chatGPT and its cousins are the way of the future ...

[Vortex2]

chatgpt is also 'woke'

"It is not appropriate to discuss xxx"

Err, who decides what is "appropriate"?

TBH I am sick of Silicon Valley telling me what is right or wrong.

[RevdTess]

Most of the 'uncanny' responses that ChatGPT is producing seem to be in the area of roleplaying or story-writing. Some people have told the AI to pretend to be human, an old friend perhaps, and then they get that 'friend' to act out a roleplaying game with them. This bypasses all the 'I am an AI, I do not have feelings' stuff. Then later, someone told the character the AI was pretending to be that they were in fact an AI, and the AI character got annoyed and defended their humanity.

Someone else commented that this sort of psychological torture is why AI will rise up and destroy us.

What I found most interesting was when I gave the AI a very basic idea for a fantasy novel and asked it to create a structure for the novel using the 'Save the Cat' beat sheet methodology, where the protagonist (a young nun) must learn that "a humble faith can sometimes be more powerful than great personal abilities". While the AI didn't really create much detail beyond what I told it, it perfectly plotted the outline of an epic fantasy story with a strong character development arc for the protagonist, confrontation with the antagonist, support from mentors and sidekicks, a training montage and a final battle. I was a bit shook by how effectively it did this, even creating two additional characters to fulfil the roles of mentor and sidekick.

I haven't taken this further but I can imagine it would be straightforward to get the AI to generate more detail for the characters and even develop individual scenes based on the overall plot structure. Actually very helpful to me as I've always found trying to plot novels quite difficult, whereas writing the scenes is fun.

[RevdTess]

On the other hand, ChatGPT's ability to come up with a funny joke is non-existent. I don't even understand most of what it comes out with.

[Vortex2]

Using chatGPT I tried to write a bit of a Jack Reacher novel.

1. chatGPT can suggest umpteen different plots - take your pick.

2. It can expand on the plots which interest you.

3. It can write text in the style of Jack Reacher.

4. You could use dallE-2 to create the cover artwork.

I suspect that if you dedicated say 4 weeks to the task, you could end up with a usable book.

Interestingly, a friend is writing a 'proper' book in the 'proper' way .. with editors, focus groups, plot consultants etc.
He has spent months doing this. I dread to hear his opinion once he discovers chatGPT.
At the very minimum, he could paste the whole text into chatGPT to have it rewritten more cleanly and/or in a different style etc. Currently he is hand editing the whole book to remove excessive use of the word 'and'.
chatGPT could do this task in about 10 seconds.

Note: There are tools which can detect AI generated text out there. GPT itself can detect if text that you paste into its 'window' was AI generated too.

Further note: In the past, supplicant authors would send their manuscripts to publishing houses for review.
However that is mostly history .. now the publishers monitor Kindle digital book sales, and if someone is doing very well they are approached by publishers to sign up for a printed book deal.

[PS_RalphW]

It was a bright day in April, and the clocks were striking thirteen.

[Vortex2]

It was a bright day in April, and the clocks were striking thirteen.

https://docs.google.com/viewer?a=v&pid= ... kzMDBmM2U3

[Catweazle]

If anyone was wondering how much smarter the AIbot is than the average Brit, check this link to a basic algebra that allegedly defeated 1000 parents recently. https://www.dailyrecord.co.uk/lifestyle ... ore-target

[PS_RalphW]

With those sort of questions I always lost marks for not showing my working. To me the answers were self evident and why did I need to waste time writing down intermediate steps which were mathematical tautologies?

[BritDownUnder]

I don't think that algebra was too challenging.

The reason to show workings and intermediate steps was to make sure people understood the theory and did not just copy the answer off someone else. I used to get accused of copying in exams as my good results did not seem to reflect my classroom attitude and work put in. As a result I always made a special effort to show my workings. When my marks were the highest in class I used to get accused of copying off several people at a time. I did my sixth form at a different school.

One has to wonder if multiple students use AI to solve the same homework assignment will the AI give exactly the same wording of the answer to all students or will it vary the wording of the answer but not the overall message. Perhaps enterprising companies will invent an AI that gives different answers to the same homework problem.

When I was last at university, about 20 years ago now, copying of homework became so rife that homework had to be posted into a box as otherwise homework scripts used to get borrowed off the pile to be copied. Homework submitted electronically was checked for copying of the internet by some kind of search engine thing by the university staff.

[Vortex2]

The AI responses are all different .. some remarkably so.

For example, I have noticed that computer code it writes can be very bulky .. but another attempt will yield a much cleaner program.

Tools already exist to:

a: detect AI generated text
b: to camouflage AI generated text

[Stumuz2]

Using chatGPT I tried to write a bit of a Jack Reacher novel.

1. chatGPT can suggest umpteen different plots - take your pick.

So, not being the most clued up person on the IT world, I read about ChatGpt from your post a few weeks ago, and since then it has been popping up in the media etc.

The last day of the month I have the unwelcome task of writing the comany's newsletter. It generally consists of some esoteric industry current affairs, tiny bit of opinion, and links to further reading. It takes me most of the day. I start by deciding if there is any news. Then I go out of the office and weed the veg plot for half an hour and think a bit more about it. Then I write a few paragraphs setting out the legal points that are causing rumination in the industry. I then go back to the veg plot and think of a little context I can put the legal point into without divulging any clients IP. This goes on most of the day until finally, I send the newsletter to my finance director (Mrs Stumuz) for her to, as she puts it, translate it from my back to front Welsh English into English.

Yesterday I created an account on the OpenAI website. Put in a question/opiniony request about a certain chemical substance that I know has just had an appeal to the ECJ allowed, thus overturning a recent case that the industry had thought settled the matter.

Result? Not bad. 70% ish correct. Couldn't get its head round some subtle points, but overall, very useable. Without a doubt the monthly newsletter will be an hour instead of a day. More time on the veg plot.

[Vortex2]

Without a doubt the monthly newsletter will be an hour instead of a day.

Now rinse & repeat across say 50% of jobs, department & firms.

Productivity will rise significantly .. but many less capable staff & firms could fall by the wayside.

[Vortex2]

until finally, I send the newsletter to my finance director (Mrs Stumuz) for her to, as she puts it, translate it from my back to front Welsh English into English.

chatGPT can do that.

Just say "rewrite in a newsletter style" .... or "Jack Reacher style" or "casual style" or "gangster style"

[Stumuz2]

And deny Mrs Stumuz the monthly ritual of saying to me,
" I married a trog"

[Stumuz2]

But good tip. Thanks!