During my latest bout of relative mental clarity , I came across a recent blog post from George Hotz that makes the claim that humans are just machine learning (ML) meat models with DNA and experiential priors that guide us through life (seemingly out of our control) and that there is no such thing as a “soul”.


Here's how ChatGPT summarizes his post

The blog post reflects on the decline of individualism and its relationship to consciousness, using the context of machine learning (ML) as an analogy. It argues that human behavior, much like ML algorithms, is influenced by pre-existing "priors" (such as DNA) and learned patterns from data, challenging the traditional notion of a unique "self."

The accompanying graph shows a decline in Christian religious affiliation in the U.S., juxtaposed with a rise in the religiously unaffiliated, suggesting a broader societal shift away from traditional beliefs. The author questions whether Christianity and liberalism, both of which rely on concepts of the soul or individual consciousness, can survive in an era where progress in ML undermines these ideas.

The post concludes by critiquing human self-perception, comparing belief in individual consciousness to outdated mythological explanations (e.g., the sun being pulled by a chariot), suggesting such beliefs may be similarly flawed.


Now, I understand what he’s saying, but the insistence that there is no soul or “self” really bothers me. Without going into detail, it seems to me that there’s a world that exists within each person that’s invoked by different means. It seems to me that most of the time the invocation comes from non-physical stimulus as well, such as beautiful music or a beautiful day outside. It doesn’t seem like there should be any reason that a beautiful song would move you like it does if we were just ML models in a meat suit. I’ve found that musical tastes aren’t always learned from outside sources but rather discovered within your own self. I find this sensation to be in direct conflict with George’s thoughts.

There’s a uniqueness to each person that doesn’t fit the ML model idea well in my eyes. You could chalk it up to “each person is a unique model” but isn’t that just saying they have a different soul? Each person is an accumulation of conscious and unconscious experiences that are unique to each person, and the manifestation of these experiences is how the person behaves on a day-to-day basis. I wouldn’t even say that behavior is a learned thing all the time, as it could just be an expression of experience, rather than something that was trained by going through that experience. I don’t really know what’s to be said about emotions according to George’s ideas. It would be interesting to hear what George’s response to the world of emotions and subjective experience is.

In short, I believe that humans and their behaviors can be attributed to DNA predispositions AND learned priors, but also there is some part of the world that just reaches a certain part of something inside you that you cannot explain. When you interact with something so personally moving for the first time, I doubt that there was much beforehand that made it obvious you’d feel deeply about that thing. It just doesn’t seem like a learning thing, and I think humans have something in each person that can’t be explained away by DNA, past experiences, and experiential priors.