Two Types of People [AI Edition]
A few weeks ago I read a comment on HackerNews which made me realise just how deep the divide goes between those who are strong supporters of the use of AI everywhere and those of us who remain somewhat more skeptical, even if we do agree there are certain cases where it can be a useful tool.
I'm not going to dig up the comment and quote it in its entirety — that's too close to naming and shaming for me, and I probably couldn't find it in any case — but in summary, the user described how his "AI agent", which has access to his email and instant messaging accounts (à la OpenClaw-style), read a text message from his wife asking about what was for dinner. The agent latched on to an unrelated email from weeks ago which mentioned a mushroom pasta, and went ahead to respond to his wife with a made up recipe for the supposedly planned dish.
For the user writing this comment, the problem was the hallucination incorrect response of the agent. For me, the problem was the agent being involved in this communication at all. Quite frankly, it horrified me.
People often express their frustration with receiving an automated e-mail reponse to a job application, to a complaint they have made to a business, to an enquiry to the local council. This is somewhat understandable, as the responses almost inevitably do not satisfy. They do not answer the question, do not provide the information requested, do not provide a meaningful update.
Surely, the AI-boosters say, if we plonk an agent in the communication, it will be able to answer all and any questions people may have. This is Progress. This might just work for updates on your job application, or a response to an inquiry to your local council about the dead dog on the berm outside your house [1], if the agent is given sufficient and necessary information to adequately answer. I have my doubts though — putting aside the problems associated with relying on an AI agent to provide accurate information about any of these things, no one is about to take accountability for the responses provided. We don't need another pointless layer of useles platitudes, but that's what we'll end up with!
That relates to situations where we have already been conditioned to expect a frustrating and worthless response from our engagement as a supplicant with a faceless corporation. Where I don't think an AI agent belongs is in inter-personal communication.
In the original comment, how did the wife feel about having an agent respond instead of her husband? How would you feel if you communicated with someone on a personal level, and received a response from a souless and faceless machine? It doesn't have to be your spouse, it could be a friend or acquaintance. I know that if I receive an obviously AI-generated response, I am extremely unlikely to respond, and I would deeply question whether I interact with this person again.
It seems as though some people see the end game as humans outsourcing everything to a machine. So many people have already outsourced how they spend nearly every waking moment to an algorithm optimised only to show more ads, and that is not going well. Our attention is being fragmented, whilst research shows that switching and skipping ahead actually lead to more boredom, not less! The algorithm drives young people towards more conformity, more groupthink, while stealing any time which may give rise to introspection, creativity, and the development of individuality. Have you ever met one of those people who is so terminally online, so permanently addicted to social media, that you realise they lack much of a personality of their own? It terrifies me.
This isn't a criticism of young people, but rather the world that we have created for them. There has to be a better way, where we can use technology to genuinely connect with others, where we can learn, without our attention being exploited for profit by a few giant corporations. It used to be this way — before Web 2.0 was taken over by money, and by greed. It used to be that there were many islands of completely different cultures and experiences online, just like in the real world. But even in the physical world, we've arrived at a time where you can travel across the globe and discover that people everywhere dress alike, all the coffee shops look the same, and we are all slaves to the same algorithm.
Inserting AI agents into our interpersonal communication will not make us any less isolated, it will not help us build relationships, and it will not help our mental health. I see no benefit in having an agent perform any act of communication on my behalf, whether that is reading a message, or responding to one.
If I ever get to the point where I want a machine to intervene in between basic communication between me and my spouse — then take me out the back and shoot me. If my spouse gets to that point, well, I would rather her "agent" just print the divorce papers and leave them on my desk.
[1] Whilst working at the public library part time during my university studies, I was on the receiving end of this phone call. Apparently, because the library is run by the council and will answer the phone on a Saturday, this person concluded the part time librarian should know and be able to carry out all functions of the council, extending to the retrieval and disposal of deceased animals.