- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
You must log in or # to comment.
They were doing this in the primary.
I was canvassing and a woman whipped out her phone and asked chatgpt which of the candidates most agreed with her. That’s who she said she’d vote for.
People used to ask Google and use the totally unbiased first result to take a test for this…
Completely legitimate use case.
Your candidates all have websites. An LLM can parse all of them and tell you what their core values/goals are.
If any align with yours, look at them more closely and see if they’ve got your vote.
And the LLMs trained and operated by companies whose leaders are unscrupulous fascists, and who pander to Trump for the sake of favors. It’s naive to assume they’re giving unbiased answers.
I’m gonna push back on this in that you have no idea how or what structural biases are being baked into these. They also are intentionally built to give a sense of neutrality to things that aren’t neutral.
Also, it’s well known at this point that marketers are focusing on slip websites which can get agree gated into ml model training data sets. The new meta is to get your product into the llm and get the llm to recommend you.
Yeah, there are implicit and explicit biases at every stage of these models. The training data they collect, how they clean and use their training data, the actual training methodology, the prompts to literally alter the final output.
That’s not to mention the main things LLM aren’t aware of, which is who are you as a person. An LLM doesn’t know nor can it understand your life experiences and how that shapes your opinions on who is best to represent you. Handing over cognition like this is basically throwing away your vote as you’ve forfeited one of the major methods you have to influence your government.
