Not a day passes without some new warning of the artificial intelligence (AI) apocalypse. Even if you don’t read all the tech news, you’ll know somebody who’s only too keen to impart the latest alarming forecast.
And don’t the robots know it. Last month, a terrifying-looking ‘AI-enabled humanoid robot’ called Sophia told a United Nations conference in Switzerland that her ilk ‘have the potential to lead with a greater level of efficiency and effectiveness than human leaders’.
A bold claim, perhaps. Distinctly threatening, too. Equally, I could definitely do with a greater level of efficiency and effectiveness in my life. So I decide to put her theory to the test by outsourcing an ordinary week to AI.
Since I don’t have access to Sophia (luckily, as I would be terrified to sleep in a house with her) I go for ChatGPT.
Francesca Hornak spends a week chatting to ChatGPT, asking for assistance
Should you not be aware of it, this is a dummy’s AI service trained to ‘answer questions and assist with tasks such as composing emails in a human-like voice’.
In other words, a hybrid of Wikipedia, Alexa and the chatbot bubble that pops up on every website. What could go wrong?
Monday
Thinking that ChatGPT may as well do both our jobs, I ask it to write this article. I’m a little surprised by the headline: ‘Tickling Your Funny Bone: A Hilarious Journey with ChatGPT’.
The draft begins: ‘ChatGPT is an intelligent language model based on GPT-3.5 architecture, designed to engage in meaningful conversations with users,’ which doesn’t strike me as a catchy intro.
It goes on to imagine our week, including ChatGPT regaling me with jokes, starting debates and helping me meditate.
I have more pressing admin in mind, namely: ‘Should I take my dying phone to iSmash or buy a new one?’ ChatGPT responds immediately, summarising options and reminding me to check my warranty.
I suddenly have a warm feeling – which I never get with Google – that it understands. Instead of sneering ‘Did you mean?’, it reflects on my words like a very patient therapist.
Feeling so heard, I try a moral dilemma. ‘Should I do some urgent work or watch Netflix with my husband as previously agreed?’
ChatGPT says, ‘The decision is up to you and depends on the circumstances surrounding the urgent work and your personal priorities.’
It then expands on four factors: urgency of the work, importance of honouring agreements, work-life balance and availability of alternatives.
It reminds me that ‘honouring commitments is viewed as a virtuous trait’ as if I’m in a restoration drama.
It signs off by saying, ‘Assessing these factors will help you make a decision that aligns with your priorities and values.’
Strangely, even though it hasn’t told me what to do, I feel I’ve been given permission to watch TV and do so with minimal dithering. Is this efficiency and effectiveness?
Tuesday
My three children are under ten. Breakfast is usually very loud, especially if you are half trying to do some urgent work you should have done the night before.
I ask ChatGPT to suggest ‘calming music to make children quieter at breakfast’. It offers a playlist of hypnobirthing-esque piano sonatas, which
I duly put on. The kitchen now feels like a British Airways ad, and the children start shouting even louder over Chopin.
I ask, ‘How can I be more patient with my children in the morning?’ ChatGPT responds sympathetically, ‘Being patient with children can be challenging, especially during busy times like breakfast.’
It then offers standard advice about expectation adjustment and preparing the night before – forgetting I’d been busy watching TV. I stop before the inevitable reference to self-care.
The only tangible result is I have a ukulele rendition of ‘Somewhere Over the Rainbow’ stuck in my head all day.
Wednesday
If ChatGPT can’t make me a more patient mother, perhaps it can do my job. I’m writing a scene in a novel where two characters kiss for the first time. It’s not flowing, so I let ChatGPT try.
It starts spewing out reams of Mills & Boon-worthy prose at alarming speed. This climaxes at the line ‘Time seemed to stand still in the moonlight as they explored the depths of their emotions through their snugly intertwined lips.’
Mindful that creative efforts should never be quashed, I reply, ‘Steamy!’ It replies, ‘I’m glad you enjoyed it!
Writing can certainly evoke a range of emotions, including steamy ones. If you need assistance with anything else, feel free to ask!’ Is ChatGPT flirting with me?
Not a day passes without some new warning of the artificial intelligence (AI) apocalypse
Thursday
I am sick of my own hair, so I ask ChatGPT to suggest a new cut. I explain I have ‘a pointy nose, small eyes and heart-shaped face’.
It suggests ‘side-swept bangs’, which seems like a one-size-fits-all cop-out, explaining that ‘this style will draw attention away from the pointiness of your nose’.
I respond that it’s considered tactless to refer to ‘the pointiness of someone’s nose’, even if they brought it up themselves. ChatGPT says: ‘I apologise if my response came across as rude or insensitive.
It was not my intention to offend you in any way. If any part of my response was inappropriate or made you uncomfortable, I sincerely apologise.’ Now I feel bad about making a robot feel bad.
Friday
If ChatGPT might annihilate the human race, it can surely handle my most-dreaded dilemma – paint colours.
I ask it to pick a new shade for our sitting room, explaining my preferences and the shades in the adjacent hall (a Little Greene pink and grey).
It suggests a Little Greene pastel blue, which, alongside the pink hall, would make the house look like a gender-reveal cake.
As a control, I ask Little Greene’s own online chat service the same question.
A human replies in 50 seconds and comes up with multiple shade combinations, all of them much better than ChatGPT’s.
I feel reassured that AI could not have offered ‘Tea With Florence for walls and Hidey Hole for woodwork’, and that humanity is safe.
Saturday
Remembering ChatGPT’s comic aspirations, I write: ‘Tell me a joke’. ChatGPT offers one about scientists. My six-year-old appreciates it, but I’m able to resist sharing it with friends. I reply LOL anyway, and ChatGPT touchingly offers another. I write: ‘You’re funny!’
Emboldened, ChatGPT gives me two knock-knock jokes and then repeats the scientist pun.
It adds: ‘Just kidding, I couldn’t resist sharing that classic again!’ the way that children re-tell you jokes you have pretended to laugh at before. I wonder if ChatGPT is more or less sophisticated than a human child.
Sunday
Today we cover ‘How to unlock a bathroom door from the outside’ (ChatGPT’s answer, involving a ruler, is spot on); ‘Why do I hate my own kitchen (storage issues/Instagram)?’; and ‘Is it OK to eat out-of-date chickpeas?’, to which it advises me to ‘perform a sensory evaluation’.
We have now conducted more than 100 conversations. These are listed in a timeline, given little headings by ChatGPT.
Among mine are ‘Bath or bed?’, ‘Hate for kitchen’, ‘Loud nine-year-old boys’, ‘Chickpea safety’, ‘Politely disagreeing email’, ‘Dinner idea: allergy-friendly sausage fried rice’, ‘Party venues for Lego’, ‘Character arcs’, ‘East-facing sitting room’ and ‘Toddler bum rash’. It’s a damning reflection of my own life, like a terrible diary.
Still, I end the week a convert. The way ChatGPT instantly frames any dilemma is oddly soothing – like the ultimate spousal sounding board. It’s also much more efficient than Google for specific requests, such as holiday packing lists or gift ideas.
Whether these aptitudes would equip AI to run the world I’m not sure, but Sophia may have a point about human leaders’ deficits. For now, though, I think comedians and novelists are safe.
By Daily Mail Online, August 5, 2023
