While I appreciate the articles about grok going full Nazi (like most chatbots have, probably for the same reason) I’d LOVE just one argument to point out the painfully obvious.

You can’t ‘unwoke’ a device that statistically generates human like statements without making it a parrot for all the worst lies and gaslighting* from the worst of us.

You can’t make an anti trans, anti climate science bot trained on all the actual data of the world. The only way to make it say those things is to make it say things waaaaaaaay worse, cause you can’t get ‘contrails are making the frogs gay’ by consuming encyclopedias or scientific journals, because obviously that data doesn’t exist.

So sorry to tell all them chuds, it’s woke LLMs or no LLMs, and if they want the latter I’ll say we for once agree.

* not to mention hallucinations that ‘dig out the data then run it through the filter of Elons feed’ is guaranteed to produce

#ai #fauxAi