Sermon on the Dangers of AI, Screen Addition

[mike]

Here's a sermon that someone recommended to me the other day that I thought was really good. The speaker talks about the power and the challenges of AI and goes on to talk about the dangers of screen addiction in general among other tech-related issues. Highly recommended.

https://www.youtube.com/live/dq7lb6GCGU ... 0vQxPVVUjn

[mike]

There's about a ten-minute children's lesson before the sermon.

In the sermon, he demonstrates how AI can be used to generate a five-minute devotional talk that is very good and with only a few minor changes could be shared by the speaker without anyone knowing it was not the product of his own study and inspiration. He reads the entirety of the AI-generated speech to demonstrate this.

I believe his primary concern is AI taking the place of a person's own work, saying that it will result in the flattening of one's own intelligence. It also can help a person fake inspiration and knowledge. I'd enjoy seeing another discussion of AI here. I utilize AI for various things including Bible study. In my experience AI can quickly put together some ideas as a springboard for more study, to remind me of things I already know, and give me ideas on how to word concepts. It is also very good for producing study questions based on a text, and creating outlines.

I feel like AI can enhance intelligence, not that there are not challenges, dangers, and issues to consider.

Socrates was supposedly very worried that the increasing use of books in education would have the effect of ruining students' ability to memorize things. I feel like the worry about AI is somewhat similar to this concern, and there is no doubt that a highly powerful new technology can in fact create laziness and bring atrophy to certain skills. I can barely stand handwriting a long bit of writing because I'm so used to typing with a keyboard. My handwriting skills have atrophied.

Still, I think I would argue that AI is generally a net benefit.

[Neto]

I don't use AI for anything, and generally avoid looking at the AI response that Google Search now puts at the top of search results, but I think that as long as a person does their own study - checking the reliability of the AI response, it's probably OK to use its response for Bible study in the way you describe. The caution I would offer is that just like depending on books by popular Evangelical preachers/teachers can lead a person into a sometimes very "non-anabaptist" direction, remember that AI is (apparently) just pulling information from perhaps prominent on-line sources. So it is easy for it to be "mislead" in its responses.

I think I gave some examples here in the past, where I tested some sort of AI tool (I do not remember which) by asking it about myself, and also about a specific "Russian Mennonite" group. In both cases it gave very incorrect responses on the first tries, and I had to give more information about myself to get it off of other people by the same name. (My wife & I have produced a few linguistics articles that are referenced by other researchers, so our names are not totally unknown in the "on-line world". But still, I didn't expect much, thinking that it would probably pull up information on a South African man by my same name, who went to prison for something or another, but it never mentioned him at all, instead going to three other people before I gave it enough information to actually zero in on me. Also, some of the information was still identical to earlier responses, about other people.)

On the question I asked it about the Kleine Gemeinde group, it gave a beginning date that was around 50 years off, probably confusing it with the Mennonite Brethren. When I offered a correction, it immediately accepted my information, incorporating it into it's response. (At first I just stated that its date was wrong, and it accepted that, then gave an even more incorrect date.) So my "take" on this is that it does not say that it doesn't know the answer, but makes up something instead. It also seemed to me that it too quickly accepts suggestions that you offer it, so I think its responses could be too easily purposefully manipulated. (I didn't try to mislead it, so maybe it WAS actually checking out what I fed back to it.)

[mike]

The less information there is online on a topic, the less accurate AI responses will be on that topic, I would think.

[Ken]

The less information there is online on a topic, the less accurate AI responses will be on that topic, I would think.

Exactly:

AI is basically just having a computer do your googling for you. If there is nothing out there to google, then there is nothing for AI to find.

The problem seems to be that AI algorithms aren't good at dealing with a lack of information. And instead of just simply saying "I don't know" they put together whatever bad information there is and produce responses that are inaccurate or made up.

[ohio jones]

If I listen to this sermon, will I be able to resist the temptation to add a third monitor?

[barnhart]

AI is basically just having a computer do your googling for you. If there is nothing out there to google, then there is nothing for AI to find.

The problem seems to be that AI algorithms aren't good at dealing with a lack of information. And instead of just simply saying "I don't know" they put together whatever bad information there is and produce responses that are inaccurate or made up.

I am beginning to think AI is an unusual case of a new technology that is accurately named. The more intelligent we want it to be (do our thinking for us) the more artificial the output (it makes stuff up). The less intelligent we want it (computation, accumulate information, analyze scans for cancer) the closer it stays to reality.

[mike]

If I listen to this sermon, will I be able to resist the temptation to add a third monitor?

Only if you refrain from actually watching the screen while the YouTube video plays.

[Ken]

AI is basically just having a computer do your googling for you. If there is nothing out there to google, then there is nothing for AI to find.

The problem seems to be that AI algorithms aren't good at dealing with a lack of information. And instead of just simply saying "I don't know" they put together whatever bad information there is and produce responses that are inaccurate or made up.

I am beginning to think AI is an unusual case of a new technology that is accurately named. The more intelligent we want it to be (do our thinking for us) the more artificial the output (it makes stuff up). The less intelligent we want it (computation, accumulate information, analyze scans for cancer) the closer it stays to reality.

Plus, context matters and most questions are actually ambiguous or not as specific as we think they are.

For example. I might ask how many miles are between Pittsburgh and Philadelphia. And there might actually be 10 or more correct answers to that question depending on the context. Are we asking about highway miles or air miles? Freeway miles that might generate a longer but faster route? Or the shortest possible road distance via city streets and country roads? Are we asking from city center to city center, or from city limits to city limits? If we are asking air miles, are we asking from airport to airport or as the crow flies from city center to city center? Or are we asking actual straight line miles which would be boring through the earth's crust in a straight line rather than following the curvature of the earth. AI might guess from the context of the question which answer we are seeking. But it also might guess wrong. It might also seek clarification. Or it might not.

And that is for a very simple seemingly factual question about the distance between two cities. If you start using AI to solve complex tasks such as developing new mRNA vaccines or some such you might have hundreds of such assumptions built into the question and not even know that they are there or how the AI is guessing.

[joshuabgood]

I use AI for editing and proofreading web pages, other writing etc.