ChatGPT: something to play with over Christmas

I’ve noticed a couple of posts recently that include a liberal sprinkling of bold font. Looks like AI to me. Is this bad or good? If someone is not confident with their standard of grammar or spelling I suppose it helps to present a readable post. Better than all capitals, no punctuation and no paragraphs!

 

Bad without fail as far as I'm concerned and I'm seeing more and more of the bullet point format

It used to be that a hearsay enquiry generated by family lore was one or two lines eg

"what can you tell me about Uncle Bob - we all called him Fred, that was a rear gunner on Spitfires"

Now I have to wade through about a page of tripe that AI has generated scatter gun like of half truths to find the underlying one liner that started it.

If there is so little effort in asking the question - why should I put in more time and effort to disect all the post.

I really feel like doing a one line answer in a similar vein - Claude answer this post - no other tokens spent on the seed.

Ross

 

nah doe's not work false information, i mean one lottery number out of 6 ! stick to my pot luck thanks

 

I'd rather have the human post, even if it's hard to read.

 

It's easy to say the result was yours entirely rather than admit it had a source - I'd far rather people were honest in all instances and sourced their references. Sometimes people genuinely wonder at what they are told - and want to query if it is right. Families tell lots of stories - and one of the historians' findings is often the fallibility of human memories. Typical Ai mistakes can often be traced back to something a human once said somewhere, or wikipedia ;-)

When I ask why was the British 24th Lancers disbanded it tells me "because they suffered catastrophic losses and had to be disbanded" - then I can query it and present it with alternative data and it rapidly changes its mind and agrees with the data - that one aspect of it I find people may not use. Don't just accept it, get further opinions and always rely on yourself and others rather than on it.

 

NFC

AI Is an Anti-Primary Source Machine

All the best

Andreas

 

Newspapers have a long history of being heavily criticized for both being economically and environmentally wasteful, as well as structurally inaccurate.

And then there was censorship and propaganda, particularly during WW2, leading to issues with trust and who or what to believe.

Anyone called Al must be considering changing their name...

On the surface, Paul Simon’s 1986 hit "You Can Call Me Al" sounds like a whimsical, upbeat pop song. However, the track actually maps a deeply personal journey about navigating a midlife crisis, feelings of cultural alienation, and ultimately finding a spiritual awakening.

[1] (You Can Call Me Al - Wikipedia),
[2] (),
[3] (https://www.quora.com/What-do-the-lyrics-to-You-Can-Call-Me-Al-by-Paul-Simon-mean),
[4] (You Can Call Me Al by Paul Simon - Songfacts)

 

I thought that was Alien.

I’ll get my coat…

 

The problem with AI is that it makes everyone think they are an expert on any subject. There is an assumption, reinforced by the polished manner in which the information is presented, that AI can't be 'wrong'. As Ramiles mentions, it rapidly changes its mind when you pull it up on its errors.

This was a good recent example I came across in a Facebook group. The confidence with which it pronounces, 'Lance Corporal France's movements can be mapped directly to the battalion's war diaries through Italy', is going to convince the reader that this must be correct, despite it being a complete lie.



And although I hate to be the 'Aschaully, I think you'll find' guy, I couldn't stop myself responding.



The OP gave that a like, but I'm sure she probably thinks "this bloke can't know more than AI".

 

I would amplify Gary I think its not that "this bloke cant know more than AI" but more that "I've done history, ticked the box - not worth my time looking at again".

As said change the seed and you get a different output - not necessarly more correct, just more targeted to another trait of machine teaching input.

I do feel in a lot of cases the initial urge is due to the social bubble being fed "look and give me a thumbs up - I did this" which leads to a corresponding need for quick result.

Put stuff in - get the "Who do you think you are" expert style result. No desire to know more or more importantly check basic fact validity, cut and paste result. There done it move onto the next fad.

Drive by posters are, in my mind, a suggestion of this mind set - post not answered within a couple of hours, forget and move on, reponse from AI almost instant.

All the above slant towards a one and done task.

Leads to an other question

Why spend time iterative altering seed/prompt to get a more correct answer ?- if some knowledge of the required result why persevere with proofing the numerous drafts from AI and not just cut to the chase by another route.

Ross

 

Quickly identified



More here.

Regards ...

 

Not an AI error von Poop!

Just AI practice try at meeting "10% chance of killing all humans by end of decade"

Ross

 

I'm beginning to find this, "End of Days" scenario very interesting (not in any morbid way). I see it's on the BBC website this morning (again).

Is it genuine, is it hype? I suppose only time will tell but, in the mean time, I'm off to stock up on water and water purification tablets

 

My take is a lot of genuine and a fair amount of hype.

There have been parallels of world ending tech in the past - nbc warfare, power looms, gunpowder etc all had potential and were not contained/nullified on release after creation - just not widely used in the way that triggers the full kill result.

AI agents are in that category - no way to erase now created. No effective regulation/kill switch that could not be nullified by human or machine intent.

What is not widely understood is the route AI armagendon would take.

There have been pointers to what disruption single events have on our connected world eg cloudfare outage, suez block by evergreen vessel and most recently UK airtraffic from a single corrupted flight plan.

The current genuine fear is intended release of AI for a limited malicious task that escalates.

The hype is terminator hunter killer skynet - but more likely is just stopping you getting replacement water and tablets after you use up your stockpile.

Take the malformed prompts with poor range limits that we see in AI slop described in this thread and think what the result of "AI - cure global warming" without specifying that it was not to remove the perceived root cause of human development.

Ross

 

This is the thread that keeps-on-giving ...mostly entertainment, but a few insights!

On a recent flight I was able to watch the film "Good Luck, Have Fun, Don't Die" a black comedy/science fiction/action movie.

Its centred around the idea that someone from the future needs to come back and save humanity.

Without giving too much away, our hero needs to apply a safety/security patch to AI to stop it getting out of control. But it also provides some entertaining ideas on where we could be going with AI & the internet. It also has a dark twist on school shootings, and a high tech (with Ads) cloning process.

This is not usually 'my thing' but I really enjoyed it, and loved the comedy.

If they don't fall into the trap of making a second movie (which I fear they might) this could turn into a cult classic 30 or 40 years from now (...if we survive!)

Here is the trailer:-


This film is currently available on Rakuten TV and Amazon Prime Video.

 

I've posted these same two pictures on the 'Wildlife' thread.

I finally managed to get a picture of a red squirrel. Unfortunately, I was too far away. So I zoomed in and asked AI to 'clean' it up.

Good or bad? Honest opinions....