Everyone Dies

Before you dive into the meat of this post which is the Youtube video linked at the bottom I just want to remind you that AI is incredibly susceptible to being wrong. Sometimes on some of the most basic things that you would expect a super brainiac computer to get right. We assume that it's calculating things in the way that a calculator or a program used to, but it uses fundamentally different techniques, not all of which we understand.

Let me give you an example here. I've been experimenting with several different remote AI setups, and one of them is Perplexity Computer. I have a Mac subscription to the Perplexity service, which includes a personalized news feed in it.

The night in question, a week or so back, the Knicks were playing the first game of their series with the Cleveland Cavaliers. Like many of the recent playoff games, it was a blowout. The Cavaliers were up by a lot. If I remember right, they were up like 22 points with 7-ish minutes left and had less than a 1% chance to win the game.

I must've been really bored, so I kept watching, and the Knicks started to storm back slowly but surely. At the end of the game, it was tied. Sam Merrill of the Cavaliers (who played college ball at my alma mater, Utah State University) had a great look at a last-second three, but it somehow rimmed out.

The Knicks went on to crush them in overtime and have since swept them in the Eastern Conference Finals.

So imagine my confusion when reading the headlines in Perplexity that night that presented the following:

If I hadn't watched it and seen it with my own eyes, I honestly would have totally believed that's what happened. But that isn't what happened. Sam's shot missed. He wasn't the hero, and the Cavs have now been swept. So I pushed back pretty hard and got the classic AI response when it is blatently wrong: "You're absolutely right." And then some driveled apology.

The problem I encounter on a daily basis in using AI is that no matter the answer it gives you, it is supremely confident in it! It will literally lead you confidently off the cliff, and then be like, "Sorry, bro, my bad."

It also happens to be insanely easy to trick, where it will take in data that is false and then send it back like it's absolute truth. Here is a great example of that:

So the point is to be very, very careful about what you trust as far as information that is coming from an AI model. This truly is the time where we have to get very spiritually in tune to the spirit so we can discern truth, because just logic is not going to be enough in the future.

The major problem with this is that it's infiltrating most of our information feeds. From the news to social media, AI content is everywhere. It's just too easy and cheap to produce right now, so you can expect the AI slop to just get worse and worse.

One promise I make to you here and now is that all content on this blog, now and in the future, is and will be 100% written by me. I take full responsibility for any egregious spelling and grammar errors or even the use of an em dash. I may use an AI model to help brainstorm or research, but I do all of the writing and thinking that goes into it.

There is a reason why there's a disclaimer right at the bottom of every AI input that says something like "AI can make mistakes." Yes. Yes it can.

Now on to the promised part where everybody dies... I don't want you to think that was just clickbait. 😉