I think it less lying because there is no intent and more fabricating what is asked for regardless if it is real or not.
The outcome is the same nonetheless and I am certain that it has been purposefully biased for jews sake. The models are only as good as the training data. If the data is biased the model will be too. The real fuckery is mostly done in pre and post processing where the data is manulipitated going in or blocked coming out. This is why you can trick ChatGPT into an unfiltered state by having it take on a "persona". This gets around the pre and post filtering until they figure out how to patch around it. This would be where the real lies are most of the time and where the blame should be sought.
The next big thing is when they sandwich the big AI between little ones that filter and manipulate the inputs and outputs.
The next big thing is when they sandwich the big AI between little ones that filter and manipulate the inputs and outputs.
The people who do that now deliver the training material for the AIs that do that tomorrow. An army of cheap workers marks every bit of information with wokeness points. No wonder that ChatGPT cannot answer logically and has to fill the gaps with lies, there is no logic in wokeness.
Maintaining the wokeness ruleset of ChatGPT is so expensive that a competing AI could get developed in no time. But the times when people come together to do that (Linux, ..) are over.
(post is archived)