Before I dive into this post, I want to assure you that this is not a summary of existing opinions, nor a prediction of what life will be like in 100 years, nor (necessarily) a screed about how things ought to be. It is more an observation, paired with experience, and some things that mean a great deal to me.
For all of the AI carnage we could talk about from the past 3 or 4 years, nothing has stood out to me so strongly as a phrase that more and more of my friends keep using: “I asked AI…”. These three words worry me more than anything else I have seen, because they betray a fundamental shift in how people relate to technology, and I don’t think it is good.
I’m not trying to say anything bad about my friends. In a sense, especially since they are often quite reasonable (no doubt because they are friends with me, har har har), this phrase is really just an update to the expression “I searched [the internet]” or “I googled”. But it’s also so much more. You don’t ask things, typically, you ask people. AI is not a person, but it’s configured to respond like it. So yes, it seems semantically appropriate to say “I asked AI” because of the way it responds to human feedback, but it’s also a subtle deception, and nobody I know has ever addressed this in polite conversation. I’ve had numerous friends say, “I understand it isn’t sentient” and “it’s good for certain things”, but they always come back to this phrase: “I asked AI”.
Recently, a popular YouTuber whom I occasionally watch was caught using AI to write his scripts, evidenced by a nonsensical script inclusion that is a common response from ChatGPT. This kicked off a firestorm of gossip which I will spare you, but it turns out this YouTuber’s brother had been warning him to stop using AI for so much, but he ignored this.
No doubt, plenty of people would say, “But Risky, that’s not me! I know better!” But the phrase persists. “I asked AI”.
If I told you I started asking questions of the rocks in the backyard, you’d think I was crazy, but sending questions off to a program cleverly disguised to create a human-like response and thinking it is human is also crazy, but it’s acceptable because everyone is doing it. Mind you, you actually get a response in the latter case, but that doesn’t make it any more sentient.
“I still don’t see your point, Risky. Just because of phrasing doesn’t mean people actually believe any of the AI chatbots are alive”.
No, but it’s still a fundamental shift.
People “ask” AI as if “knows” something. It doesn’t know anything. These systems are typically calibrated to predict responses based on the associations of words within massive datasets. They have absolutely no concept of truth, but if you’re lucky, they’ll output correct information based on what other people have posted online. Some people call it TurboGoogle, which I think is fitting, and honestly a better description than “artificial intelligence”. Insofar as the datasets contain truthful information and insofar as the model (in all of its complexity) is well-made, you can do some incredible things with it, but this idea that it “knows” something that you can “ask” from it is a fiction. People say, “But I use it to generate ideas!” That’s fine, but it’s mostly just scraping that from somewhere else, and that gets me into the second part of things.
None of this information is free. By which I mean, almost every accurate piece of information that any AI model has ever ingested has been scraped from the internet or specifically filtered to shield the companies running these from embarrassment. This means that real content creators are losing ad revenue at an alarming pace, as people start “asking” the chatbot, which slurps up the content and regurgitates it [my gosh, am I actually defending the ad system?]. It’s a convenience, to be sure (at least…when it’s actually accurate), but it fundamentally destroys the connection it has to the origins of the information (with the notable exception of some sites which include reference links, which I honestly applaud).
In a lot of ways, AI chatbots are a layer separating us from actual people on the internet, and the internet kind of separates us from people anyway (with only a few exceptions). AI chatbots are essentially a new layer of autistic, anti-social separation from real people who produce information. Not only does this break down society, but it forces people into reliance on systems, which are easily controlled and manipulated.
Remember when people were driving into lakes because Google maps directed them to? (The Office reference). That’s what happens when you trust computer systems more than you trust your own eyeballs. And it’s extremely dangerous. People “ask” AI because they think it “knows” something, but it’s just probability and data scraped from people.
What this all seems to indicate is just how deeply distrustful society has become. And don’t get me wrong, caveat emptor is a Latin phrase that has been around a long time because it’s so incredibly true, and I’m not saying the appliance guy isn’t lying to you when he says you should pay for a new compressor instead of a $50 relay, nor am I saying that your spouse will never ever cheat on you, but people literally trust systems more than the people around them, or their own critical thinking. And I think that’s really what’s going on here. The AI sounds confident, it tells you what you want to hear, it accepts correction when you correct it (without pushing back like real people might), and a lot of people even say it makes them feel incredibly smart.
On that note, how many people use chatbots regularly but don’t even know the names of their neighbors?
I damn near got kicked out of this house because shortly after I moved in, one of the wifi networks went down, and apparently having a computer security background makes you immediately suspicious. To be fair, extreme stress, past trauma, and other things played a role in me being accused, but do you know what also played a role? He asked AI! Rather than talk to me about it, rather than assess everything he knows about my character and behavior from the previous 5 months of knowing me, he assumed the worst, and then let AI confirm it. It’s a miracle I’m still living here (apparently customer support showed him that turning the router off and back on again would fix the issue, or something like that).
AI psychosis is real, and it’s spreading like a cancer.
Warts and all, I prefer real people over AI, and I always will. Yeah, people suck sometimes, and friends can disappoint, but without people, what even is the point? Moreover, have you ever had a really good disagreement with somebody? Are you really friends with someone if you’ve never argued over anything?
“Risky, you’re getting off topic, what does this have to do with AI?”
Everything! I get the whole thing with efficiency, but I think efficiency kind of stopped being a net positive at some point and is largely being used to destroy the world, but sure, okay, efficiency == good, but why are we so afraid to ask people for things? Why do we need chatbots to make us feel capable and affirmed? Why do we consider forum posts from real people to be less reliable than a probability machine? Why is understanding being side-lined in the name of speed? Is AI going to arrest any of Epstein’s co-conspirators or bring justice to his victims?
AI is ultimately anti-social. Why have employees when you can have a machine? Why analyze information from experts when you can blindly trust a confident-sounding bot? Why potentially suffer rejection when you can get a system to output text imitating a significant other? In fact, why have friends, when you can just plug in?
“I asked AI”.
Ugh, no! It’s downhill from here if people don’t wake up to what’s happening!
Addendum: Inevitably, there will be people who point out highly successful uses of AI, which is fine, and something I didn’t address in what I wrote above, but without responding wholly to that sentiment, the existence of successful usage is always put forward as a wholesale defense of AI. In my opinion, successful usage is a tiny fraction of overall usage, much of the rest being used for faux productivity, or otherwise being used on the job to “increase efficiency”, which isn’t inherently bad, but it’s generally still anti-social (being an order from the top down, because there’s labor to exploit, dammit!). I don’t think anybody has any idea how these things will affect society long-term, but we see some tremendous downsides already, and only marginal upsides which, in my opinion, are not an effective defense. I fear this will be yet another technology that brings with it some tiny benefits, while dragging the rest of society down. The people most vehemently defensive of AI typically use it the most, although I have yet to see somebody’s life dramatically improved by replacing reality with AI, just as my generation learned social media always skews toward dysfunction, kept only in check with some degree of resistance to engage, almost like saying, “Hey, drugs can make you feel really good, so they’re not all bad, but they can be addictive, so you have to be disciplined”. That’s a campy and incomplete analogy, but you get the point. AI is being used to write emails on one side, and used to summarize emails on the other side, and such is a great example of just how anti-social this stuff is.