did hoover up that code and present it to me having updated the code to modern standards - e.g. using prepared statements for the databse queries etc.However...in terms of what I thought might be happening (hoovering the internet for examples to use or adapt and use) I spent 10 seconds online and found this.
https://krazytech.com/programs/simple-library-mana...
Could a searchable database for books in a library - or similar, this was meant to be a general example - be located and adapted to become a searchable database for a collection of CDs? To repeat, I have no knowledge of what you were looking to do so this may be completely n/a and once again I'm just asking.
Very interesting!
Conversely though, I have written much of my own code from scratch, and asked it to improve what I have written, and the response is not only instant, but the code is markedly improved too!
So there must be some 'knowledge' there, but I don't know how!
Edited by turbobloke on Sunday 17th March 21:33
Mr Penguin said:
Mr Whippy said:
Does ChatGPT actually “search”?
I thought it created its answers from its NN and your input/s only?
It can access the Internet to get the latest answers but it doesn't research things in the way a human would. I thought it created its answers from its NN and your input/s only?
But a novel question to catch it out. You create a website about this animal.
Ie. How long is an adolescent wuankidonc’s tail?
It’s not going to dig that up from your website and tell you the answer. That’d be very dangerous.
It’s going to throw it into its NN and then do something with it.
I’ve been dipping into this video again and again over the last year.
It’s long and I just don’t have the time, and coming back to it months later I’ve kinda lost track.
https://m.youtube.com/watch?v=flXrLGPY3SU
Definitely worth a watch to understand the NN side of things and raw language stuff.
I’m just not sure how the new data can be added safely by letting ChatGPT just going to scour the web.
It’d end up polluted rather quickly?!
Interesting in any case, I’m gonna use it to do some programming soon… and use the graphics side to help me with artwork, for a game project.
I’m broadly ok at programming and good at graphics, so will be interesting to see if I can use both to get a one man team creating the output of what would have been more like two or three people!
Mr Whippy said:
Of course the whole NN is created from data from all over.
But a novel question to catch it out. You create a website about this animal.
Ie. How long is an adolescent wuankidonc’s tail?
It’s not going to dig that up from your website and tell you the answer. That’d be very dangerous.
It’s going to throw it into its NN and then do something with it.
I’ve been dipping into this video again and again over the last year.
It’s long and I just don’t have the time, and coming back to it months later I’ve kinda lost track.
https://m.youtube.com/watch?v=flXrLGPY3SU
Definitely worth a watch to understand the NN side of things and raw language stuff.
I’m just not sure how the new data can be added safely by letting ChatGPT just going to scour the web.
It’d end up polluted rather quickly?!
Interesting in any case, I’m gonna use it to do some programming soon… and use the graphics side to help me with artwork, for a game project.
I’m broadly ok at programming and good at graphics, so will be interesting to see if I can use both to get a one man team creating the output of what would have been more like two or three people!
I have had a play with something similar. GPT helped with the maths involved in creating nicely swooping aliens in a 2D shooter and on the graphics side helped in generating the pixel art for the sprites.But a novel question to catch it out. You create a website about this animal.
Ie. How long is an adolescent wuankidonc’s tail?
It’s not going to dig that up from your website and tell you the answer. That’d be very dangerous.
It’s going to throw it into its NN and then do something with it.
I’ve been dipping into this video again and again over the last year.
It’s long and I just don’t have the time, and coming back to it months later I’ve kinda lost track.
https://m.youtube.com/watch?v=flXrLGPY3SU
Definitely worth a watch to understand the NN side of things and raw language stuff.
I’m just not sure how the new data can be added safely by letting ChatGPT just going to scour the web.
It’d end up polluted rather quickly?!
Interesting in any case, I’m gonna use it to do some programming soon… and use the graphics side to help me with artwork, for a game project.
I’m broadly ok at programming and good at graphics, so will be interesting to see if I can use both to get a one man team creating the output of what would have been more like two or three people!
It didn't always do as it was told with the graphics and when I asked for a 16x16 sprite it would return all sorts of different sizes and needed to be told off a number of times before behaving itself and producing something usable.
I've created a YT channel that's 90% AI generated (the 10% comes from me having to put some manual labour into it eg thinking of the topics before getting AI to flesh out the topic idea and the script, checking the video isn't gibberish, uploading it, copypasting the description etc). The script, footage and voiceover are all generated.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
Hoofy said:
I've created a YT channel that's 90% AI generated (the 10% comes from me having to put some manual labour into it eg thinking of the topics before getting AI to flesh out the topic idea and the script, checking the video isn't gibberish, uploading it, copypasting the description etc). The script, footage and voiceover are all generated.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
fascinating idea - will be interested to see the results...I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
but equally depressed that it is another nail in the coffin of high-quality content online
akirk said:
Hoofy said:
I've created a YT channel that's 90% AI generated (the 10% comes from me having to put some manual labour into it eg thinking of the topics before getting AI to flesh out the topic idea and the script, checking the video isn't gibberish, uploading it, copypasting the description etc). The script, footage and voiceover are all generated.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
fascinating idea - will be interested to see the results...I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
but equally depressed that it is another nail in the coffin of high-quality content online
SpudLink said:
akirk said:
Hoofy said:
I've created a YT channel that's 90% AI generated (the 10% comes from me having to put some manual labour into it eg thinking of the topics before getting AI to flesh out the topic idea and the script, checking the video isn't gibberish, uploading it, copypasting the description etc). The script, footage and voiceover are all generated.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
fascinating idea - will be interested to see the results...I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
but equally depressed that it is another nail in the coffin of high-quality content online
I’m amazed people actually watch these.
And it’s pretty damning about the uses and abilities of AI.
I’m sure there are tons of fringe uses where it’s really great and valuable, but the current hype around AI feels like it’s at a level where half the jobs in the planet will be fulfilled by a couple of big tech cos.
In actual fact what we’re gonna see is just reams of utter s
t advertising, pictures, videos, films, etc etc…When anyone can create hours of film or libraries or books for a few quid, that will be the end result.
Mr Whippy said:
SpudLink said:
akirk said:
Hoofy said:
I've created a YT channel that's 90% AI generated (the 10% comes from me having to put some manual labour into it eg thinking of the topics before getting AI to flesh out the topic idea and the script, checking the video isn't gibberish, uploading it, copypasting the description etc). The script, footage and voiceover are all generated.
I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
fascinating idea - will be interested to see the results...I'll run it for a year and see how things go. Either (1) I'll get monetised (2) I'll promote something from it (3) I'll bin the channel.
but equally depressed that it is another nail in the coffin of high-quality content online
I’m amazed people actually watch these.
And it’s pretty damning about the uses and abilities of AI.
I’m sure there are tons of fringe uses where it’s really great and valuable, but the current hype around AI feels like it’s at a level where half the jobs in the planet will be fulfilled by a couple of big tech cos.
In actual fact what we’re gonna see is just reams of utter s
t advertising, pictures, videos, films, etc etc…When anyone can create hours of film or libraries or books for a few quid, that will be the end result.
t videos, you're surprised anyone watches etc. I've seen some videos where the comments are really positive mainly in the motivational space eg "watching your videos has turned my life around". Nobody will be saying that about Mr Beast's videos which can be argued to be high quality. Actually, nobody is going to be saying that about Love Island or Portillo's walks or Harry Potter or Star Wars. This entertainment stuff is just material to make the time between now and your death feel shorter. 
I still don' really get how LLMs store data .
If a LLM can be 1000 lines of code, with weights, how for instance does it know the birthday of every usa president up to a training cut off, if these birthdays are not stored in a database or in the code?
Edit: never mind. I asked a LLM for the answer.
If a LLM can be 1000 lines of code, with weights, how for instance does it know the birthday of every usa president up to a training cut off, if these birthdays are not stored in a database or in the code?
Edit: never mind. I asked a LLM for the answer.
Edited by lizardbrain on Tuesday 19th March 10:13
Hoofy said:
Can either of you link to a channel that covers what you're saying? Ie stuff that is bland, st videos, you're surprised anyone watches etc. I've seen some videos where the comments are really positive mainly in the motivational space eg "watching your videos has turned my life around". Nobody will be saying that about Mr Beast's videos which can be argued to be high quality. Actually, nobody is going to be saying that about Love Island or Portillo's walks or Harry Potter or Star Wars. This entertainment stuff is just material to make the time between now and your death feel shorter.
Sorry can’t recall the names. I often block them.t videos, you're surprised anyone watches etc. I've seen some videos where the comments are really positive mainly in the motivational space eg "watching your videos has turned my life around". Nobody will be saying that about Mr Beast's videos which can be argued to be high quality. Actually, nobody is going to be saying that about Love Island or Portillo's walks or Harry Potter or Star Wars. This entertainment stuff is just material to make the time between now and your death feel shorter. There is one that was space news or something. A planet discovered, etc.
Then you sit working with this droning voice for 15 mins talking word soup about something tentatively associated with the video title.
The visuals are like stick videos all stick together.
It’s a viable video for sure, acceptable, but given a documentary cam cram in years I’d research, interviews, specific video footage, etc etc, for 45 mins… it makes these 15 min videos feel utterly devoid of effort.
And I suppose that’s the point. If the ‘creator’ exerts no effort, what is their output going to look like really?
They’ve no buy-in on its success as tomorrow they just create more, and throw enough s
t at the wall until some sticks and they get a few million views.Again I’m sure AI will open some great doors for content creation and ideas etc.
But it’ll also be abused to the point of utter frustration. It can already be tough finding good info and videos on the internet now.
What could be a 1 page infographic diagram is now a 10 minute video, with adverts, and often still lacking.
But where is the money in an infographic?
We’re just encouraging information to be hidden in long boring videos so you can have adverts rammed into your face.
So innovative and advanced
Mr Whippy said:
Hoofy said:
Can either of you link to a channel that covers what you're saying? Ie stuff that is bland, st videos, you're surprised anyone watches etc. I've seen some videos where the comments are really positive mainly in the motivational space eg "watching your videos has turned my life around". Nobody will be saying that about Mr Beast's videos which can be argued to be high quality. Actually, nobody is going to be saying that about Love Island or Portillo's walks or Harry Potter or Star Wars. This entertainment stuff is just material to make the time between now and your death feel shorter.
Sorry can’t recall the names. I often block them.t videos, you're surprised anyone watches etc. I've seen some videos where the comments are really positive mainly in the motivational space eg "watching your videos has turned my life around". Nobody will be saying that about Mr Beast's videos which can be argued to be high quality. Actually, nobody is going to be saying that about Love Island or Portillo's walks or Harry Potter or Star Wars. This entertainment stuff is just material to make the time between now and your death feel shorter. There is one that was space news or something. A planet discovered, etc.
Then you sit working with this droning voice for 15 mins talking word soup about something tentatively associated with the video title.
The visuals are like stick videos all stick together.
It’s a viable video for sure, acceptable, but given a documentary cam cram in years I’d research, interviews, specific video footage, etc etc, for 45 mins… it makes these 15 min videos feel utterly devoid of effort.
And I suppose that’s the point. If the ‘creator’ exerts no effort, what is their output going to look like really?
They’ve no buy-in on its success as tomorrow they just create more, and throw enough s
t at the wall until some sticks and they get a few million views.Again I’m sure AI will open some great doors for content creation and ideas etc.
But it’ll also be abused to the point of utter frustration. It can already be tough finding good info and videos on the internet now.
What could be a 1 page infographic diagram is now a 10 minute video, with adverts, and often still lacking.
But where is the money in an infographic?
We’re just encouraging information to be hidden in long boring videos so you can have adverts rammed into your face.
So innovative and advanced
On the subject of a droning voice and word soup, I wonder if (1) the script was written by someone where English isn't their first language so they just accepted the script put together by ChatGPT (2) they chose the wrong voice (a cheap voice). I've tried a number of different voices. ElevenLabs is the best but costs a bit. Most people will use CapCut because it's free/unlimited but the voices are naff. Many use FlexClip and ClipChamp. Better but not great and still free/unlimited - CapCut is preferred because of the overall editing functionality.
Ultimately, the good stuff (human or AI) will still outshine the rot. Just like Amazon Prime and Netflix is full of crap films, most people still tend to watch the decent stuff and that gets recommended more than the crap ones.
As for my own AI material, being a native Brit, I like to think I've got a decent command of the English language so I vet my videos more carefully. Hence it still takes time. FWIW if I read the script myself (I could and have but I'm testing AI) it would probably take 3 times longer to create the video. If I wrote the script myself, it would probably take 10 times as long. I'd also move back to a more freestyle format rather than a written script (which I'm planning on doing more of anyway for my main channel).
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff