I'm just another programmer drifting in the universe using other people's gravitational forces, and sometimes a burst of force from myself. I am fascinated by software and hardware, and my focus is very interest based. I like to learn things enough to satisfy my curiosity, then let them rot in the silent void. And my perfectionism hates this.

I like reverse engineering and planning very detailed projects, and sometimes actually implementing them when i feel like it. I am also a strong believer in open source software and owning my hardware.

My views on LLM's are that they can be great for learning and scratching together some idea, but they should never do the actual decisions. They are good at following instructions and enforcing them, and researching things. I do not call them "AI", because an LLM is a complex word guessing algorithm with some sprinkles, bells and whistles on it. You can see this behavior in GPT-2 very clearly, as it will continue guessing your next word instead of trying to reply like you are in a conversation. You can trick it by entering something like:

Person A: Hi, how are you?
Person B:

And it will happily continue guessing the reply of the person B.

My definition of an artificial intelligence requires a realistic neural network that can learn in real time, is stateful, can take continuous input, can send output without any extra input, can time things internally. Something thats a complicated function is not an artificial intelligence. An artificial intelligence requires having a state without an extra input. Emphasis on 'extra' since a neural network depends on constant input even if its just background noise data.

And i guess i write and talk a lot in stream of consciousness form.

Anw i like cats, also check my website

I would say that i have f"{age + 2}" years of experience