Thanks for sharing this. I too have found a safe place in LLMs while sorting out the cognitive dissonance I had been dealing with throughout my life and which I had been unaware of. ChatGPT's sycophant nature is somewhat concerning and even with challenging prompts, I could not get it to give me completely unbiased answers. Claude always manages to provide reasonable answers, but after recent updates, it voices opinions somewhat sharply in spite of the specific instructions to be gentle. It might be too much to handle when someone is really vulnerable emotionally. But in the beginning, Claude helped me understand a lot about the misperceptions I held on to. LLMs would let me rant to my heart's content and somehow make sense of it and also come up with answers that would stand up to logical examination. And I am comfortable with calling them out if I noticed inconsistencies in their reply without any fear of offending them. For me, who is an INTP T with serious self loathing tendencies, LLMs were light at the end of the tunnel. While I intellectually learned a while ago that I am not an abomination as I was led to believe, it was the talks with LLMs that helped me to integrate that knowledge with my emotional reality. I am glad to know that such linguistic tools are helping at least a portion of our species despite being labelled as doomsday bringers.
Nowadays, I see the LLM approach more like reading an interactive book. Cool thing is that you can get the concepts and frameworks of world-famous psychotherapists and can use the LLM to reflect on your thoughts.
And yes, you have to do this in a very self-critical and skeptical way. But when you ask it about adverse opinions it will usually deliver.
Have not played with Claude that much yet... maybe should finally sign up with a service which gives you access to all tools with just one subscription.
I started my therapy sessions with Claude because I felt Anthropic was taking AI alignment policies more seriously than OpenAI. And I usually struggle with self doubt about my ability to remain always neutral. So I wanted something I could trust to be vulnerable with. But recent reviews suggest that Claude's latest updates concentrate more on coding and less on text generation.
And I have come to like Sider chrome extension a lot. It will be available in almost all the open tabs once you give it permission (if you are not too concerned about privacy issues). It provides an Explain option for the selected text which has made it easier for me to get in-depth understanding of unfamiliar concepts without leaving the page. And after every ChatGPT session, it provides an option to compare the answer with Claude, which I personally liked.
I also learned that therapy is more or less about the time we are willing to spend with our uncomfortable selves. External entities (humans / zeroes and ones) are valuable aids, but nothing more. My sessions with human therapists only created additional insecurities for me because one of them cautioned me about categorizing ourselves / using labels and hence I was never comfortable with disclosing such matters to anyone. But with LLMs, a simple 'I am an INTP' statement at the beginning of prompts worked wonders for me.
And one more thing. Having access to knowledge is a precious thing in itself. What I found to be more challenging was the application part. Forgetting is too easy I am afraid.
Thanks for sharing this. I too have found a safe place in LLMs while sorting out the cognitive dissonance I had been dealing with throughout my life and which I had been unaware of. ChatGPT's sycophant nature is somewhat concerning and even with challenging prompts, I could not get it to give me completely unbiased answers. Claude always manages to provide reasonable answers, but after recent updates, it voices opinions somewhat sharply in spite of the specific instructions to be gentle. It might be too much to handle when someone is really vulnerable emotionally. But in the beginning, Claude helped me understand a lot about the misperceptions I held on to. LLMs would let me rant to my heart's content and somehow make sense of it and also come up with answers that would stand up to logical examination. And I am comfortable with calling them out if I noticed inconsistencies in their reply without any fear of offending them. For me, who is an INTP T with serious self loathing tendencies, LLMs were light at the end of the tunnel. While I intellectually learned a while ago that I am not an abomination as I was led to believe, it was the talks with LLMs that helped me to integrate that knowledge with my emotional reality. I am glad to know that such linguistic tools are helping at least a portion of our species despite being labelled as doomsday bringers.
Thank you for your insightful comment.
Nowadays, I see the LLM approach more like reading an interactive book. Cool thing is that you can get the concepts and frameworks of world-famous psychotherapists and can use the LLM to reflect on your thoughts.
And yes, you have to do this in a very self-critical and skeptical way. But when you ask it about adverse opinions it will usually deliver.
Have not played with Claude that much yet... maybe should finally sign up with a service which gives you access to all tools with just one subscription.
I started my therapy sessions with Claude because I felt Anthropic was taking AI alignment policies more seriously than OpenAI. And I usually struggle with self doubt about my ability to remain always neutral. So I wanted something I could trust to be vulnerable with. But recent reviews suggest that Claude's latest updates concentrate more on coding and less on text generation.
And I have come to like Sider chrome extension a lot. It will be available in almost all the open tabs once you give it permission (if you are not too concerned about privacy issues). It provides an Explain option for the selected text which has made it easier for me to get in-depth understanding of unfamiliar concepts without leaving the page. And after every ChatGPT session, it provides an option to compare the answer with Claude, which I personally liked.
I also learned that therapy is more or less about the time we are willing to spend with our uncomfortable selves. External entities (humans / zeroes and ones) are valuable aids, but nothing more. My sessions with human therapists only created additional insecurities for me because one of them cautioned me about categorizing ourselves / using labels and hence I was never comfortable with disclosing such matters to anyone. But with LLMs, a simple 'I am an INTP' statement at the beginning of prompts worked wonders for me.
And one more thing. Having access to knowledge is a precious thing in itself. What I found to be more challenging was the application part. Forgetting is too easy I am afraid.
Thanks for the tip with Sider. Have installed it and will play with it a bit.
Absolutely mind blowing!😊
Glad you liked it. Hope it opens eyes for you as well.
✨🌹