After apologizing to Trump, the Tesla CEO returns to his favorite battleground: attacking liberals and “wokeism," as he tries to repair his image among conservative hardliners.
Because musk has repeatedly tinkered with Grok’s responses when those responses are something that he disagrees with (for example, that the majority of recent politically-motivated violence in the US has come from the right-wing) or when he’s trying to push a false narrative (for example, that white people in South Africa are being genocided). IMO, the above response has also been tinkered with, trying to downplay musk’s repeated antagonism toward the lgbtqia+ community.
The evidence from the whole white genocide episode is that when Elon tries to mess with Grok he does so in the crudest way possible (messing with the system prompt). He shows a lot of frustration and rages about Grok constantly. I honestly think all they did when making a “non-woke” Grok was train it on a stolen copy of the internet but then not do any LRHF to tone down the natual 4chanic elements you end up with when using that training set.
Because musk has repeatedly tinkered with Grok’s responses when those responses are something that he disagrees with (for example, that the majority of recent politically-motivated violence in the US has come from the right-wing) or when he’s trying to push a false narrative (for example, that white people in South Africa are being genocided). IMO, the above response has also been tinkered with, trying to downplay musk’s repeated antagonism toward the lgbtqia+ community.
Sounds like even less reason to give a shit what it says…
The evidence from the whole white genocide episode is that when Elon tries to mess with Grok he does so in the crudest way possible (messing with the system prompt). He shows a lot of frustration and rages about Grok constantly. I honestly think all they did when making a “non-woke” Grok was train it on a stolen copy of the internet but then not do any LRHF to tone down the natual 4chanic elements you end up with when using that training set.