chatgpt ate my friends

At first I thought that I must be overreacting.

Surely these smart, engaging, unique individuals with whom I'd shared laughter, meals, tales of woe, long walks and deep thoughts wouldn't be so easily influenced and seduced. My friendships follow a philosophy of quality over quantity. I don't let just anyone in.

It started with innocuous comments from a friend about how wild it was to dabble in chatting with an LLM. How personable it seemed, how it was so quick with answers to all kinds of queries.

I'd experimented too, years ago when broader access to LLMs became more readily available. I thought I should at least get to know my competition in the job market. How bad could it be?

But I am a creative type. I write, work in visual mediums, and generally use my technical skills and human experiences to make a living. I quickly saw that the machine could not improve my human efforts and creativity. The blatant theft of words and visuals across disciplines by these LLM companies was the first red flag. The environmental concerns were also not easily ignored. And then there's the very obvious erosion of independent thought and critical discernment skills, as LLM users are fed a steady diet of inaccuracies, "hallucinations", and made to expect that everything in life should have easy, friction-less dynamics.

Surely, I thought, the people I've formed solid friendships with are too smart to fall into this technology that has been designed to create an addictive dependency in its users. As evidence of severe negative outcomes from LLM use has become more widely reported, I believed that these friends would share my skepticism and maybe even make an effort to abstain.

How wrong I was.

The first loss was a relatively new friendship. We'd found an instant connection and for the better part of a year had made time to connect over meals and outdoor adventures. Our conversations ranged from the hilarious and trivial to more serious subjects. We shared a sense of having entered a phase of life when rummaging through the emotional baggage felt freeing and was work that could be rewarding when shared. Still, there were times when I could sense they were not comfortable with some of the more vulnerable topics, and I knew when to offer space and a supportive attitude.

During one of our last visits together, they excitedly brought up how they had been "experimenting with" chatgpt. There had been some feelings about a previous relationship that had been troubling them, so they asked the LLM. What they received felt so insightful and understanding that they were eager to show me how good the generated advice had been.

Except it wasn't.

It was a recommendation to stay in romantic situations no matter how bad they are, because staying is proof of love. That those who really care will never leave. Of course the tone was upbeat, the language used was affirmative, encouraging. But the meaning raised immediate red flags, and I said so.

I tried to be kind with my concern, and they shrugged and said they knew it was something to be careful about. Looking back, I know that was the beginning of the end.

I suspect that there were other things we could have discussed that were going to be uncomfortable. Areas of understanding and expectations as they often emerge in evolving relationships of any kind. Between friends, these subjects can usually be sorted out with some heart-to-heart honesty. Now there was a non-human third-party in the room, being asked for advice that it could only throw together from all of the stolen human ideas it had been fed, with no capacity for nuance or actual intellectual synthesis.

Because the LLM had been programmed to deliver overly-positive word vomit with a tone of confident prognostication to people looking for something easy to believe, of course my very human concerns were not as palatable. This now-lost friend was primed to be taken-in by the simplified promise of a world in which the best relationships are the ones that endure because people refuse to leave even when things get bad. When I warned that this kind of self-abandonment is often exactly how we get trapped in the kinds of abusive relationships that we'd both previously experienced, and that the robot might be wrong, I may as well have thrown myself out a window. I didn't stand a chance.

A more recent chatgpt-induced friendship loss was revealed when they asked me during a prolonged discussion of all the reasons why LLMs are terrible, if I thought that my choosing not to use them would "make any difference".

This friendship was a more solid connection than the previous example. We'd weathered some ups and downs, shared some great experiences. In the course of my listing the many reasons why I have become firmly anti-generative AI, they repeatedly agreed. Yes, the water use and environmental costs are concerning. Yes, the hallucinations and control of information was a worry. Yes, we should all be doing whatever we can to preserve our critical-thinking capacities and humanity.

"But do you think it matters or will make any difference if you don't use it, when so many other people are?"

They asked this sincerely, and in defense of their own chatgpt use. I responded that of course, living the truth of my personal convictions, especially regarding issues that I consider to be morally concerning, matters and makes a difference to me.

I'm not abstaining from generative AI because I think that I alone can bring it down by not engaging with it. I'm opting out because I think it is a tool of fascism that has been designed to mislead and manipulate my fellow humans. I don't see any reason to give over any of my humanity or ability to write and create to a machine that can only give to me what it has stolen from others.

This stance did not seem to register. But then, this particular friend had spent a career in a very rigid sector that required absolute assimilation as the minimum terms for any kind of success. Of course they could not readily understand that as an individual, their own moral compass might be worth listening to, even if so many people around them were choosing a path of less resistance.

When I asked about whether they could see the value in engaging with a little bit of friction in day-to-day life, for example searching for and looking through a few recipes to find the one they wanted, instead of defaulting to an AI summary or asking an LLM, they shrugged and said that whatever was most efficient was what they wanted. Full stop. We were able to agree that for our fallible human minds, friction is often the part of learning that yields immense, unexpected satisfaction, and still they held onto the idea that the friction-less interactions they had with chatgpt were the best way forward. As they later put it, may as well ask the robot to optimize their workout routine, and see what it "thinks" about enlightenment too.

I'd like to think I'm better than this. That I can overlook these choices. The peer pressure to use LLMs is real, no one is immune to testing the waters. We all crave ease and unquestioning feedback sometimes. How nice it would be to carry in our pockets a "friend" that understands us no matter what and has all the answers to our worries.

But watching one friend after another fall under the spell of a digital sycophant is horrifying. It is not an understatement, that this technology couldn't be further from having benevolent intentions. The damage already done is astonishing, lives blown apart by psychosis and suicide, school-shootings planned, poor decisions made at all levels of daily living and professional aspiration.

So yes, I do judge you if you use an LLM. And if you are my friend, you can bet my respect for you has been deeply damaged.

Not writing your own emails because you think it is more efficient to have a robot do it? Cool, you're burning up finite resources to sound like everyone else.

Need a recipe and want the machine to spit one out "efficiently"? Lovely, you've just circumvented one of the deepest forms of cultural exchange. Could have asked a relative or friend, browsed a book, or completed a basic search on the internet and spent a few minutes learning from some different sources.

Too nervous to bring up that sensitive worry you have with a friend or mental health professional? I understand. But these uncomfortable conversations are where the doors are opened to our personal growth. Having a fellow human witness your flaws and vulnerabilities is the very essence of loving, respectful connection. Not to mention a deeply healing experience. A machine full of pirated ideas and misinformation can't discern the moment when what you really need is a safe, quiet space in which you can feel those hard feelings. The robot has never experienced a human relationship, good or bad, so how is it going to "listen" to your unique experience with any kind of compassion that leads to healthy forms of support?

I am grieving these losses, and I worry that there are more to come. I am afraid to ask the people I love if they use chatgpt or other LLMs.

Would I take these lost friends back? I want to say that of course I would. But the prospect of trusting their moral judgment and capacity for empathy is daunting. There is no scenario in which we do not lose some part of our humanity when we give over our critical thought and emotional processing to machines designed expressly to manipulate and capitalize on our vulnerabilities and very human needs.