I think I might be a Large Language Model. At least partially. You probably are too.

Most of us know what (who) Chat GPT is. Open AI’s Large Language Model. For those that don’t, some call it autocorrect on steroids, and others have shown it seems to understand what we say to it. It is a system that takes in words, chews on them, then spits out its own words. (I won’t go on, because many have written clear explanations of LLM’s and Chat GPT’s capabilities).

I was thinking about how an LLM has no consciousness while it chews on the words (we assume). It is given a prompt, which is just a set of words that define context and what type of response is required, and it spits out the response.

Sometimes it spits out words that sound very reasonable, but it is making stuff up. This most often happens when it is asked facts. This hallucination issue is why one should treat LLM’s as analysts, not as search engines.

So while I was thinking about this I realised that I do the same thing as an LLM when I’m writing. The words you see here aren’t carefully selected by some conscious part of my brain for me to consider carefully before typing. The words here appear at the same time as they appear in my head as a voice and on my keyboard as finger patterns. I am not conscious of the act of typing other that I know my fingers are moving. I don’t consciously spell anything to myself, most words are patterns I’ve used many times and my fingers just do them.

Sometimes I read what I’ve written and realise I want to say it a little differently, or that I’ve made a typo (where my finger pattern went for a word I’ve used more than the word I wanted I guess so my brain triggered the wrong finger type pattern), and I go back to that bit and correct it.

But this all happens after what I’m writing is written. Not before. My brain is spitting out words right now and while I do this my mind is conscious of the roadworks happening outside and of the tinnitus in my right ear (very annoying and quite loud today).

So I wonder this, am I partly just a Large Language Model? Have I been trained to spit out words in certain orders because of the tens of millions of words I’ve already taken in? I’m mid fifties, so I’ve only got maybe a 100 yrs left on the planet. I’ve been reading solid for most of the fifty years. I could not read very well when I was first born, which suggests my ability with words has been acquired. Trained into the structure of my brain.

And the part that fishes out the words isn’t really a conscious part. The words just appear. Some part of me must be deciding what words to pick, but that part of me is the same part that talks about how terrible the 60th Anniversary Doctor Who is.

There is no effort on my part that I am aware of to pick the words that express my thought. And the thought itself is happening with the words. The two are bonded together. The thought itself is a construct just like Lego can be turned into a castle. Words are my internal Lego. My software.

I imagine my dog’s limited mental Lego is one of the main reasons he operates somewhat differently in the world to what I do. His inability to type would not be the limiting factor on his communication. I do know that if you give dogs buttons that each trigger a word, that they tend to demonstrate a level of thinking after a while, and I assume that is because the buttons give them ways to encapsulate their thought. To Lego together an idea.

I think I might be at least partly LLM.

I know I hallucinate when sharing facts. My brain isn’t the best place for dragging facts out. Some facts that I am certain of, turn out to be my hallucinations. LLM mode.

So that’s my thought for today. Are you part LLM? And if my brain’s function is part LLM, then when we get the other parts that make a brain built, and we wire them together the same way, might we not end up with an artificial consciousness? Hmmm. We may never be certain, no matter what the machine may tell us.