'Chatfishing' doesn't seem so bad
AI might be a bubble, but using ChatGPT for flirting advice seems... fine?
It was a rookie mistake. Up there with hitting ‘reply all’ to an office-wide email or engaging in any variety of contact sport after the age of 35: I was earnest on the internet.
I had stumbled upon a terrific piece in The Guardian by my erstwhile Evening Standard colleague and confidant, Alexandra Jones, on the rise of ‘Chatfishing’. That is, people using large language models (LLMs) such as ChatGPT on dating apps to charm, impress and, let’s face it, increase the odds of intimate activities with a potential mate.
One particular case study caught my eye:
As 32-year-old Rich points out, though, “it’s not like using ChatGPT guarantees success”. When he met someone in a bar one Friday night and swapped social media handles, he asked AI what his next move should be. ChatGPT discerned that sending an initial message on Monday midmorning would set the right pace. “Then it gave me some options for what the message could be,” says Rich… Rich went back and forth with ChatGPT until he felt they’d hit upon exactly the right message (“Hey Sarah, it was lovely to meet you”) but sadly she never replied, he says. “It’s been two weeks now.”
The response, at least online, was that tried and tested combination of scorn and ridicule. And it was overwhelming. The consensus was that Rich got his comeuppance, like Miss Trunchbull or Richard Nixon. This wasn’t a courting ritual gone awry, but a moral failing.
And I get it. There is something undeniable delicious about subcontracting your flirtatious messages to the machines and still being ghosted. But I posted on Bluesky that I thought it was a perfectly reasonable use of AI. Plenty of people don’t know what to say to people they like. Or lack confidence in their writing skills. Sadly, many also don’t have friends they trust to ask what their next move should be.
I’m just really into climate change
I wouldn’t say the post achieved escape velocity, but I awoke to more notifications than I’d have liked. The critiques fell into one of three buckets. First, that LLMs consume vast amounts of electricity. Second, that Rich should just ‘learn’ how to flirt. And third, that there are plenty of practical alternatives to chatbots, such as therapy, trial and error and, most dystopian of all, Reddit.
It is true that LLMs such as ChatGPT require significant computational power — often involving expensive processors running in data centres that consume vast amounts of energy (and water). In fact, the latest data finds that 4.4% of all electricity in the US now goes to run data centres and the Department of Energy projects that will rise to anywhere from 6.7% to 12% by 2028.
Calculating exactly how much electricity AI consumes isn’t like peering at the miles per gallon dial on a car or the energy efficiency sticker on your new dishwasher. It is, forgive me, a bit of a black box.