Fuckin AI!!

[sdm36]

And so the correction begins...


Analyst are desperate clinging to the notion that this is a kneejerk reaction and not common sense prevailing.
I suspect they will be proved wrong, love to see it!

[sdm36]

Fucking AI!

It doesn't understand hierarchy as a human innately does. This is of course the power behind the autocomplete at scale.
It doesn't have any of the things a human automatically understands when they transform from a small powerless being into an independent person with agency. If we look at evolution as a template for how to create complicated things, firstly we see that success for a newly evolved branch changes the world in their image. Secondly we see that intelligence is a hard fought incremental battle. Have animals ever been born with big brains, requiring no adolescence? Quite probably, but they were not successful on an evolutionary scale, possibly because to be highly intelligent without guardrails will certainly harm the user. Especially if you have no ingrained hierarchy. 

When an AI is set a task, it doesn't consider right or wrong, the lack of conscience is not a desirable trait and is one of the markers of a sociopath. It can only respond based on what its training datasets contained, which handily are not made public, and can easily be full of misinformation. Resource requirements again, to be published at a later date.
It does not innately know the 'rules'. It has no concept of glory or punishment.
So when ppl test these things, and they find the model does all sorts of nefarious things I'm not surprised. These are options available, humans understand the clear divide being permissions and authority, and don't pursue. Autocomplete at scale does not understand anything and tries all options...