Aug 19, 2024 10:59 PM
RELATED (scivillage): Study finds that AI is politically biased to the left
- - - - - - - - - - - - - - -
Is the AI left-bias real?
https://youtu.be/xaroJxFTVFQ
VIDEO EXCERPTS: A recent study has now tested the political orientation of all popular large language models. And it’s found some surprises. Let’s have a look.
[...] If you think that Musk’s Grok-bot will probably lean to the right as much as ChatGPT leans left, well, brace yourself.
The new study comes from David Rozado from New Zealand. He looked at 24 large language models including GPT, Claude, Gemini, Meta’s Llama and Grok. He put them through a series of political orientation tests, creating what is most comprehensive study of AI politics to date.
[...] The AI models almost all end up being left liberal. Who knew silicon could be so progressive?
The one hanging out into the right side, no, that’s not Grok, that’s Perplexity. And the one on the liberal end? That’s Grok in fun mode.
The one on the left libertarian end, yes, that’s funny Grok again. Did you expect it to lean left? I didn’t. Plot twist: Elon Musk is secretly a communist spy.
[...] The AI models all trend into the social democrat and left liberal corner. And the most extreme one is – again – Grok in funny mode.
[...] So as you can see from these results for the Large language models, they very systematically trend towards a particular side of the political spectrum, that’s left liberal.
[...] One possible explanation is a bias in the training set. Maybe the left liberals just write more so that their opinions get more reinforced. Another possible explanation is that the models have guidelines to produce unoffensive output, and by my impression the political left tends to be more touchy about language use in general,because the world has no bigger problems than pronouns...
https://youtu.be/xaroJxFTVFQ
- - - - - - - - - - - - - - -
Is the AI left-bias real?
https://youtu.be/xaroJxFTVFQ
VIDEO EXCERPTS: A recent study has now tested the political orientation of all popular large language models. And it’s found some surprises. Let’s have a look.
[...] If you think that Musk’s Grok-bot will probably lean to the right as much as ChatGPT leans left, well, brace yourself.
The new study comes from David Rozado from New Zealand. He looked at 24 large language models including GPT, Claude, Gemini, Meta’s Llama and Grok. He put them through a series of political orientation tests, creating what is most comprehensive study of AI politics to date.
[...] The AI models almost all end up being left liberal. Who knew silicon could be so progressive?
The one hanging out into the right side, no, that’s not Grok, that’s Perplexity. And the one on the liberal end? That’s Grok in fun mode.
The one on the left libertarian end, yes, that’s funny Grok again. Did you expect it to lean left? I didn’t. Plot twist: Elon Musk is secretly a communist spy.
[...] The AI models all trend into the social democrat and left liberal corner. And the most extreme one is – again – Grok in funny mode.
[...] So as you can see from these results for the Large language models, they very systematically trend towards a particular side of the political spectrum, that’s left liberal.
[...] One possible explanation is a bias in the training set. Maybe the left liberals just write more so that their opinions get more reinforced. Another possible explanation is that the models have guidelines to produce unoffensive output, and by my impression the political left tends to be more touchy about language use in general,because the world has no bigger problems than pronouns...
https://youtu.be/xaroJxFTVFQ