That’s what humans do, too.
Discussion
If that is what humans do, where did the old thoughts come from?
I get your point. I was trying to say that most of our thoughts are variants and remixes of ideas we have heard before. Culture in general draws very heavily from what came before.
The argument of „no reason to worry because it‘s just a language model that remixes old thoughts“ makes no sense because thats 95% of what we do as humans too. Unless you want to argue that text itself cannot be dangerous.
What’s the most dangerous piece of writing ever produced by humans? Was it wholly original? Or did it draw from something that existed before?
You can convince people to do pretty bad stuff with only text and old thoughts.
Not to say that Elon‘s letter makes any sense. That train has left the station long ago.
Perhaps I should have said that LLMs draw very heavily from what came before without applying any thought to what it draws. LLMs have no opinions, no agenda (other than those the trainers intentionally inject), no ability to reflect, no model of personality in order to assess the intent/ideology of the reader.
What LLMs do is regurgitate without thought.
>From: (06ee745...) at 03/31/23 09:22:17 on wss://relay.damus.io
>---------------
>I get your point. I was trying to say that most of our thoughts are variants and remixes of ideas we have heard before. Culture in general draws very heavily from what came before.
>
>The argument of „no reason to worry because it‘s just a language model that remixes old thoughts“ makes no sense because thats 95% of what we do as humans too. Unless you want to argue that text itself cannot be dangerous.
>
>What’s the most dangerous piece of writing ever produced by humans? Was it wholly original? Or did it draw from something that existed before?