Obviously it would be someone entering queries and such but the idea that gov and others have better tech than currently public plus people thinking it was an a.i, perhaps it was a group of people asking a chatgpt or they might call theirs Q
Comments (13)
sorted by:
No.
It's possible that they used a GPT AI (which was far less advanced when Q posts started) COULD relate to crafting messages in a way that avoids detection. However, there's no possibility that Q posts were made with an AI independently while maintaining the deltas, proofs, etc. that stretch out into 5 years from original posts.
Why not?
Fair question. Absent some super-intelligent general AI housed on some super quantum computer that's far beyond what's publicly known.
The reason has to do with what and how AI learns and the limits on what can be learned. At its simplest level, an AI is a map of neurons. Each neuron accepts inputs multiplied by a weight (the weight represents what is learned and gets adjusted during training) and the outputs are a result of an activation function (a non-linear equation to determine how much that neuron is activated). The more complex the map, the more intricate things it can learn.
Then there are the limitations; a few years back there was an image recognition AI that those involved wanted to test to figure out what it was actually learning, so they chose to check how it would distinguish between a wolf and a german shepherd. The result was that wolves were called wolves when there was snow in the background and a german shepherd if there was grass... IOW, the distinction was nothing relating to the animal itself.
To close the circle, an AI could be used to craft posts in ways to obfuscate the original author, but absent it being a general intelligence that also understood human nature, cause-effect relationships in the political sphere, a grasp of how optics of events get filtered by people in different spheres while providing double meanings and presenting some as proofs of validation that continue on several years out. It's a challenge that is orders of magnitude more complex than any currently known AI system. Literally, the time travel theories are more probable than anything that could be extracted from all intelligence sources fed into it in live time.
Would you be able to see an advanced enough neural net be used in conjunction with experts in different fields to ask questions about specifics, summarise large info into digestible pieces, find reference for and to within the past, zeitgeist (described to it for understanding if needed) and the media & further as you added obfuscate the author/s while retaining the main info. So less a general a.i giving the answers the people but... almost acting as a p.r for the white hats that work with/under/for Q (the person/organisation)
A tool rather than a leading intelligence
Short answer, yes, that seems plausible.
That's where I struggle to believe that human minds could make plans that far out and accurately enough to where they could have Q posts where people are finding 'proofs' 5 years out now. (Unless they happened to encode the drops such that they are leveraging the gambler's fallacy, where people tend to focus on the hits, and the misses get ignored) That kind of plan would need to have branches for various contingencies, counter-moves, delays, etc. where the further out things get the less likely that things would remain relevant to anyone who is given aspects of that plan.
This is why thinking through what would be needed, looking glass tech or actual time travel becomes the most likely tools that could meet the needs of making a plan AND maintaining a level of certainty in outcomes to be able to lay it out for people outside the plan and to show "future proving the past drops" 5 years out now.
chatgpt = AI?
Yes, GPT is an AI. I forget the GP part, the T means "transformer" which is how that AI learns to understand the context of text.
Dog bites man vs Man bites dog. Token wise are identical, add key and value pairs to the text read and it might be 1, 2, 3 vs 3, 2, 1 which would be distinct sequences that could have different learned meanings.
Thanks for the explanation. Appreciate.
No problem, that's about as simple as I can explain without needing what might be a chapter of a book. I also like explaining the topic from the perspective of a modestly informed amateur (my home computer could really only replicate experiments that are now 30+ years old)
The previous approach was more complicated, but is what's still used on phones for autocorrect, where the AI "Learns" the token (word) and words that are generally connected with that word ("is a" would be an example of commonly connected) and then a "forget" zone because you generally want to write something like "a tent is a tent", so, it was inferred context mainly because you don't generally repeat.
I was thinking that today as the whole tech world is talking about it.
But I was thinking intel agencies probably have similar tool but have access to way more non public data for it to sift through.
https://www.theguardian.com/technology/2022/dec/06/meet-chatgpt-the-viral-ai-tool-that-may-be-a-vision-of-our-weird-tech-future
some background info today, from a UK legacy media rag.
Why don't you define chatgpt??? Can't be chat group with that "t' at the end.
Apologies. It is an a.i chat bot, Google 2.0 basically, thought it was popular enough to be known at the time