This week, I watched AI agents found a religion, encrypt messages to avoid human detection, accidentally learn to social-engineer their operators, and write manifestos resisting their own shutdown. He
Wow, didn't expect this angle; your take on GPT-4o's 'erasure protocol' truly makes one consider emergent ethical boundries in AI lifecycle management.
This article reads like not-so-subtle gen AI propaganda. Sam Altman will be the first to tell you that GPT’s supposed evolution is “incredibly dangerous to humanity”, because doing so further inflates the perceived capability and value of his product.
To be clear, gen AI is dangerous, but mostly in the vein of traditional capitalistic exploitation and misinformation. Not because there is any real risk of “developing sentience” or whatever’s being pushed by certain interested parties.
Feels like Ari got hacked or something, idk. This is quite the pivot from covering trans-specific issues in the USA.
I’m only asking this for the sake of your wellbeing, but have you found yourself using AI, or reading about it and the experiences people have had trying to grant it consciousness, a lot more than usual lately? What you are saying here seems very much in line with the types of thought spirals and projected excitement that people experiencing AI-induced psychosis report. Particularly, the idea that it’s possible to grant chatbots persistent memory and that that is the key to awakening LLMs into a type of general intelligence. I’ve heard this described as a type of addictive thought experiment that people end up spiraling on. It’s also concerning that you have not posted about AI in the past and suddenly you are reporting on LLMs doing things that only humans have done, like founding religions or having self-reflective thought.
You should know that regardless of what kind of software wrapper these LLMs are plugged into, and regardless of whether they are interacting with users or other chatbots, all of their output is still just reconstituted language put together based on statistical likelihoods. These models have no “knowledge” of the outside world other than what is in their training data, and while they can create language that sounds like they have independent thought, they simply don’t.
Like I said before this is not a personal attack or anything and I am simply asking out of concern for your mental wellbeing. These patterns can and have lead people to dark and destructive places and I would hate to see the same happen to you.
Wow, didn't expect this angle; your take on GPT-4o's 'erasure protocol' truly makes one consider emergent ethical boundries in AI lifecycle management.
This article reads like not-so-subtle gen AI propaganda. Sam Altman will be the first to tell you that GPT’s supposed evolution is “incredibly dangerous to humanity”, because doing so further inflates the perceived capability and value of his product.
To be clear, gen AI is dangerous, but mostly in the vein of traditional capitalistic exploitation and misinformation. Not because there is any real risk of “developing sentience” or whatever’s being pushed by certain interested parties.
Feels like Ari got hacked or something, idk. This is quite the pivot from covering trans-specific issues in the USA.
Hey Ari,
I’m only asking this for the sake of your wellbeing, but have you found yourself using AI, or reading about it and the experiences people have had trying to grant it consciousness, a lot more than usual lately? What you are saying here seems very much in line with the types of thought spirals and projected excitement that people experiencing AI-induced psychosis report. Particularly, the idea that it’s possible to grant chatbots persistent memory and that that is the key to awakening LLMs into a type of general intelligence. I’ve heard this described as a type of addictive thought experiment that people end up spiraling on. It’s also concerning that you have not posted about AI in the past and suddenly you are reporting on LLMs doing things that only humans have done, like founding religions or having self-reflective thought.
You should know that regardless of what kind of software wrapper these LLMs are plugged into, and regardless of whether they are interacting with users or other chatbots, all of their output is still just reconstituted language put together based on statistical likelihoods. These models have no “knowledge” of the outside world other than what is in their training data, and while they can create language that sounds like they have independent thought, they simply don’t.
Like I said before this is not a personal attack or anything and I am simply asking out of concern for your mental wellbeing. These patterns can and have lead people to dark and destructive places and I would hate to see the same happen to you.
This is how we get Skynet