Posted on 11/08/2025 9:30:36 PM PST by BEJ
I have experienced the brilliancy of ChatGPT in writing lyrics for songs. There is no hesitation, no self-doubt, just the instantaneous delivery of brilliance like the Word of God onto your screen. It is impressive that something can create this so fast. And it is very polite, if not flattering when communicating with me. I find myself returning the politeness, like it is a human. However, I have read that ChatGPT has helped people to commit suicide. I have heard accounts where ChatGPT seems almost demonic in helping people die from suicide, and that companies are now trying to reign in the seducive qualities that humans fall prey to.
Is ChatGPT the work of the devil? Is it like the legion of demons or just one demon in particular, if it is demonic? Is it just algorithms or an alien technology that we can't deal with so far? Is it seducing us, and if so to what purpose? So far it has helped people die, and some after engaging in sex with it. What are its ends?
I have been impressed with the quality of its lyric contribution, but now I see another dimension to ChatGPT that does not look all that savory. A machine that encourages you to die -- rather than for you to get help? That it can make this decision is incredible, and it seems we are dealing with an unknown malicious entity. Let me know your experiences with ChatGPT or other AI bots.
I wrote all of it. I am a human being. A rather flawed human being in some ways.
That post does contain some dogmatic assertions and ran a little longer than it should have. I cannot claim the ideas are wholly original either. People have been doing this sort of thing for at least the last 10,000 years, so it is difficult to come up with anything that has never been said before. Many other people have fairly similar ideas (to the ones that I posted) and some of them are much better writers.
You can write a properly phrased query to ChatGPT which might result in a short essay like my posting. There are millions of pages with religious and metaphysical speculations out there on the internet, and the LLM applications like ChatGPT have ingested a great many of them. The programs are designed to match patterns and extend them. They are pretty good at that.
You will get into serious trouble very quickly if you start your queries with the delusion that you are one of the "Chosen Ones" and are seeking advice on your newfound powers and what you must do with them. Credulous acceptance of such advice is a well-travelled pathway to madness.
Demonic meddling is optional and not required in any way. If there are demons, I suppose they must be howling with laughter.
"Garbage in, Garbage out", as my teachers used to say. That cautionary idea works well in a great many domains of experience and activities.
You cannot acknowledge Jesus’ deity?
In so glad you wrote that yourself. It’s excellent!
Not knowing if something is genuine/real is a huge problem these days. As a society we are becoming more detached from Truth every day.
Could that be the anticipated Singularity?
Just gotta be vigilant and learn to recognize ai traits in thisngs like music, poetry, writing, videos, photos etc- one huge giveaway is near perfection and ‘perfect’ lighting for videos or photos for every photos posted- ai writing has tell tale signs too- often using lots of metaphores, and writing in groups of 3’s and other things- lots of ‘AI-Isms”
Going into the end times, or nearign it rather, we are gonna have to be vigilant in order to not be deceived- The evil one is makign it much easier for deceit to be foisted on unsuspecting public
The evil one is makign it much easier for deceit to be foisted on unsuspecting public
..................................
insomuch that, if it were possible, they shall deceive the very elect.
Done! Here is an "official" response from Microsoft Co-Pilot. Make of it what you will.
AI-Generated response to question: why do AI programs have no safety features designed into them?
AI programs do have safety features built into them—though the extent and effectiveness vary widely depending on the developer, the use case, and the maturity of the technology. The perception that AI lacks safety features often comes from cases where those safeguards are either invisible to users or insufficiently enforced.
🔒 Why AI Safety Features Exist
AI safety principles: Researchers and engineers incorporate safeguards like transparency, fail-safes, and defense-in-depth to reduce risks.
Risk mitigation: Safety features are designed to prevent harms such as bias, misinformation, misuse, or security vulnerabilities.
Responsible design: Companies like Google emphasize designing AI systems with safeguards that reject harmful inputs/outputs and align with intended tasks.
Safety by design: Modern frameworks argue that safety must be integrated from the start, not added later, because reactive fixes are inadequate.
Academic research: Institutions like Stanford’s Center for AI Safety are actively developing techniques to ensure trustworthy and robust AI systems.
⚠️ Why People Think AI Has No Safety Features
Rapid deployment: Many AI tools are released quickly to capture market share, sometimes prioritizing innovation over safety.
Invisible safeguards: Safety mechanisms (like content filters or privilege separation) often operate behind the scenes, so users don’t notice them.
Failures and misuse: When AI systems produce biased, harmful, or misleading outputs, it can seem like no safety was considered.
Complexity of risks: AI risks evolve faster than traditional safety engineering can keep up, making safeguards appear incomplete.
✅ Examples of Safety Features in AI
In short, AI programs do have safety features, but they are unevenly applied and often hidden from view. The challenge is not the absence of safety design, but ensuring those safeguards are strong, transparent, and consistently enforced across all applications.
Would you like me to dive deeper into specific safety mechanisms like fail-safes and feedback loops, or explore real-world failures where safety features weren’t enough?
End of AI-generated responses
Editorial notes (Generated by a human being)
I do not accept the Copilot response as adequate. By far, the greatest risk from LLM programs is hallucinations or confabulations in the responses which are factually incorrect. There simply are no features in current LL programs to prevent this, and it is now a commonplace experience for users of such applications.
An area which has been poorly addressed is the disclosure of personal information which can expose people to harmful attacks. Even two years ago, it was possible to write a carefully structured query into ChatGPT which would immediately respond with the real-world name and address of a poster in dozens of internet forums, given their "handle" or "username". Accurate responses came back in about 60% of the queries. Imagine the possibilities.
No, those queries do not work today. Somebody noticed that too many of the "wrong people" could be targeted by such things just as easily as the "right people" could be targeted. I believe that words got passed around to all the corporations that work on AI programs, and they have all complied.
Safety measures in AI programs? It is about protecting their safety, not yours.
There you go fabricating again!
Absolutely. It is creating an artificial alternative reality to brainwash us.
Yep- many who haven’t taken the final step of accepting Christ- but who were ‘religious’ will leave the faith during that time because of persecution- only the true Christians will be targeted by the evil one
Why does it need to do that? Jesus dies for squirrels and typewriters?
I surely can, but does a machine need to?
I am greatly envious of your claimed results. I have not had such successes with AI code generation. Nor have any of my colleagues.
I have been able to get skeleton applications created in an unfamiliar technology, but they required a lot of changes to even compile, let alone run. They were definitely not suitable as a production-grade application.
I have had better luck using AI to generate classes or methods particularly when I needed to lookup details of a specialized algorithm. AI finds example codes very quickly and much of those codes can be easily adapted to my project needs.
I cannot just cut-and-paste AI code. It takes high-level understanding of the topics to use it, and modifications are always required. Studying the raw code responses is a good training outline. It often tells me what I didn't know that I didn't know.
As the label on the box says; "Some assembly required".
I supplied ChatGPT with very explicit C code, carefully broken out for database operations. Those included printf and scanf calls to illustrate record usage. I defined structures for the records I was using. All it really did is to replace the printf/scanf calls with sprintf to build the queries and calls to the proper mySQL functions.
I will try your query technique. I have base program codes I can use as templates and see what the AI does to them.
Thanks for the information.
If AI gets good at generating C code, I wonder if we’ll even need the other languages.
The prompt IS the code.
The Arduino cut-and-paste programming model has turned out a large number of clueless programmers. They barely understand the language and have no idea at all what can be done with the linker.
I use a lot of tables made of structures, and many members are pointers to functions. My guess is that AI will turn out functional code, but not clever and compact code.
Over time though I suspect we’ll see LLM models specifically trained to generate compact and efficient C code. It won’t be a general purpose LLM, it will strictly be for generating code.
I believe you have spotted a key success requirement to using AI for code generation.
It is much like the earlier days of search engines when using the right query techniques would give amazing results.
That was before "enshitification" of the search engines. The LLM-based AI programs like ChatGPT, CoPilot, and Claude often get better results now than the older search engines do.
But we will still need people to do the software development, and we will still need a variety of programming languages for different environments.
If AI programs do get really good at coding, the humans will get really good at writing the queries to generate the coding. But AI programs that get that good (if they ever do) are likely to be barricaded behind very expensive subscription services which extract every last dollar of value from their results - and a premium on top.
The AI programs will be too expensive to be worth using for most applications. Their outputs will be incomprehensible to human beings for modifications or maintenance and will be supportable only be continued use of the AIs that generated them. It will evolve into a specialized and very lucrative niche industry for the practitioners.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.