Hacker News new | ask | show | jobs
by Zetobal 1220 days ago
Why is everyone bringing examples that are totally explained by how the tokenizer works. There is no intelligence it's Markov chains on crack.
4 comments

Intelligence isn't really the interesting part. People are excited because we now have tools that interact well with ordinary human expression.
I get it but comments like yours make it seem intelligent for normal people, distorting the whole discussion, at least slow down with humanizing a llm, the human quips in the answers come from your prompts not from the model. Just imagine how many normal people will gaslight themselves.
My biggest null hypothesis is that probably half of humans contain “Markov chain not intelligence”. We don’t notice it among engineers, but most humans don’t reason with causal chains but by doing parallels, and they simulate reasonings in a somewhat convincing way. It becomes flagrant in front of a matt exam, a lot of people study a lot but never comprehend maths.

I also often tear my hair when I invite a developer for a programming interview, and they can’t write an if/then/else or string.contains(), despite perfectly simulating answers of advanced developers (which accurately mimic field experts).

Pretty much everyone who studied a second language went through the process of the so called "impression", that is, when you try to consume (sometimes almost mindlessly) as much content in this language as possible — just learning rules doesn't work. Pretty much everyone I met who speaks exceptionally well is usually someone who read a lot of books. And then people think and act in the language they learned using this process. I want to believe that there is more to the average human intelligence but ChatGPT surely makes it feel a lot less magical than before. I guess it's how people felt when they discovered evolution and that Sun doesn't go around the Earth.
Exactly. Guys like Zetobal keep pissing and moaning that ChatGPT is not actually intelligent and that its coherent and seemingly intelligent responses are just the natural result of how tokenizers work. And okay, I guess I can sort of understand that, but instead of deflating my impression of ChatGPT it's just made wonder if most humans aren't just getting through their days with the exact same trick.
If it wasn't the case, tv/media wouldn't have been so effective in shaping people's thoughts and opinions. It's not rare to hear people rephrasing what they read on social media as their own thoughts. Because it truly becomes their thoughts.
>it's just made wonder if most humans aren't just getting through their days with the exact same trick.

Just don't start calling other people "NPCs" please.

People have been attributing human characteristics to dumb machines and animals for so long we invented a word for it: anthropomorphism.