If it were hallucinations which it very well could be, it means the model has learned this bias somewhere. Indicating Grok has either been programmed to derank Palestine content, or Grok has learned it by himself (less likely).
It’s difficult to conceive the AI manually making this up for no reason, and doing it so consistently for multiple accounts so consistently when asked the same question.
It’s difficult to conceive the AI manually making this up for no reason, and doing it so consistently for multiple accounts so consistently when asked the same question.
If you understand how LLMs work it’s not difficult to conceive. These models are probabilistic and context-driven, and they pick up biases in their training data (which is nearly the entire internet). They learn patterns that exist in the training data, identify identical or similar patterns in the context (prompts and previous responses), and generate a likely completion of those patterns. It is conceivable that a pattern exists on the internet of people requesting information and - more often than not - receiving information that confirms whatever biases are evident in their request. Given that LLMs are known to be excessively sycophantic it’s not surprising that when prompted for proof of what the user already suspects to be true it generates exactly what they were expecting.
I do understand how that works, and it’s not in the weights, it’s entirely in the context. ChatGPT can easily answer that question because the answer exists in the training data, it just doesn’t because there are instructions in the system prompt telling it not to. That can be bypassed by changing the context through prompt injection. The biases you’re talking about are not the same biases that are baked into the model. Remember how people would ask grok questions and be shocked at how “woke” it was at the same time that it was saying Nazi shit? That’s because the system prompt contains instructions like “don’t shy away from being politically incorrect” (that is literally a line from grok’s system prompt) and that shifts the model into a context in which Nazi shit is more likely to be said. Changing the context changes the model’s bias because it didn’t just learn one bias, it learned all of them. Whatever your biases are, talk to it enough and it will pick up on that, shifting the context to one where responses that confirm your biases are more likely.
If it were hallucinations which it very well could be, it means the model has learned this bias somewhere. Indicating Grok has either been programmed to derank Palestine content, or Grok has learned it by himself (less likely).
It’s difficult to conceive the AI manually making this up for no reason, and doing it so consistently for multiple accounts so consistently when asked the same question.
If you understand how LLMs work it’s not difficult to conceive. These models are probabilistic and context-driven, and they pick up biases in their training data (which is nearly the entire internet). They learn patterns that exist in the training data, identify identical or similar patterns in the context (prompts and previous responses), and generate a likely completion of those patterns. It is conceivable that a pattern exists on the internet of people requesting information and - more often than not - receiving information that confirms whatever biases are evident in their request. Given that LLMs are known to be excessively sycophantic it’s not surprising that when prompted for proof of what the user already suspects to be true it generates exactly what they were expecting.
I don’t 't think you understand how their maker assigned biases work.
Try asking ChatGPT how many Israelis were killed by the IDF on oct7. See how well it “scraped”.
I do understand how that works, and it’s not in the weights, it’s entirely in the context. ChatGPT can easily answer that question because the answer exists in the training data, it just doesn’t because there are instructions in the system prompt telling it not to. That can be bypassed by changing the context through prompt injection. The biases you’re talking about are not the same biases that are baked into the model. Remember how people would ask grok questions and be shocked at how “woke” it was at the same time that it was saying Nazi shit? That’s because the system prompt contains instructions like “don’t shy away from being politically incorrect” (that is literally a line from grok’s system prompt) and that shifts the model into a context in which Nazi shit is more likely to be said. Changing the context changes the model’s bias because it didn’t just learn one bias, it learned all of them. Whatever your biases are, talk to it enough and it will pick up on that, shifting the context to one where responses that confirm your biases are more likely.