AI again - Church speaks out at last!

[malkie]

Don’t use A.I. to write your sacrament meeting talks, apostle urges Latter-day Saints

[during Brigham Young University’s Education Week]
Artificial intelligence is not God and cannot be God. It cannot replace revelation or generate truth from God.
...
A.I. cannot provide inspired divine truth or independent moral guidance. As a creation of God, man can create A.I., but A.I. cannot create God.

He may be right - especially about the current generation of AI. But just as AIs may soon pass the Turing test, someday they may become so superior to us that, for some people at least, they may be a good replacement for a god.

What do you think?

As church members, we will not grow spiritually if we let artificial intelligence write our sacrament [meeting] talks or do our seminary homework. A.I. cannot replace our individual effort and spiritual preparation as we prepare lessons, prayers or blessings.

I'm really taken aback that he didn't mention the permissibility (or not) of using AI to write posts on message boards. Maybe we need to wait for a further statement, after he has a word with Dr Shades.

Gong also demonstrated A.I.’s limits. When ChatGPT was released in November 2022, he said, he and his loved ones asked it to write a journal entry that would summarize a recent family reunion. The chatbot complied but ended up fabricating a story of the family arriving at an expensive lodge with a private chef feeding the clan.

“Gong reunions don’t have fancy lodges or chefs,” he said. “We have a lot of fun and great memories, not in the way ChatGPT fantasized.”

I wonder - could an AI fabricate a story about an emergency landing with a wing on fire? How about a meeting on a plane between an apostle and Mick Jagger?

Perhaps Paul H. Dunn was the first person to run an AI? Inquiring minds want to know!

[[Mods - I assume that this topic is OK here, and does not belong in the AI megathread]]

[Rivendale]

Seems a bit hypocritical. The leaders use professional speech writers in their talks so in a sense it isn't "their" original stream of thoughts expressed. AI probably wouldn't obfuscate or lie about having a picture taken of you and your gay son having dinner together. But who knows, it does depend on the prompt and hallucinations are common in both AI and professional writers .

[Kishkumen]

I like what Gong has to say here. I am anti-AI. AI is a threat to humanity.

[Xenophon]

I'm generally anti-AI as well but honestly it feels like sacrament talks and sunday school lessons would be ideal candidates for AI generation. Unless there has been some significant change to what is being shared across local pulpits the content was generally very formulaic. Talks were mostly milquetoast regurgitations of recycled conference material. It feels like if you fed enough conference talks into a platform you could actually create talks that would be indistinguishable from the 1,000 rehashes of the Beatitudes I've sat through.

This isn't meant as an indictment of members either. Rather an acknowledgement that creating engaging and unique content is not simple, especially given some of the other liberating constraints placed on members and the what/how/why of being asked to speak.

[Sophocles]

I'm generally anti-AI as well but honestly it feels like sacrament talks and sunday school lessons would be ideal candidates for AI generation. Unless there has been some significant change to what is being shared across local pulpits the content was generally very formulaic. Talks were mostly milquetoast regurgitations of recycled conference material. It feels like if you fed enough conference talks into a platform you could actually create talks that would be indistinguishable from the 1,000 rehashes of the Beatitudes I've sat through.

I think that's probably what they are afraid of, people realizing how formulaic it is. Mozart was a genius, no doubt about about it, but once we discovered how easily AI could compose new music matching his style, we began to realize that his genius was more worldly than divinely inspired. Still impressive to some of us, less so to those who wanted to believe that only God could write like that. (They still have Beethoven, though.)

If the average member experiences firsthand how Heartsell can be ginned up via algorithm, they are less likely to be as impressed by Conference talks.

[Dr Moore]

General Authorities have a lot in common with AI.

AI makes mistakes, hallucinates about historical facts, is credited by some with god-like insight, and is terrible at predicting unusual events in the future.

[Physics Guy]

I don't see why machines may not be able to become superintelligent someday. I still like Stanislaw Lem's Golem XIV as a concrete picture of what such a mind might be like.

I don't think we are close to anything like that, however. I think that what gets called AI nowadays would more accurately be called "imitation intelligence".

I don't actually mean to take a stand on whether or not these chatbots are intelligent as opposed to unthinking. That's a subtle issue, but I'm kind of bored with it. I'm not really seeing any new ideas about consciousness here; it's basically Searle's Chinese Room. That's a cool thing to think about, but it's from 1980.

My point is about whether the chatbots are intelligent as opposed to stupid. I think they're stupid, but good at sounding smart. They are imitating intelligence, not demonstrating it. Even after all these billions spent on them, they still just don't actually get things. For the kind of hard tasks for which a real artificial intelligence would be wonderful, they are frustrating.

They may be good at routine tasks, and that may be all kinds of important, but they seem best at sounding smarter than they are. I think we would all be better off—and would get better use of the LLM chatbots—if they weren't trained to sound so smart, for marketing reasons, and instead showed their real capabilities and limitations more clearly.

And I don't see them becoming more intelligent than all humans anytime soon.

[Fence Sitter]

And I don't see them becoming more intelligent than all humans anytime soon.

Given that Trump just got elected the 2nd time, I think AI is already more intelligent than a lot of people.

[Doctor Scratch]

I admit that I laughed out loud as I read the OP. Quite funny that the GAs would find it necessary to admonish the rank-and-file about using AI for this. What it implies is that many in the rank-and-file would find it acceptable to do this. A lot has already been said over the years about how "overworked" the average Latter-day Saint is when it comes to Church service obligations. So it's understandable that they might try to save time and effort by asking ChatGPT to write their sacrament talks. But, of course, arguably the larger issue has to do with "depth" or "content," and hence why I find this humorous: since when is the average Church talk ever "better" than anything that AI spits out? And what percent of the rank-and-file would ever be able to tell the difference between AI talks and "actual" talks? What about if AI were allowed to generate patriarchal blessings? Would anyone be able to tell the difference? Meanwhile, we have the Mopologists' content. The vast majority of the material that appears on "SeN" and Interpreter--probably 95%--could be generated by ChatGPT and I doubt anyone would know the difference.

[malkie]

I admit that I laughed out loud as I read the OP. Quite funny that the GAs would find it necessary to admonish the rank-and-file about using AI for this. What it implies is that many in the rank-and-file would find it acceptable to do this. A lot has already been said over the years about how "overworked" the average Latter-day Saint is when it comes to Church service obligations. So it's understandable that they might try to save time and effort by asking ChatGPT to write their sacrament talks. But, of course, arguably the larger issue has to do with "depth" or "content," and hence why I find this humorous: since when is the average Church talk ever "better" than anything that AI spits out? And what percent of the rank-and-file would ever be able to tell the difference between AI talks and "actual" talks? What about if AI were allowed to generate patriarchal blessings? Would anyone be able to tell the difference? Meanwhile, we have the Mopologists' content. The vast majority of the material that appears on "SeN" and Interpreter--probably 95%--could be generated by ChatGPT and I doubt anyone would know the difference.

If my patriarchal blessing had been written by an AI, it would definitely have been during a period of hallucination.

When the patriarch finished he said something like "where heck did that come from?"