"Can ChatGPT do my job and live my life? I test it out"
"Hey Cosmo UK readers! Have you ever heard of ChatGPT? It's a pretty cool AI technology created by OpenAI that can hold conversations, answer questions, and even write text."
If you're wondering why that intro sounds like it was written by an enthusiastic robot, that's because… it was. By ChatGPT – an advanced AI chatbot that can do things like write poems, scripts and even texts you don't quite know how to word. When it launched in November last year, people were quick to experiment. Some used it to write messages to their Tinder matches, while one school in south London announced it was likely to ditch at-home essays after teachers gave a ChatGPT-written essay an A*. All you need to do is create a free account with OpenAI and you're good to go.
There's more: ChatGPT could change the way we source our information, too. Tech giant Microsoft just announced a multi-billion dollar partnership with OpenAI and is set to integrate ChatGPT into the search engine Bing. Google is also trialling its own rival AI chatbot called Bard, which will be rolled out via its search engine: you could ask it a question, like, 'What should I consider before buying a family car?' and it'll respond with handy points written in a conversational manner. But, it’s already proved it isn't perfect – in a promotional video for the launch of Bard, the chatbot shared factually inaccurate information.
This may all sound like a lot of change in a short space of time, but chatbots are actually nothing new – many of us have slowly lost the will to live while trying to get an answer out of an automated online chat service, and some millennials may remember furiously corresponding with SmarterChild on MSN back in the 00s.
So, what makes ChatGPT different? And could it really change our lives – and how we work – forever?
Can ChatGPT write emails and do my life admin?
"It's the level of sophistication and the ways it's able to respond to the nuances of human language," explains Mhairi Aitken, an ethics research fellow at The Alan Turing Institute, when asked what sets ChatGPT apart. "It's trained on lots of different kinds of human language and to recognise genres or types of language, so it can create different outputs too."
As someone whose job revolves around writing, I don't love the idea of being replaced by a robot, but maybe it could help me out with some simple life admin instead? Okay, so it can't book a dentist appointment for me (yet) or clean the bathroom, but maybe it could write my emails and assist with arranging dinner with a friend? I give it a go. In the name of journalism.
The test
When I give ChatGPT a prompt and ask it to write a work email thanking somebody for letting me interview them, it starts typing. I'm hopeful. Then the screen goes blank and there's a network error. I persist, and on the second attempt, it works.
At first, I’m impressed, but on second glance I realise the message is incredibly long and oddly formal: "I wanted to take a moment to express my gratitude for taking the time to answer my questions. Your help was invaluable and greatly appreciated. Your knowledge and expertise provided me with valuable insights and helped me understand the matter much better." I bin it off and write my own one-line response.
Next, I try to make my brief more specific and ask for a less formal tone. ChatGPT's response includes three exclamation marks in one short email (far too many) and also gets confused, writing the email as if I'm the expert, rather than the other way round: "Thanks for reaching out! I'd love to chat with you about ChatGPT and share my thoughts on the topic. A phone call at 9am tomorrow sounds good to me. Let's do it! Just let me know if that time still works for you and I'll make sure I'm available. Talk to you soon!"
Maybe robots can't completely outsmart us, after all...
I press on with the experiment, but by this point I'm getting a bit impatient, so I ask ChatGPT to draft a reply to an email and 'sound like a normal person and not a robot'. Amazingly, this works. In fact, the email it writes is eerily similar to the one I actually write and send (a simple: 'That's great - I'll give you a call then. Look forward to speaking with you.')
Next, I trial it over WhatsApp, given all the praise ChatGPT has received for being able to understand and deliver 'human' tone, I wonder if a friend will notice it's an AI bot asking her if she wants to grab food together? (“Hey there! Would you be interested in going out for dinner soon?”)
Once her reply is in (she is interested - and because she's not a bot, asks how I am too), I set ChatGPT to work, and it comes up with this: “I'm good, thanks for asking! Glad to hear that you're interested. How about we plan for this weekend? Would that work for you, or do you have a preferred date in mind?”
My friend says she’s busy that weekend, and asks me to send some other days. I brief ChatGPT with a date to suggest, but it makes me sound quite rude (again). Rather than throwing out a date as an option, it skips ahead: “Sure, let's plan for the 24th of Feb.” The whole exchange feels pretty awkward. My friend stops replying. I then have to quickly explain that I’ve been communicating via a chatbot, which all in all, has not been a great success. So much for technology making our lives easier.
The verdict
Having played around with ChatGPT for a few days, it best delivers when I ask for help unravelling a very boring – but infuriating – admin issue with my doctor's surgery, something I've been trying to resolve for years. I need to be polite but firm, and figured ChatGPT's go-to formal tone would actually be quite useful here. Turns out, it is.
But given I'm an actual human, contacting other actual humans, I feel a bit weird about sending AI-written messages. I keep editing or completely discarding the text for this experiment, fearing ChatGPT's prowess for being curt, and also ultimately pondering the ethics.
"I think it depends on how it's being used," Aitken says. "So if it's being used as a tool to help structure a letter or an email, but you still write it, then that's okay, that's not that different from existing tools that we already use, like spellcheck and grammar check. But if people are using it to actually draft content, that becomes much more problematic."
Aitken adds that in the future, we could have a ChatGPT interface built into our email providers. "If I have my ChatGPT interface write an email and then your ChatGPT interface writes an email back to me, it's just bots communicating with bots. It raises some fundamental questions about what things need to be reserved for actual human interaction." This is particularly worrying when it comes to communicating with those we're close to, Aitken points out: "When you start thinking about it in terms of intimate relationships or friendships, automating those kinds of interactions feels like a slippery slope."
So, what does the future hold for this kind of AI technology? "We're seeing the tip of the iceberg at the moment," Aitken notes. "With OpenAI's deal with Microsoft, we're likely going to see it integrated fairly soon into products such as Bing, Outlook, Word and PowerPoint." But ChatGPT doesn't know what's true or false, so, at this stage, it's best to think of it as an advanced spelling and grammar checker.
There's no doubt that ChatGPT and AI technology will continue to change the way we do things, but Aitken is also keen to stress that there can be a tendency to overhype what this kind of tech is capable of. "They're not intelligent, they just do what they're programmed to do," she says. And ultimately, that’s the issue here – however advanced it is, ChatGPT can't respond to things in the exact same way a human would. It can't pick up on nuances or get the tone exactly right.
Aitken also stresses that, rather than getting caught up in how powerful the new tech is, our main focus should be looking at how people plan to use it. "We need to make sure that organisations and people making decisions about AI systems are held accountable. It's important as these systems are about to play bigger and bigger roles in our lives." That might be true, but when it comes to the smaller things, like writing emails or messages to other humans, the bots still have a lot to learn.
You can give ChatGPT a try here (at your peril)
You Might Also Like
