AI is not the problem. We are

[The Liturgist]

I had no interest in making that hash. The reason I asked AI about it was because I had watched a video of a renowned South Carolina barbecue chef who said that the meat ingredient they use is either the butt or the head of the cow. He didn't mention pork. He used the head - face meat (yuk). I was curious to see if AI would give that as an optional ingredient. Just now I asked chatGPT the same question about how to make it, and it said it's made with pork. Didn't mention beef as being an option. I thought I had caught AI making a mistake, but I perused a few websites, and yes, pork is most common. But it's pork offal. If I'm ever in South Carolina and someone offers me hash, I'm afraid I'll have to say no thanks.

Maybe if I used the paid version of chatGPT it would have given me more ingredient choices? I don't know.

No, rather, you need to ask it about the beef version. By default, its going to tell you the most common ingredient, so if you want something non-standard, you need to express that to it.

Also by the way the extent to which chatGPT is able to satisfy the demands of an end user depends on whether or not you stay logged in, because its in the course of extended conversations where good things tend to happen with that particular AI. Unlike with Grok, which is basically the same on each execution, but which is not as bright or capable as chatGPT.

 

[The Liturgist]

"You got me - I don't know so I took a guess. Here is my best guess..." They always present the answer with some level of assurance.

ChatGPT 5 is specifically designed not to do that. In addition, there were models of chatGPT 4 you could have selected which were more performant in this regard, but less…creative.

Part of the tradeoffs with these AI systems has been between creativity and reliability.

What will improve the experience of anyone using an AI is to maintain an ongoing conversation with it, which allows it to adapt to your particular standards and requirements. For example, you can direct chatGPT to verify all of its answers by checking with a search engine (early examples had no ability to use a search engine or do research; this was introduced with the 4o model and has become progressively better).

With chatGPT 5, there are now two models, the default model and the “thinking” model, which is really a reasoning model, based on the o4 series, whereas default chatgPT 5 is a refinement of the prototype 4.5 but with more flexibility. The thinking model, like Grok, allows you to see the steps it is taking in answering your question, which is diagnostically useful, but it also takes longer to answer queries. The stock version is faster.

 

[Chesterton]

No, rather, you need to ask it about the beef version. By default, its going to tell you the most common ingredient, so if you want something non-standard, you need to express that to it.

Also by the way the extent to which chatGPT is able to satisfy the demands of an end user depends on whether or not you stay logged in, because its in the course of extended conversations where good things tend to happen with that particular AI. Unlike with Grok, which is basically the same on each execution, but which is not as bright or capable as chatGPT.

I don't use chatGPT or Grok because I'm not a fan of signing in to too many things. I avoiding handing over my email address as much as possible. Do you know of any good quality AI site which doesn't require signing in?

 

[JustaPewFiller]

ChatGPT 5 is specifically designed not to do that. In addition, there were models of chatGPT 4 you could have selected which were more performant in this regard, but less…creative.

Part of the tradeoffs with these AI systems has been between creativity and reliability.

What will improve the experience of anyone using an AI is to maintain an ongoing conversation with it, which allows it to adapt to your particular standards and requirements. For example, you can direct chatGPT to verify all of its answers by checking with a search engine (early examples had no ability to use a search engine or do research; this was introduced with the 4o model and has become progressively better).

With chatGPT 5, there are now two models, the default model and the “thinking” model, which is really a reasoning model, based on the o4 series, whereas default chatgPT 5 is a refinement of the prototype 4.5 but with more flexibility. The thinking model, like Grok, allows you to see the steps it is taking in answering your question, which is diagnostically useful, but it also takes longer to answer queries. The stock version is faster.

But, you may need to change conversations sometimes due to differences in how we and the AI "think" (for like of a better way to put it).

Here is an example. I didn't realize I needed to start a new conversation until I saw the results...

I was playing around on Gemini making a few images the other day.

I made a few images of a pile of autumn leaves with a leaf rake and work gloves on the ground beside the leaf pile.

After making a few iterations of that image I changed gears. I asked for an image of 2 cats curled up asleep on the sofa. I did not mention anything about leaves, rakes or gloves in the prompt.

Gemini gave me an image of 2 cats asleep on the sofa with a leaf rake and a pair of work gloves on the sofa beside them. Clearly the either the cats were exhausted from raking leaves or someone had done a fine job of raking the cats up into a pile on the sofa.

It hadn't dawned on me to start a new conversation. I was still just making images after all. But, what I really wanted was a totally different image than I was making before. I started a new conversation and bingo, just cats, no rakes and no gloves.

 

[The Liturgist]

Gronk

Grok, surely? X.com’s AI is named based on a word popular among programmers going back to the UNIX and Arpanet hacker culture of the 1970s (hacker as in programmer, not criminal hacker - the word hacker later took on a perjorative context within the mainstream media, but even today among the major operating system development groups for example, people who work on the kernel are called “kernel hackers” and so on.

AI is capable of producing "AI Slop" in anything it does from art to code. Much of if generates slop or not depends on the quality of the model, its training, and the instructions it is given. But what it produces is often "good enough", in addition to being fast and cheap.

AI Slop is a known effect, but it can also do good work, depending on the model. For cxample, chatGPT’s image generation system has been massively upgraded and is now capable of accurately drawing human anatomy. The mistakes that used to be made when drawing details like the hands have been rectified.

I don't use chatGPT or Grok because I'm not a fan of signing in to too many things. I avoiding handing over my email address as much as possible. Do you know of any good quality AI site which doesn't require signing in?

In the case of AI, all good AIs need a login, because the way to use them for optimal performance is to have one singular, session-persistent conversation with them, which requires having a user account with the platform.

What I do is use a separate e-mail address for log-ins to online services I don’t want spamming my primary account.

 

[JustaPewFiller]

Grok, surely? X.com’s AI is named based on a word popular among programmers going back to the UNIX and Arpanet hacker culture of the 1970s (hacker as in programmer, not criminal hacker - the word hacker later took on a perjorative context within the mainstream media, but even today among the major operating system development groups for example, people who work on the kernel are called “kernel hackers” and so on.



AI Slop is a known effect, but it can also do good work, depending on the model. For cxample, chatGPT’s image generation system has been massively upgraded and is now capable of accurately drawing human anatomy. The mistakes that used to be made when drawing details like the hands have been rectified.



In the case of AI, all good AIs need a login, because the way to use them for optimal performance is to have one singular, session-persistent conversation with them, which requires having a user account with the platform.

What I do is use a separate e-mail address for log-ins to online services I don’t want spamming my primary account.

Actually mostly chatgpt for Linux thus far. Its worked, and so, I've stuck with it.. It is honestly a good "all around" model at least for my uses.

Yeah - people talk about AI having trouble with hands. But that is understandable. Hands are hard. Go back and look at some of the "practice works" of the old masters. They painted, drew or sculpted hands after hands until they got them right. The new kid on the block having trouble with them is nothing to be ashamed of.

 

[The Liturgist]

Yeah - people talk about AI having trouble with hands. But that is understandable. Hands are hard. Go back and look at some of the "practice works" of the old masters.

This problem has been corrected with chatGPT, to be clear, which is extremely impressive. They basically had to give it a model of human anatomy so it could model us realistically.

 

[RDKirk]

Grok, surely? X.com’s AI is named based on a word popular among programmers going back to the UNIX and Arpanet hacker culture of the 1970s (hacker as in programmer, not criminal hacker - the word hacker later took on a perjorative context within the mainstream media, but even today among the major operating system development groups for example, people who work on the kernel are called “kernel hackers” and so on.



AI Slop is a known effect, but it can also do good work, depending on the model. For cxample, chatGPT’s image generation system has been massively upgraded and is now capable of accurately drawing human anatomy. The mistakes that used to be made when drawing details like the hands have been rectified.



In the case of AI, all good AIs need a login, because the way to use them for optimal performance is to have one singular, session-persistent conversation with them, which requires having a user account with the platform.

What I do is use a separate e-mail address for log-ins to online services I don’t want spamming my primary account.

You can also name conversations and tell it to remember them.

 

[Chesterton]

In the case of AI, all good AIs need a login, because the way to use them for optimal performance is to have one singular, session-persistent conversation with them, which requires having a user account with the platform.

What I do is use a separate e-mail address for log-ins to online services I don’t want spamming my primary account.

Is it always important to have a persistent conversation? I can imagine how that could be important if I were using AI to work on an ongoing project, but does that matter if I'm just a casual user who, for example, might want to ask about quantum mechanics today and ask about making tuna salad tomorrow?

 

[The Liturgist]

Is it always important to have a persistent conversation? I can imagine how that could be important if I were using AI to work on an ongoing project, but does that matter if I'm just a casual user who, for example, might want to ask about quantum mechanics today and ask about making tuna salad tomorrow?

With chatGPT, yes, because the model will adapt to your interests and idiolect, and also it shares data between conversations, and much of what you might care about, for example, your preferred approach to output, how rigorous you want it to be, and so on, can be defined.

With Grok on the other hand there is much less of a benefit, but Grok is not as competent an AI platform; the main thing it is good for is generating pictures of historical figures which chatGPT won’t do, or at least wouldn’t before the recent upgrades.

 

[The Liturgist]

You can also name conversations and tell it to remember them.

Yes this is correct, and you can do a lot more than that with them.

 

[The Liturgist]

But, you may need to change conversations sometimes due to differences in how we and the AI "think" (for like of a better way to put it).

Here is an example. I didn't realize I needed to start a new conversation until I saw the results...

I was playing around on Gemini making a few images the other day.

I made a few images of a pile of autumn leaves with a leaf rake and work gloves on the ground beside the leaf pile.

After making a few iterations of that image I changed gears. I asked for an image of 2 cats curled up asleep on the sofa. I did not mention anything about leaves, rakes or gloves in the prompt.

Gemini gave me an image of 2 cats asleep on the sofa with a leaf rake and a pair of work gloves on the sofa beside them. Clearly the either the cats were exhausted from raking leaves or someone had done a fine job of raking the cats up into a pile on the sofa.

It hadn't dawned on me to start a new conversation. I was still just making images after all. But, what I really wanted was a totally different image than I was making before. I started a new conversation and bingo, just cats, no rakes and no gloves.

ChatGPT 5 is specifically trained to recognize a change in conversational topic, and also I’ve never had that issue, but of course what one should do whenever changing the subject or output preferences is make it clear that a new subject is being generated.

That said, the use of conversations themed to particular subjects or styles of artistic output is highly desirable.

Things have improved so much for the average user with ChatGPT version 5 and the other new systems such as the new image generator. Grok also recently made version 4 available for some users but I’m not into it.

 

[timewerx]

former Open AI researcher Daniel Kokotajlo warned that Artificial Intelligence will become an existential threat to humanity within two years.

Not for me at least.

AI saved my life or God used it as an instrument of our rescue.

Luke 19:40
“I tell you,” he replied, “if they keep quiet, the stones will cry out.”

 

[The Liturgist]

Not for me at least.

AI saved my life or God used it as an instrument of our rescue.

Luke 19:40
“I tell you,” he replied, “if they keep quiet, the stones will cry out.”

God can absolutely rescue people through something of AI; that I have no doubt. Additionally, the major models (chatGPT, Google DeepMind, Microsoft CoPilot et cetera) are not going to intentionally help the end user do something which could be harmful; they have safeguards to prevent this.

 

[RDKirk]

God can absolutely rescue people through something of AI; that I have no doubt. Additionally, the major models (chatGPT, Google DeepMind, Microsoft CoPilot et cetera) are not going to intentionally help the end user do something which could be harmful; they have safeguards to prevent this.

Providing wrong information can certainly be harmful, and they do provide wrong information...frequently.

 

[The Liturgist]

Providing wrong information can certainly be harmful, and they do provide wrong information...frequently.

Which is why ChatGPT and other responsible providers include a warning label.

By the way, one can protect against inaccurate information to a large extent through careful use of the prompt. People don’t understand the real capability of these systems is that they represent the ultimate macro processor or text-transformation utility, as if one can speak the word and animate all of the text processing utilities of the UNIX system like m4, grep, less, sed, awk, et cetera.

Also the errors tend to relate primarily to questions posed in natural language not checked against external data, for example, with a web search, and never relate to the output of questions processed by the built-in programming language.

So whereas on the one hand I have made the point that if people refer to AI as a reliable source for information in their posts they are making an appeal to unqualified authority, but on the other hand, AI can be a source for information that is as potentially reliable as any other web search, with additional processing capabilities.

 

[Jerry N.]

This is a very interesting thread. There are two things about AI that I’ve been wondering about:

• AI is running out of good data. Except for biological data that can be used for creating new and better medications, and some forensic studies, and similar studies in electronics and such, AI can’t exceed human knowledge. Researchers warn we could run out of data to train AI by 2026. What then?.
• I wish there was some type of marker on images and videos so I know it was generated by AI. I often view harmless videos that I find interesting. I show them to my wife, and she can often spot that they were AI generated. It is deflating, and I wish there was a required water mark or something.

 

[JustaPewFiller]

This is a very interesting thread. There are two things about AI that I’ve been wondering about:

• I wish there was some type of marker on images and videos so I know it was generated by AI. I often view harmless videos that I find interesting. I show them to my wife, and she can often spot that they were AI generated. It is deflating, and I wish there was a required water mark or something.

Some social media platforms do have some type of "Made by AI" that flag that can be set when posting an image or video. However, it often seems its up to the person posting if they tag it that way or not.