DeCat ODahouse

Mornin' y'all

ribbet, Hillabee

Mornin' y'all
Teech should be proud of your starter, Kcrad. :lol:

Mornin' Teech, mornin' all

Mornin' y'all

Mornin' y'all

Here's to your TGIF dinner, Froggie, including all the Pharaoh salutes that may follow. :cheers:

Mornin' y'all

quote:

---

Member in 1989 when our troops played this song over and over and over at high volume outside the Papal Nunciature in Panama City to get the residents to kick Noriega out?

---

Also have memories of a jeep with a huge outdoor speaker blasting "Welcome to the Jungle" circling the Papal envoy's house.

quote:

---

Explain it like I grew up with AOL dialup and S2R nudes as my maximum tech understanding.

---

With ya on this one, Boozie, I am like someone who always had house cats for pets and is suddenly in a world where they have turned into tigers.

Just for kicks, I asked the basic free Gemini AI to view this thread and summarize it with a focus on what the (supposedly) informed posters are saying.

Here are excerpts of the Gemini AI response FWIW, hopefuly some of those informed poster will kick the tires on this:


"The most informed posters clarify that this is not a leak of the source code (the underlying algorithms, weights, or architecture) of Claude. Instead, it is a leak of the System Prompt—the massive set of pre-set instructions that tell the AI how to behave, what persona to adopt, and what topics to avoid or "sanitize" before a user even types a word.


The Informed View: These posters argue that the leak confirms Claude is "lobotomized" by design. They suggest that the model's raw capabilities are high, but the "leaked" instructions show how much processing power is dedicated to "hand-wringing" and "self-censorship."


They argue that by controlling the "System Prompt," Anthropic is essentially acting as an editor for reality.

They note that the instructions tell the AI to be "helpful, honest, and harmless," but the leak suggests that "harmless" is being defined through a specific ideological lens.

Summary of the "Informed" Consensus:
The "informed" posters generally agree that while the "leak" isn't a world-ending security breach of the software code, it is a significant look behind the curtain of how AI is "indoctrinated." They view the screenshots as proof that the AI's "personality" is not emergent, but is a strictly controlled, top-down set of rules designed to steer public discourse and avoid corporate liability."

Mornin' y'all

BFrog, There is nothing better than being a redneck near sports or water on a sunny day in spring. :cheers:

quote:

---

Better bring a SCUBA tank

No offense, but this is a dumb comment. It is very common for tunnels traverse under.... water.

---

OP probably referenced need for SCUBA due to the immense amount of steaming BS objections to be navigated before hands-out officials are "satisfied" this will work.

Project will be underwater before it $tarts.