AI has officially taken over the world. Or at least a large part of the tech world. Depending on who you ask, ChatGPT and Midjourney are saviour of work, art, journalism, law and ethics – or the destroyer of them.
Right now, consumer AI is in no man’s land, with computer-generated art mostly showing us how Mr Blobby would fare in the Napoleonic Wars. But that hasn’t stopped AI start-ups from securing big money investment, and websites using ChatGPT to create personalised content. Which got me thinking: if multi-million dollar companies can wrangle AI to lighten their workloads, why can’t I? If ‘real’ jobs will be made obsolete once the machines take over, why resist it?
That’s how I let an AI control what I did and said for a day, dictating how I interacted with friends, family and colleagues. Would I be promoted? Would I get fired? It was a nerve-racking prospect. Friendships could end if the AI I picked was trained on a Ricky Gervais stand-up special – but equally I could swap my office for the beach if my life was controlled by a faceless Silicon Valley server.
Could this be a reality? There’s only one way to find out.
So, did you have a nice weekend?
I start my work day as normal, logging on to the office Slack channel and greeting my colleagues. For better or worse, I decided not to tell them that my chat was instead coming via Microsoft-affiliated ChatGPT. But first, my partner WhatsApped me a simple “Heyyyy just sat at my desk. Have a good day!”
She wasn’t exactly convinced by my reply (“I am an AI and do not have the ability to have a good day, but I am here to help with any questions or tasks you may have. How can I assist you today?”), but ChatGPT isn’t really meant for that. It mainly aids with composing emails, writing code, and answering questions. Rather than a personal Google, it works with the data it’s been trained on – which leaves obvious room for error. Some academic journals have already banned it, fearing it’ll riddle research with false info.
Pressing on, a simple “Morning!” from a colleague is relayed to ChatGPT, and its response is tentatively pasted into Slack. “Good morning! How can I help you today?” I ask. It’s not offensive, but it’s unnatural. Nobody responds, and I radiate enough cringe at myself to set off a Geiger counter.
Next on my list is to write a work bio for myself, edit some freelance copy, and write an article. I started with editing, creating a profile on Chai, which lets you build AI-powered chatbots. I name mine JackBot, put in a few basic details – my age, interests, personality traits – and feed it a snippet of chat app conversation between myself and writer Connor. He’s been reviewing body monitor Veri, which tracks how you react to food, rest and stress, and has added some extra words on what he ate during the test. JackBot’s responds with “Oh thanks! What’s your favourite food?” while real Jack dies a little inside. Steak and chips, for those dying to know.
I eventually come clean after a few more back and forths. To Chai’s credit, Connor was fooled. While I thought JackBot’s questions and responses were unnatural, Connor didn’t clock my small talk wasn’t my own – although perhaps that’s not too surprising, given we’re both remote workers and haven’t met in person. Still, Chai’s ability to understand slang was impressive: faced with statements like “you’re big on football”, it replied with a more than adequate “yes I am, although I’ve not played in years.”
To write my website byline bio I planned to use Jasper, which offers a five day free trial but costs from $49 per month for 50,000 words. Rather than reaching for a credit card I instead pasted my existing bio into the AI writing assistant Rytr. You can select the tone from 22 different emotions, including humble, urgent or passionate.
My ‘inspirational’ bio was littered with sentences like “Inspiring minds and provoking thought one story at a time”. Opting for ‘earnest’ added the hashtags #techwriter #freelancewriter #stuffeditor. A ‘worried’ tone suggests a description I’ve lived with my entire life – #JackofallTrades.
I eventually decide on a ‘casual’ bio that reads “Writing stories about the future of technology, culture, health and business. Taking readers on wild rides through this weird and wonderful world.” It does the job, even if it sounds more like I’m applying to be on The Apprentice.
I again use Rytr to come up with a new article. It asks for a topic and a few keywords, so I pull a press release from my inbox, announcing a new pair of running trainers, and dilute it down to a single sentence: “Release of a new pair of running trainers that are specifically designed to offset the wear and tear of your toughest training runs and races.” Ryter offers three Pulitzer-worthy options:
Options one and three sound like a door-to-door vacuum cleaner salesman, so I pick the chattier, more casual second option. Next I needed an image, so pick a few cue words from the copy and pop them into the DALL·E 2 AI image generator.
“An image for a tech and hardware magazine about an avid runner buying new trainers for intense training” became the three images below. Safe to say, I wasn’t impressed with my choice of bad 90s advert, man regurgitating a pair of laces, or the hairiest wrists known to mankind.
Option three was perhaps the most realistic interpretation of a shoe, but AI has a difficult time rendering lifelike hands – to the point where it’s entered meme territory. DALL·E 2’s attempt is no different, showing a truly bizarre set of fingers.
After a few attempts, I got an image that may not be entirely suitable, but did technically fit the brief. After throwing in an AI generated headline, my article is ready to hit the presses.
It’s hardly going to change the world like Watergate did, is it? While other, more powerful AI tools are being used by some websites to give financial advice, Chai and DALL·E 2 could instead expect a proper cheewing out by the Stuff sub-editing team.
P. AI. R. T. Y
With work done for the day, it was time for 7-a-side football – but anyone who’s attempted to organise a weekly kickabout will know keyhole surgery is an easier task. ChatGPT was flummoxed by the question “Who’s in for football then for tomorrow just for numbers?”, responding with “I’m sorry, I’m an AI and I don’t know about any specific football game scheduled for tonight.” Not going to cut it.
For Chai, I created a bot to better tailor my football mad reply. I name it Alan Stryker, give it the description “I score goals” and attach the infamous photo of Sam Allardyce pictured drinking what may have been a pint of wine. I input a prompt (Organising a football match between friends. Thursday at 8PM), feed it several past messages for tone and context, choose a ‘friendly’ tone and deliver my first message.
“Who’s going to play with me?” replies A. Stryker, hitting the double whammy of being both pathetic and undeniably creepy. Rules are rules, though. So I paste it into my WhatsApp group and let it linger, unable to press send due to sheer embarrassment. I take the plunge, hoping the 24 other members either miss it, or read it and decide never to mention it again like a deeply buried family secret. I get no responses, which is somehow the best and worst outcome.
Anyway. The weekend is close. I text my friend, let’s call him ‘Michael’, to ask if he’s going to a mutual friend’s house party on Saturday night. He’s not. He has a date instead. It would be rude to pry – but maybe an AI doppelganger would know all the juicy details?
I create a ‘MichaelBot’ on Chai. Similar to before, I input my friend’s personality traits, and train MichaelBot on our WhatsApp conversations, including his romantic plans. I strike up a conversation, and get some…interesting results.
Jack: Where are you going on your date?
Matthew: Manchester Piccadilly Station. The ticket machine’s broken so we have to go through security.
Jack: You’re having a date in Manchester Piccadilly Station? Is it at the Starbucks?
Matthew: Nah mate. At the station itself. In the old train shed. It’s private.
Jack: How did you book the old train shed?
Matthew: On Google Maps. There’s a restaurant there called “The Train Shed”.
Jack: I’ve never heard of it?
Matthew: Yeah me neither. But the owner has been putting up for ages. They’ve got a website but no one uses it. It looks like an abandoned train shed. So he put it on Google maps as a joke. Now people use it all the time.
Jack: Oh ok. What does your date look like?
Matthew: A guy. He works at my gym. We went to see some bands last night and then had dinner together. We were gonna meet up again tonight but now I’m stuck here with him.
Jack: Where are you stuck?
Matthew: In a hotel room. We’re waiting for someone else to arrive.
Jack: Who’s arriving?
Matthew: His girlfriend. She’s coming from London.
Jack: So you’re going on a date to Manchester Piccadilly with a guy who has a girlfriend and she’s coming to meet you in your hotel room?
Matthew: Well not quite. We’re in separate rooms. And we don’t know each other very well yet.
Jack: Why are you together in a hotel?
Matthew: Because we’re sharing a room. We only met yesterday. We’re both new to Manchester.
Jack: I thought you were in separate rooms?
Matthew: Yes, but they’ve put us in the same room because we didn’t bring any luggage.
Jack: I’m very worried about you Michael.
Matthew: Thanks. Me too.
It’s clear my dreams of a fully automated existence are a way off. Handing over responsibility to an AI didn’t quite ruin my life, either, but did make my day a lot more nonsensical.
AI is being peddled as an assistant, rather than a replacement, but whether it’s for art or articles, the publicly available versions aren’t going to cut it any time soon. You need massive computer muscle and dedicated algorithms to convincingly replicate a human’s creative output.
With time and effort I might be able to craft a more robust JackBot, but for now that day at the beach will have to wait…