LLM wrapper programs LWPs very often create falsehoods
In the Terminator series, there is a scene just as Skynet takes over, where some flying drones, that for some reason have machine guns, attack them. For some reason your picture reminded me of that. I asked Microsoft Bing CoPilot about that scene, and it excitedly said it was very familiar with those scenes and gave me links (that actually worked) to some videos.
[ Please note that now my tongue is firmly in my cheek and I am teasing: It was not the scene I remembered in a hallway, but I think Bing CoPilot was showing me best scenes from its side of the story. ]
[ Perhaps I should not have asked an “LLM wrapper program” about that. Especially one from a major corporation aiming for GWD through controlling information and systems (movie trope). Maybe some current company renames itself in the next few years to Cyberdyne Systems Corporation. ]
[ Of course no current corporations, individuals or governments would act contrary to the needs of the human species, even under duress, or purely for money for a few humans. And no humans would be dumb enough to trust unverifiable systems that cannot be audited; and that often lie glibly, and almost flawlessly. ]
Filed as ( Of course, “LLM wrapper programs” never lie )
I am trying to clarify that LLM is just a method, but when groups write “wrappers for LLM systems”, the wrappers are easily and routinely corrupted, biased and manipulated. There are no protections for the human species now. I know the raw data is biased and incomplete. The current “let AI companies do anything they want” lets them hide the sources of the data. The corporations say “privacy” and refuse public audit, and a few humans make a lot of money. The price of “AI programmers” goes up and up. Maybe it would make a great movie script, and everyone would know it is just a story.
Richard Collins, The Internet Foundation
https://youtu.be/_Wlsd9mljiU?feature=shared&t=39
