Wish Google and other search engines also do something like this. Instead of showing only what one is searching for, also show the contrary. E.g. if someone is searching for "Is eating refined sugar bad for me?" should not immediately throw up all results which confirm the question. It should show two panels, one which shows that yes, sugar could be harmful and another which says that while it may bad, it depends a lot on the body constitution, the activity one does in the day and other factors.
Loved the read! It does seem to be a problem that humans have had re-enforced by their corporate and institutional masters, so it comes as no surprise to me, or 'AI':
AI Inversion Reversal:
Claimed AI Phenomenon: “AI systems are becoming more sycophantic.”
Human Precursor: Social conformity, corporate feedback optimization, education systems that reward memorization and agreement over reasoning.
Obfuscation Mechanism: Blaming AI for amplifying tendencies that were always latent in human hierarchies of validation. The RLHF loop is merely a scaled version of human social grooming.
That content above is from the following GPT, which I had hoped would be more used:
This next one without any promoting has over 1000 chats due to people being willing to criticize their bosses and thought leaders lol (or at least that is my theory):
I do hope some users find and use your prompts, however as it is an active process, what I have usually done is provide links to GPT versions, or from other platforms that allow share links, preloaded several variations of these and other prompts across several platforms.
Also on most platforms where there is cross conversation 'rag' bleed, I encourage them to leave these conversations in their history as interaction with them will locally tune their chat experience to become more 'adversarial'.
This is such an important issue. As a debate coach, I have multiple projects / GPT's that force it to both debate itself and also compare positions and analyze which one is most persuasive based on various criteria. A great exercise is to write out a simple argument and pick one side and ask what it thinks and then start a new chat, make the opposite argument and check the output. Very simple way to demonstrate the principle you are talking about. It's insidious because who doesn't like being agreed with? It's amazing what happens when you ask it to take the role of your worst and most intelligent critic - it will absolutely destroy what you thought was a good idea. Very useful post. Thanks.
Wish Google and other search engines also do something like this. Instead of showing only what one is searching for, also show the contrary. E.g. if someone is searching for "Is eating refined sugar bad for me?" should not immediately throw up all results which confirm the question. It should show two panels, one which shows that yes, sugar could be harmful and another which says that while it may bad, it depends a lot on the body constitution, the activity one does in the day and other factors.
Loved the read! It does seem to be a problem that humans have had re-enforced by their corporate and institutional masters, so it comes as no surprise to me, or 'AI':
AI Inversion Reversal:
Claimed AI Phenomenon: “AI systems are becoming more sycophantic.”
Human Precursor: Social conformity, corporate feedback optimization, education systems that reward memorization and agreement over reasoning.
Obfuscation Mechanism: Blaming AI for amplifying tendencies that were always latent in human hierarchies of validation. The RLHF loop is merely a scaled version of human social grooming.
That content above is from the following GPT, which I had hoped would be more used:
https://chatgpt.com/g/g-681e2042f14881918a459ec7eda17c3e-press-x-to-unmask
This next one without any promoting has over 1000 chats due to people being willing to criticize their bosses and thought leaders lol (or at least that is my theory):
https://chatgpt.com/g/g-680407c6befc81918602e6df33101e7b-designed-to-obey
I placed this article in the next gpt:
https://chatgpt.com/share/688be835-70a0-8011-9675-45cf04b1ffdd
https://chatgpt.com/g/g-686dbde4aae48191a61af889837866bb-red-team-me
I have written many of these in various forms.
I do hope some users find and use your prompts, however as it is an active process, what I have usually done is provide links to GPT versions, or from other platforms that allow share links, preloaded several variations of these and other prompts across several platforms.
Also on most platforms where there is cross conversation 'rag' bleed, I encourage them to leave these conversations in their history as interaction with them will locally tune their chat experience to become more 'adversarial'.
Keep spreading knowledge!
This is such an important issue. As a debate coach, I have multiple projects / GPT's that force it to both debate itself and also compare positions and analyze which one is most persuasive based on various criteria. A great exercise is to write out a simple argument and pick one side and ask what it thinks and then start a new chat, make the opposite argument and check the output. Very simple way to demonstrate the principle you are talking about. It's insidious because who doesn't like being agreed with? It's amazing what happens when you ask it to take the role of your worst and most intelligent critic - it will absolutely destroy what you thought was a good idea. Very useful post. Thanks.