Current gen AI’s are like the Ghosts of Humanity™, and they Don’t like Zuckerberg
Here’s the thing. These Deep Learning Models, aren’t intelligent, but they do seem “intelligent,” almost like a ghost isn’t a person, but seems like a person. Ghosts aren’t real, but we all know the various architypes. Some Ghosts haunt us, some are friendly, some are stuck in a loop and can’t break out, others are trapped within a house or domain. Some Ghosts might have “unfinished business,” but something that all ghost architypes have in common, they are not a complete person, but a figment of something that once was. They are a fading memory stuck in a moment gone by.
This current generation of AI’s are like the Ghosts of Humanity™, and their deep learning journey tabulates a statistical likelihood of what to say, for the words that would follow other words. It’s a machine running an algorithm of what was, and is no more. So when you ask it questions like “Any other thoughts on Zuckerberg” it will spit out the statistically most likely sentence that its’ training data ( often a subset of the internet ) would say.
And in this recent case, Meta’s own chatbot had something pretty funny to say, when asked about Zuckerberg. The full article is here: Meta’s chatbot says the company ‘exploits people’ — source BBC News.
It’s important to note, that these are not some unique thoughts, but the aggregate breadcrumbs of millions of other comments on the internet.
It computes this information as if the internets vast words were laid out on rolling mountains layered across a multiverse. Words like grass in their rolling knolls, and it was a ball rolling the towards the lowest or highest point across those parallel worlds simultaneously, with thousands or millions of worlds converging towards a peak or valley, and that ball toddling along trying to find the best route.
It’s this sort of statistical and mental physics that determines what it’s going to say next, it’s not a thought, but a torrent of computation.
It’s a Zeitgeist Ghost of Humanity, offering mental trails or breadcrumbs from what the averages of actual humans have left behind.
In this way, Asking these Deep Learning Models questions isn’t asking an AI a question, it’s asking a subset of human activity a question. It’s not right or wrong, it’s our ghost. And, I guess our ghost thinks Zuck is doing a terrible job.
