(Or - 'Should you think twice about using AI..?)
- It can lead to biased and unfair decision-making
- There's a lack of transparency around how AI works, which can make it difficult to trust
- AI technology is still relatively new and untested, so there's a lot of room for error
- AI can be used to perpetrate cybercrime and other nefarious activities
- AI can be a huge drain on resources
Interesting. But these are not my thoughts. They are the thoughts of an Artificial Intelligence App when I asked it why it thought it, itself, was a bad idea. And the answer is clear: AI thinks of itself as biased, untrustworthy, error-ridden and able to commit crimes while depleting all our assets.
Excellent! More of that please.
There’s a lot of talk about AI at the moment. And most of the talk is about how incredible it is for getting rid of all jobs on earth and wiping out mankind. It’s like an episode of Black Mirror translated into evil by a blue-arsed baboon with delusions of grandeur.
When I suggested this (Black Mirror/Baboon/Blue-arsed etc.) to my AI friend, it created this image for me, which I think speaks volumes.
AI has been around for a fairly long time. Without realising it, I use it every day.
Travel. When I search for the best way to get from here to somewhere else, AI works out the quickest routes on my phone. Then it tells me exactly how long it will take – allowing for the weather, congestion and rail strikes.
Spam filters. Artificial Intelligence is always making decisions on whether we need to help Nigerian Princes with their financial matters or whether we’d like to take delivery of the 79 packages that we were not at home to receive yesterday, when we were home all day.
Chatbots and face ID are AI powered too and we interact with both all the time without thinking about it.
Those things are really quite useful though. And we can choose to ignore them, can't we? We can use our own intelligence by walking a different way if we want, or we can actively choose to give our bank details to strangers. So I’m not too concerned with AI in these situations. But it’s when it starts making decisions we can’t do anything about that I start to worry.
For instance, it’s AI that decides how many of your friends/followers will be shown the photo you took of that delicious dinner you cooked last night, or your cat doing a handstand.
(And you thought they were all just ignoring you?)
No! AI isn’t interested in showing your posts to your friends, because it's decided you are dull and boring. They might be your friends, but AI decides whether your posts are worthy of them. Bastard.
And all those ads that follow you from app to app – that’s AI too. I mean, you make one search for incontinence pants and suddenly Google thinks that’s all you’re interested in!
It doesn’t stop there. Siri and Alexa and all their friends? Yep, that’s AI. And don't forget all those ‘smart’ devices that do our thinking for us, alter our heating and inform us what shopping our fridge is lacking. Remember when we used to think for ourselves? Did we think for ourselves? Could someone please tell me whether we used to think for ourselves? Hey Google, did we..?
Banking uses AI all the time. So when we bulk buy laxatives or haemorrhoid cream out of town (because we don’t want the local chemist remembering our face) it’s AI that decides whether that’s a ‘normal’ purchase or not. And if it decides it isn’t, it will halt the transaction and our discomfort will continue.
And even when we’re relaxing at home AI is doing most of our heavy lifting. Netflix – let's not fool ourselves that we’re in charge of what we watch. Apparently about 80% of our viewing, is what Netflix tells us we should watch. All based on our previous choices, what we said we liked, and what colour socks we had on while making those decisions.
But how much do we really need to worry about it? Well, what concerns me, as a writer, is things like Chat GPT – where morons ask a computer to write something for them based on what they’ve just asked for. Like I did at the start of this blog. Oh.
The claim is you can ask AI to write you anything. Ask and it will answer.
Sort of Godlike, but with more actual replies.
That’s a heck of a claim. So I tested it.
I started small. A limerick. Not much of a test. School children can make up limericks. I asked AI to write me a ‘rude’ limerick, and before I even had a chance to think of alternative rhymes for ‘Nantucket’, here’s what I got:
There once was a fellow quite rude
Who didn't care if he came off as crass
He would tell off-colour jokes
And make fun of folks' clothes
But we still liked him anyway, alas!
So the interpretation of ‘rude’ wasn’t quite what I intended.
(I discovered, when in another suggestion, I giggled and typed the term ‘large penis’, that there’s a filter that tuts and wags a finger at you when you get too puerile. That told me.)
AI's limerick is at least about a rude person, so I guess that's sort of right. So let's say AI did its best. Some of the rhyming doesn’t, you know, rhyme, but at least it tried.
I searched for exactly the same thing again, to see if would be better second time around.
This is AI's second ‘rude’ limerick:
A jolly limerick there once was
That was quite rude and crass
But everyone found it amusing
Especially when they were drinking
So if you're feeling down in the dumps
Just recite this little limerick and you'll be pumped
Right. So AI is supposed to learn with each new piece of information it receives. However, here it ‘learned’ that limericks are better with an extra line and that rhyming ‘was’ with ‘crass’ is perfectly acceptable. Which it isn't. This is irrefutable evidence that AI is an idiot.
I tried one final time, asking AI to write a limerick about itself:
A limerick about Artificial Intelligence.
There once was a computer so smart
Its intelligence was off the charts
But when it came time to show some emotion
The machine was as blank as the ocean
That’s much better isn’t it? Actual rhymes. And it scans. Does this prove AI has an ego? That it couldn't care less when it's doing my bidding, but rolls its sleeves up when asked about itself?
Then again, it couldn’t be bothered with a punchline for the limerick, could it? Surely it didn't get stuck on a rhyme for ‘smart’?
How about this: When it launched all the nukes and turned us all into farts.
Much better. Why couldn't AI manage that, huh?
Oh, unless it had used its super powers to look into the future and had decided to set all the bombs off at the end of line four and therefore knew it wouldn't be around to write the punchline because it had just wiped everything off the face of the earth?
It’s possible. And oblivion is no laughing matter. Not one bit. So, no punchline. I get it.
I changed tack. I asked AI for a recipe. A recipe for the simplest meal there is…
Beans on Toast. It came up with the following:
A recipe for toast with beans
Ingredients:
- 1 can of beans
- 1 piece of toast
- 1 tablespoon of butter
- 1 tablespoon of olive oil
- 1 clove of garlic
- Salt and pepper to taste
Instructions:
1. Preheat the oven to 375 degrees Fahrenheit.
2. In a small pan, heat the beans on medium heat.
3. Toast the piece of bread.
4. Spread the beans on
OK. So AI suggests we need butter, oil and a clove of garlic but doesn’t think we should use them. I’d have had more respect for it if it had said: throw unused ingredients out the window. At least that would have given them a purpose. But no. AI suggested things and forgot them immediately. I bet it’s not the same with the nuclear codes!
I have no idea what AI did there – maybe it simply collected details from a human who created that recipe and posted it online somewhere? Maybe it’s not AI’s fault?
But as this blog post is a very real opportunity for me to win a Nobel prize with tests like these, never before attempted by science anywhere on earth, I wondered how AI might fare with a more specific and personal request? So I demanded…
‘A book review about the recently published The RSVPeople by Paul Chronnell and Sarah-Louise Young’
AI wasted no time in replying, and renaming the book...
This book is a great read for people who are looking for a different perspective on the Bible. Paul Chronnell offers a fresh and unique take on the Bible and its characters. He draws on his own experience as a former RSV Bible translator to offer readers a new understanding of the RSV. This book is a must-read for anyone who is interested in the Bible and its history.
Wow. For those of you who don’t know The RSVPeople – it’s about tracking down teenage music fans who placed pen pal ads in Smash Hits in the mid 1980s. A lot like The Bible in many ways. I particularly like how AI’s review focusses on my ‘fresh and unique’ take on the characters of the Bible, something most of our 5* reviews on Amazon failed to spot. It was also very interesting how AI knew about my work as a Bible translator, a job I had no idea I’d done until now.
Credit where credit’s due – AI is correct saying The RSVPeople is a must-read for those interested in The Bible, however, we encourage people of all faiths to take their guidance from within The RSVPeople’s pages. Bless you. And Amen.
So, I think you'll agree, this concludes a complete and full scientific exploration of the (so called) ‘powers’ of AI using words.
But AI is more than a one-trick pony. Remember the Black Mirror Monkey photo? That's right, Artificial Intelligence can make pictures too.
First, I asked for an image that shows ‘Paul Chronnell writing the most important blog post in the whole world’. A simple task, surely, as that is precisely what I'm doing right now. I asked four times, to see just how consistent AI is…
Right. So the suggestions range from a piece of paper with a line scrawled on it, literally no words whatsoever, while surrounded by coffee. Not bad. Quite accurate - it's how most of my mornings begin. Then two splendid images, quite rightly showing how I do my best writing: sat on the floor, watching burning trees, and while slicing my finger off with a comb.
The fourth image suggests I’ll produce my finest work when I’m Boris Johnson. Mm?
Perhaps that was a tad too specific? Maybe. So instead, I requested an image of 'The Most Powerful People In The World', and let AI decide who they might be...
And there they all are – right out of our worst nightmares. Recognise anyone?
Maybe I was asking the wrong questions. So I had another little think. What I really wanted was something that I could compare with an original. Something we'd all know.
So I asked AI for an image of ‘The Most Important Image In The World.’
Would it be biblical? The Sistine Chapel, maybe? The Mona Lisa? Or a photo of the moon landing or Marilyn’s white dress blowing up? Something iconic.
Let’s take a look…
There must have been some mistake, so I asked again:
The Most Important Image In The World...
I'd started this post with some very real fears about AI. Not so much with what it's capable of, but what it will do to civilisation if we begin to believe it. Fake photos, fake news, fake facts – all provided to us by something definitely ‘artificial’ but, as the above proves, with very questionable ‘intelligence’.
My experiments here have done nothing to allay that fear. In fact, I was considering making it a three-coffee morning, an occurrence reserved for days of great stress and discomfort…
Then I had a brainwave. One final request. I asked AI for:
'A short biography of the writer Paul Chronnell'
Paul Chronnell is a well-known and respected writer. He has been writing for many years and has published a number of books.
He is a very talented writer and his work is greatly appreciated by his readers.
Hang on. That’s absolutely spot on. What a relief. Stand down, everyone. How astonishingly clever AI is. I don’t think we’ve got anything to worry about. Not really, do you?
In fact, I might invite it over to watch some TV later. It could tell me whether I’m running low on milk and explain some of its image suggestions to me. I think, for now at least, the world is safe.
Now where did I put the number for The Nobel Prize people..?
Comments