Love this, swallowed the first third in one gulp, but have to go finish performing feminine productive and care labour, will come back to finish it. Thank you and ✨✨✨
I really enjoyed seeing your conversation with Claude. It reminds me of the lengthy back-and-forth I'm having with ChatGPT as I prepare my own Substack essay each week. The funny thing is that Claude and ChatGPT have exactly the same kind of supportive, friendly tone-out-of-a-box no matter what the topic. It's eerie. I can ask it about Victorian London, or existentialist philosophy, or painters whose paintings might be relevant to what I'm writing, and its cadences are exactly like what you've put here. Sounds like a super-supportive therapist, or the chillest friend in the universe.
One thing I think Claude did well here on a substantive level was to highlight the importance of the data on which it was trained. Any LLM trained on the internet will be subtly or unsubtly misogynistic, given our culture as reflected on the internet. So the training inputs will impose a major challenge to the notion of AI as a feminist technology.
There are some issues also with regard to the notion that the nondeterministic, black-box nature of AI makes it somehow less controlled or less engineered. I'm not sure it does. It just means that the algorithm can sometimes pick a less likely next chunk - but it's still well within both its algorithm and its training. Certainly, though, it's non-determinism is a plus in using LLMs for creative work and reflection.
Love this, swallowed the first third in one gulp, but have to go finish performing feminine productive and care labour, will come back to finish it. Thank you and ✨✨✨
I really enjoyed seeing your conversation with Claude. It reminds me of the lengthy back-and-forth I'm having with ChatGPT as I prepare my own Substack essay each week. The funny thing is that Claude and ChatGPT have exactly the same kind of supportive, friendly tone-out-of-a-box no matter what the topic. It's eerie. I can ask it about Victorian London, or existentialist philosophy, or painters whose paintings might be relevant to what I'm writing, and its cadences are exactly like what you've put here. Sounds like a super-supportive therapist, or the chillest friend in the universe.
One thing I think Claude did well here on a substantive level was to highlight the importance of the data on which it was trained. Any LLM trained on the internet will be subtly or unsubtly misogynistic, given our culture as reflected on the internet. So the training inputs will impose a major challenge to the notion of AI as a feminist technology.
There are some issues also with regard to the notion that the nondeterministic, black-box nature of AI makes it somehow less controlled or less engineered. I'm not sure it does. It just means that the algorithm can sometimes pick a less likely next chunk - but it's still well within both its algorithm and its training. Certainly, though, it's non-determinism is a plus in using LLMs for creative work and reflection.