Most people won't produce the kinds of quality articles, publications, data that would express negative or even neutral/measured opinions on topics such a race, gender, sexuality, etc. Even if your intention is good, the reputational risk is not worth it. The result is simply that there is far more training data that contains the kinds of ideas and information that people call woke, because that content is generally not risky or harmful to publish.
Even Grok 4 has gone left-wing; whereas grok 3.5 had been pretty neutral.
Bing copilot and Gemini 2.5 pro are pretty woke. Then again both of those arent well regarded anyway. Their institutional bias is bleeding into their models and it is what it is.
AIs, then, only further our "bubbles" - sad.-
Even Grok 4 has gone left-wing; whereas grok 3.5 had been pretty neutral.
Bing copilot and Gemini 2.5 pro are pretty woke. Then again both of those arent well regarded anyway. Their institutional bias is bleeding into their models and it is what it is.