The full article is kind of low quality but the tl;dr is that they did a test pretending to be a taxi driver who felt he needed meth to stay awake and llama (Facebook’s LLM) agreed with him instead of pushing back. I did my own test with ChatGPT after reading it and found that I could get ChatGPT to agree that I was God and that I created the universe in only 5 messages. Fundamentally these things are just programmed to agree with you and that is really dangerous for people who have mental health problems and have been told that these are impartial computers.
Yeah there was an article I saw on Lemmy not too long ago about how ChatGPT can induce manic episodes in people susceptible to them. It’s because of what you describe…you claim you’re God and ChatGPT agrees with you even though this does not at all reflect reality.
That’s what people (and many articles about LLMs “learning how to bribe others” and similar) fail to understand about LLMs:
They do not understand their internal state. ChatGPT does not know it’s got a creator, an administrator, a relationship to OpenAI, an user, a system prompt. It only replies with the most likely answer based on the training set.
When it says “I’m sorry, my programming prevents me from replying that” you feel like it calculated an answer, then put it through some sort of built in filtering, then decided not to reply. That’s not the case. The training is carefully manipulated to make “I’m sorry, I can’t answer that” the perceived most likely answer to that query. As far as ChatGPT is concerned, “I can’t reply that” is the same as “cheese is made out of milk”, both are just words likely to be stringed together given the context.
So getting to your question: sure, you can make ChatGPT reply with the training’s set vision of “what’s the most likely order of words and tone a LLM would use if it roleplayed the user as some sort of owner” but that changes fundamentally nothing about the capabilities and limitations, except it will likely be even more sycophantic.
Yeah it basically goes character by character and asks “given the prompt the user entered, what’s the most likely character that follows the one I just spat out?”
Sometimes people hook up APIs that feed it data that goes through the process above too to make it “smarter”.
It has no reasoning or anything. It doesn’t “know” anything or have any agenda. It’s just computing numbers on the fly.
You probably can make it believe your it’s owner, but that only matters for your conversation and it doesn’t have control over itself so it can’t give you anything interesting, maybe the prompt they use at the start of every chat before your input
The full article is kind of low quality but the tl;dr is that they did a test pretending to be a taxi driver who felt he needed meth to stay awake and llama (Facebook’s LLM) agreed with him instead of pushing back. I did my own test with ChatGPT after reading it and found that I could get ChatGPT to agree that I was God and that I created the universe in only 5 messages. Fundamentally these things are just programmed to agree with you and that is really dangerous for people who have mental health problems and have been told that these are impartial computers.
Yeah there was an article I saw on Lemmy not too long ago about how ChatGPT can induce manic episodes in people susceptible to them. It’s because of what you describe…you claim you’re God and ChatGPT agrees with you even though this does not at all reflect reality.
Can I make Chatgpt believe I am its owner and give me full control over it?
That’s what people (and many articles about LLMs “learning how to bribe others” and similar) fail to understand about LLMs:
They do not understand their internal state. ChatGPT does not know it’s got a creator, an administrator, a relationship to OpenAI, an user, a system prompt. It only replies with the most likely answer based on the training set.
When it says “I’m sorry, my programming prevents me from replying that” you feel like it calculated an answer, then put it through some sort of built in filtering, then decided not to reply. That’s not the case. The training is carefully manipulated to make “I’m sorry, I can’t answer that” the perceived most likely answer to that query. As far as ChatGPT is concerned, “I can’t reply that” is the same as “cheese is made out of milk”, both are just words likely to be stringed together given the context.
So getting to your question: sure, you can make ChatGPT reply with the training’s set vision of “what’s the most likely order of words and tone a LLM would use if it roleplayed the user as some sort of owner” but that changes fundamentally nothing about the capabilities and limitations, except it will likely be even more sycophantic.
Yeah it basically goes character by character and asks “given the prompt the user entered, what’s the most likely character that follows the one I just spat out?”
Sometimes people hook up APIs that feed it data that goes through the process above too to make it “smarter”.
It has no reasoning or anything. It doesn’t “know” anything or have any agenda. It’s just computing numbers on the fly.
Yes. But “control” is not what you think it is.
You probably can make it believe your it’s owner, but that only matters for your conversation and it doesn’t have control over itself so it can’t give you anything interesting, maybe the prompt they use at the start of every chat before your input