|
|
|
|
|
|
by fabbari
1242 days ago
|
|
|
I have a point to make, that I'm not totally convinced I believe in myself. Take this as a devil's advocate argument. Nick Cave - correctly so - says that you can't write if you haven't experienced something to write about and you need to have an inner self that experiences things to do that in the first place. But AI being trained on a vast amount of human writing, writing that was the consequence of the life experience of millions of selves, could be considered - if not a 'self' - as having experienced all the digested human experiences. The mockery - the pastiche effect - is the result of asking ChatGPT to write something in the style of 'Nick Cave'. I bet it would read the same if you asked anyone to write something 'in the style of...'. I don't honestly know. |
|
|
Consider that, if the AI "actually" has experienced all the digested human experiences, the vast majority of them -- even those, in the phase space of human experiences, "close" to Nick Cave, that the AI might utilize in order to seem most like Nick Cave -- are not those of Nick Cave, and detract from the machine's ability to authentically impersonate Nick Cave just as Nick Cave's writings add to that ability.
But, like, I don't think it actually has experienced all that stuff! Don't fall into the trap of thinking that human language correctly encapsulates (or even comprises!) human thought. You'd be in good company -- great philosophers like Bertrand Russell made that error, too. But human language is a messy abstraction for Something Else that's going on while we think. I can read inspiring or evocative words, but I don't really think that even the most vivid description can give me the visceral experiential knowledge of what it's like to climb Everest, or whatever. I can, however, read a lot about Everest and the experience of climbing it, and then write something that might trick readers into thinking I speak from firsthand experience. I suspect that this is closer to what the AI is doing.