Exploring and Evaluating Generative Artificial Intelligence, Number One…

AI or Artificial Intelligence, is a hot topic at present with governments weighing up how to invest in the sector, insiders from the industry warning of the potential dangers and the media stoking fears from job losses to the eradication of the human race. During the recent A to Z Challenge 2023, I was highly impressed by the images selected by Misky in her blog IT’S STILL LIFE – A.I. ART & POETRY and decided to investigate the whole subject of AI for myself. I also wanted to know about Generative Text AI because my work demands various forms of word creation and I had heard that this might be one of the first areas to suffer job losses to AI.

But first, a little thought about where I understand us to be right now. There is Soft AI (think Alexa) and then there is the dreaded Hard AI (think Terminator). Soft AI is already in use all around us, Grammarly, Predictive Text on your smartphone, Alexa, Google and other search engines but the reason people are not running around screaming “The terminators are coming!” as a result of all this Soft AI action, is because the intelligence quotient of this sort of AI is pretty low. For example, say to your “smart” speaker Alexa “Alexa! Tell me a joke!” and Alexa will oblige with a joke of variable humour, but then say “Alexa! Tell me another!” and Alexa will not know what you want – unless you specify “joke” in which case you could leave the “another” out since it is irrelevant to Alexa that she has just told you a joke. This means that Alexa is hardly conducting a conversation. On the other hand, she can sort through literally millions of bits of data to find and play you the song you want to hear. So this sort of AI is doing exactly what we want -conducting the repetitive, boring, difficult (because of the magnitude of data) tasks which we none of us really enjoy doing, or given the scale of data today, could even do! I was reading how when the sea-change brought about in Accountancy when computers could sit on every desk and more particularly, when those computers could run spreadsheets, there were fears of mass redundancies in accounting firms. And so there were at the clerk level of the business, but on the other hand, amongst the higher level accountants, jobs boomed like never before, because the things they could do with spreadsheets, the new things they could offer their clients had been undreamable before – so let us take the tales of mass unemployment with a pinch of salt – after all, I wouldn’t mind betting that some of the accounting clerks were able to upgrade their skills and move up the ladder with the help of a computer…

Above is an example of a tricky task for any graphic designer to do manually – to design a seamless tile repeat and I will come back to this particular task in the future, but to say that this task was accomplished in under sixty seconds, is for now simply amazing!

Text AI

So first of all, I want to look at my first encounter with ChatGPT which was developed by OpenAI and which they describe as a Large Language Model Generative pre-trained transformer Chatbot. What that means is that an AI computer programme has been fed a huge amount of written material, novels, adverts, and non-fiction from the World Wide Web and not only does the AI have access to all this material, but it can stitch information together according to very clever algorithms to write “original” material. So I thought I would test it out by asking it to write a piece about one of the more colourful phrases from my recent A to Z Challenge – phrases we mostly know the meaning of but have forgotten the origins of. First, here is my explanation:

“Cold Enough to Freeze the Balls Off a Brass Monkey!” – you may be beginning to see the “lost in translation” aspect of sayings overheard from sailors by landsmen in which case, you will not be surprised to learn that this expression has nothing to do with statues of simians cast in brass losing their genitalia!

A “monkey” on a wooden ship, especially a warship, was a (usually) wooden tray with rounded depressions in which cannon balls were stacked in preparation for the battle – the last thing you wanted was heavy cannon balls rolling around the deck of a pitching deck – let alone a “loose cannon” – so cannons were secured (against recoil on firing) by strong ropes and cannon balls were kept on a monkey. I said that monkeys were usually made of wood, but admirals or very lucky captains, who had made a lot of money from their share of “prizes” (captured ships, evaluated and paid for by the Navy) – were allowed to prettify their ships with gold leaf, dress their crews in custom, fancy uniforms, and purchase brass monkeys rather than the standard issue wooden ones. Now here’s the thing, a monkey was carefully designed so that you could pile the cannon balls up in a pyramid to maximise your supply of cannon balls in as small a space as possible but the thermal coefficient of expansion of brass (the monkey) and iron (the cannon balls) is different – the brass monkey shrinks more than iron in very cold conditions. Now, the carefully spaced second and subsequent layers of the pyramid are too big for their positions and can roll off the monkey hence “Cold Enough to Freeze the Balls Off a Brass Monkey!” – Simples! Not!

Now here is ChatGPT’s first attempt:
“The expression “cold enough to freeze the balls off a brass monkey” is a well-known phrase that means that it is extremely cold outside. Although it is a common phrase, its origin and accuracy are often disputed.
One theory suggests that the phrase originates from the age of sailing ships when cannonballs were stored on the deck of the ship in a brass frame called a “monkey.” The frame consisted of several round holes where the cannonballs were stacked. During cold weather, the brass frame would contract more quickly than the iron cannonballs, causing them to fall out of the frame. Hence, the phrase “cold enough to freeze the balls off a brass monkey.”
However, there is little historical evidence to support this theory. The phrase “brass monkey” itself is not found in any nautical dictionaries or records, and the idea that brass contracts more quickly than iron is also questionable.
Another theory is that the phrase originated from a Victorian-era pub game called “brass monkey,” which involved placing brass balls on a tray and attempting to knock them off with a ball bearing. The game was typically played in cold weather, and if the temperature dropped low enough, the brass balls could freeze to the tray, making it impossible to play. Hence, the phrase “cold enough to freeze the balls off a brass monkey” may have originated from this game.
However, like the previous theory, there is little concrete evidence to support this theory.
Overall, the origin of the phrase “cold enough to freeze the balls off a brass monkey” remains uncertain. It is likely that the phrase has evolved over time, and its origins may be lost to history.”

So what do we think of this? Well firstly, the “the idea that brass contracts more quickly than iron is also questionable ” is simply not true – science tells us that brass both expands and contracts more than iron – so there is a lie right there! Secondly, Chat GPT says that there is no evidence of the phrase in “nautical dictionaries or records. Whether or not such dictionaries even exist, there are plenty of references to this particular phrase and it’s origin in many places on the web and in literature – I first heard of it in the books of Patrick O’Brien who spent much of his life carrying out research for his books in the Admiralty records and I would trust him implicitly. You have to find as many references as you can when researching a subject like this. There are, of course, always people who will dispute the origins of various phrases and so perhaps ChatGPT gave undue credence to these doubters. Still, then ChatGPT goes even further and completely fabricates an alternative theory! The idea of a “Victorian -era pub game called ‘brass monkey’…'” does not show up in any search of the internet whatsoever, so ChatGPT has not only dissed the correct explanation, but suggested a completely fictitious alternative – all done with a convincingly authoritative tone. Luckily, there is a Regenerate button and on dressing this, without any refining the prompt or critiquing the first result, ChatGPT came clean and produced a correct version of the explanation:
“The expression “cold enough to freeze the balls off a brass monkey” is a colloquial phrase used to describe extremely cold weather.
There are a few different theories about the origin of this expression, but the most widely accepted one is that it comes from the era of sailing ships.
In the days of sail, cannons were an important part of a ship’s armament. The cannonballs were made of iron and were stored on deck in piles or pyramids. To prevent them from rolling around in rough seas, the cannonballs were often stacked on a flat brass plate called a monkey.
When the temperature dropped below freezing, the brass monkey would contract more than the iron cannonballs, causing them to fall off. Hence, it was said to be “cold enough to freeze the balls off a brass monkey.”
However, it’s worth noting that there is no definitive proof for this theory, and other explanations have been proposed as well. Nonetheless, the phrase has become a part of English language and culture, and it is still used today to describe extremely cold weather.”

The second iteration is both factually more correct and more concise, but for a first encounter with ChatGPT to throw up such a blatant fabrication is worrying. It suggests that however miraculous the task of researching and collating this story might be, it definitely needs checking by a human being because otherwise, another false news item might find its way onto the Internet and go echoing around and around… Just such a story of defamation by a ChatGPT “news” story, happened to an Australian Mayor who rather than being portrayed as the whistleblower he actually was, he was reported to have gone to jail for corruption – he is suing ChatGPT‘s developers, OpenAI.

Not that I am fishing for compliments but what do you think of my explanation of the phrase compared to ChatGPT’s? After all, you might think I was too verbose and ChatGPT more concise…

Image AI

So far, I have only tried out the Midjourney bot (short for Robot, which I discussed in my “R” post in April) but I will be trying and comparing them and reporting back to you Dear Reader, but for now, I just want to share some images that I have created which were successful, although sometimes after several attempts to refine the prompts – more of that next time…

I then decided to turn this into a tiled image (below).

So we (if you want to come along with me) are on our way to investigating AI – so far Text AI dodgy, and ImageAI – amazing!