One Question Quiz

InternetMay 6, 2024

Here’s what happened when I replaced my personality with the Meta AI chatbot for a day


Keeping up with online communication can be exhausting, so Fran Barclay enlisted the help of Meta’s new ‘intelligent assistant’ to respond to all her messages. Could her mates tell the difference?

For centuries, technology has ruled the ways in which we communicate. From the dawn of written language, to the printing press and the internet, our interpersonal relationships have evolved alongside technological progress.

In the latest instalment of this ancient symbiosis, Meta’s new chatbot, Meta AI, has arrived. Heralded as an “intelligent assistant”, the tech giant claims that its artificial intelligence will “expand your knowledge and capabilities”. 

As someone generally incapable of keeping up with social media messages, I put this assertion to the test. Could the bot effectively replace me in my digital friendships? For most of us, keeping up with loved ones is a key motivator for using social media. Can Meta AI help us to maintain these valuable connections by taking the effort out of crafting an “intelligent” response?

As you may have noticed, the bot is now embedded in WhatsApp, Instagram and Facebook Messenger. You can chat with Meta AI directly, like your childhood imaginary friend – only this one feeds your secrets to one of the most powerful companies on the planet. Alternatively, you can tag it into group chats and private messages like that Grecian oracle you’ve always wanted to consult. 

In a bid to save my relationships from a slow and painful death by unread messages, I asked the chatbot to write my responses for a day. Of course, this would only prove a viable solution if my pals couldn’t tell the difference. As ethically as possible without their explicit consent to this process, I set about copying and pasting messages from friends into my private chat with the algorithm, where I requested an appropriate response. Within a few seconds, the chatbot would generate perfectly personable answers. What could possibly go wrong?

‘Creepy as hell’

Unlike in the 2013 AI movie Her, large language models are not trained on our personalities – they’re trained on the internet. That means anyone who writes with a degree of personal style, slang, or even trademark typos is going to sound unusually generic when Meta AI speaks on their behalf. 

To be (human), or not to be: that is the question

Faced with a message from a former flatmate, I turned to AI to keep the conversation flowing with a suitable dose of enthusiasm and rapport.

Nothing like a bike ride to bond with friends

Fantastic news indeed! And even better was the speed with which I could show my delight at her goal-crushing antics, with a brief ad-lib about my own week. 

Clearly this bot was invested in my friendship if it said we should meet up IRL. Still, there was something deeply unpleasant about outsourcing communication with someone I used to see on a daily basis. Yes, by replying I was showing my desire to stay connected, but it came at the cost of sounding strangely like Mark Zuckerberg – or as this friend put it, “creepy as hell”. 

Fortunately, Meta has some sage advice for customising content to your style and friendship:

  1. Know your friend’s sense of humour: If your friend appreciates sarcasm or playful teasing, use that to your advantage. If they’re more serious, tone down the humour.
  2. Reflect their language and tone: Mirror the level of enthusiasm, formality, and language your friend used in their initial message.
  3. Reference shared experiences or inside jokes: This helps create a sense of connection and makes your response feel more personal.
  4. Be authentic and genuine: Use your own voice and avoid using overly generic or cheesy phrases.
  5. Keep it concise: Don’t overdo it with a long, elaborate response. Keep it brief and to the point.
  6. Pay attention to your friend’s preferences: If they mentioned specific interests or goals, show interest and ask follow-up questions.
  7. Use emojis wisely: Emojis can help convey tone and friendliness, but use them sparingly to avoid coming across as overly enthusiastic or insincere.

As the chatbot said, the goal of any relationship is to “be yourself” and show “genuine interest.” Words to live by in this age of disillusion and ennui. 

‘I’m gonna unfriend you’

Now that I was better equipped to work with this technology in a more collaborative manner, I tried these tips on a former colleague, Dan. 

Dan had previously advised that he finds Meta AI “boring” and would not tolerate its use in our relationship. Meta’s AI policy asserts that the bot should not be used to “deceive or mislead others”, so I turned to the algorithm to craft a response to this ultimatum:

Nothing more old-fashioned than instantaneous communication online.

Either this was just the bot’s attempt at “playful teasing”, or Meta AI had breached its own ground rules. I did my best to keep the deceit “concise” and match Dan’s language, but true to his word, he blocked me for the rest of the day. 

Still, there was something endearing about the concern with which the algorithm tried to counsel my relationship:

 “I hope your conversation with Dan goes well and that you’re able to stay in touch in a way that feels authentic and meaningful to both of you,” it said.

If that all sounds incredibly two-faced, that’s because it’s a machine. As Allyn Robins, AI lead at the Brainbox Institute, explained to me, Meta AI “doesn’t ‘understand’ things” and nor does it have a moral compass.

“Don’t expect consistency, or for it to be particularly useful for anything that requires specific knowledge or authenticity,” Robins warned. 

Missed opportunities

Authenticity is a nebulous goal in digital communication. What does it mean to be “real” on a social media platform known to manipulate our thinking patterns and behaviours? Most of us don’t have a clear idea of who we are in the eyes of our beholders, but we do exert some agency in how we express ourselves – even if it’s through a well-placed emoji or a meme. 

By the end of this experiment, I was undeniably anxious about the harm I could have caused to my relationships. At the same time, I was encouraged by the ways in which the algorithm seemed to reinforce the importance of human connections. When I asked the chatbot to articulate the risks of what I had been doing, it listed “missed opportunities” for “meaningful moments of connection and deeper conversations with your loved ones”. 

Arguably, that means getting off the internet and trying to see said loved ones in person. But for those of us who live too far from home or are fiscally obliged to make a living, digital communication will continue to play a key role in maintaining our relationships.

According to Meta, its artificial intelligence is meant to “augment your relationships, not replace the authenticity and emotional depth that comes from direct, human interaction”. It seems that being ‘authentically’ yourself, with all your problematic punctuation, is one function that AI can’t yet replicate.

Keep going!