So today I want to dive into a dark corner in the world of AI. To be honest, I was hesitant to continue researching this story more than a few times as I found it disturbing, but thought it was too important not to explore. It’s the dark side of personified AI.

What is Personified AI?

I think the most popular depiction of this is the movie Her. At the time it was widely successful and at the time considered science fiction. Well it’s not very fiction anymore. With the proliferation of ChatGPT, there’s also been a handful of apps targeting the more casual use of AI. Instead of being used to help you at work, it’s more of your friend. Or companion. Or lover.

The leader in this space has been character.ai (or c.ai) for short. When I first heard of this company a few years ago, I thought nothing of it. I thought they would target the video game / virtual reality market. Making NPCs more realistic in games for a more immersive experience. But that doesn’t seem to be the case. On joining the website, you’re recommended things from a crush, to a ‘mafia boyfriend’ (whatever that means), to a psychologist.

Instead of staying in the virtual realm to make it more like real life. They’re pulling people out of the real world into the virtual world. And it isn’t so ‘fun’ for some.

AI Addiction

So I stumbled upon this story last week while researching general Sycophancy (I’ll get to that in a later article now). And down my rabbit hole I found a handful of reddit posts about Character AI being ‘scarily addicting’. About someone getting rid of it to finally go outside. Oh and there was the post referring to itself as being, ‘just another deleting C.ai post’. People deleting their accounts so often as to now be sort of cliche.

This is a blurb from the most recent one.

It’s become so common, that someone even set up a Recovering C.AI subreddit with the advice mentioning things like ‘relapses’ and ‘going cold turkey’.

Who Uses It?

So I know what you must be thinking, it’s just a small group of older men like in Her? And it’s just occasionally a teenager who stumbles upon it, right?

Unfortunately the truth is much worse.

The minimum age for Character AI is 16, which may sound ‘old enough’. But this is a self given age with no official verification. So anyone, at any age, can lie on their account to gain access.

And it isn’t a small group of people. With 60M active monthly users, Character AI is the second biggest AI application in the world. Even beating out Google’s Bard.

60% of their users are between 18-24, which is abnormally young compared to other AI applications.

They’re increasingly on mobile which is often a more intimate form of technology.

And they spend over an hour on the app, per day

We are seeing a large and growing amount of young adults carrying around virtual friends, romantic partners, and therapists without any sort of oversight. And even worse, there’s definitely more minors in these statistics than shown because they’re able to lie about their age.

Where Does This Go?

It seems as all these AI applications want to become more and more immersive, they’ll move to make it more and more human. We’ve already gone from text based chat to having the ability to speak to ChatGPT. Character AI has the same ability.

Eventually I think we’ll see a merging of deep fakes with chat bots to have a fully virtual persona. (NOTE: This isn’t the real Tom Cruise)

The ‘crush’, ‘boyfriend’, and ‘psychologist’ may become even more dystopian at that point.

What To Do

While this article may sound harsh, I think it’s important topic to be discussed. We’re starting to see the data come in about how social media is affecting teens and I have a feeling this phenomenon can have a similar affect. For kids to develop in a healthy environment, they need to be away from their phone and socializing with others. Its the main thesis of the hottest new book, The Anxious Generation. These characters seem to distract from that socialization. Creating a short-term dopamine hit that seems to pacify the need to socialize. Only for it to do long-term harm along the way.

As we progress with breakthroughs in AI, I think we need to think of how these things can be accessed and by who. And I think before there’s enough data to understand how this affects childhood development, it’s best to be on the safe side.

We have age verification for things like online gambling so don’t think that’s out of the question on consumer AI applications. Even if it affects profit and growth rates of the startups creating these technologies.

Thanks For Reading

As always thanks for reading, even the darker ones 😬. In the weeks to follow, I’ll continue to cover things within the world of AI that aren’t as often discussed. Like safety, morality, and ethics. I hope you enjoy it and be sure to subscribe, if you haven’t already!