How do you use AI, and what do you recommend be avoided?

[Voboy]

Probably no one uses it for things their brain can do just as well.

You are quite wrong. Several of my colleagues use it for things they used to just do themselves. It's very pervasive.

 

[AG31]

I'd do this, involve an engineer who knows what they're doing. Right from the start.

And where do I find such an engineer? I'm just hoping one of the condo owners might have such skills.

Didn't you just say up above that the AI produced the wrong manual?

I'm not sure what your point is. In my first contact with AI, I didn't send it the manual I had, I just gave it the model number.

The AI is only as good as the raw material it's given. If the installation manual is useless, the AI composite of all similar installation manuals is going to contain an element of that wrongness. You've said that yourself, it gave you the wrong manual.

And your point is? Sounds like your making a case FOR AI. Just be sure to give it a good prompt?

 

[ElectricBlue]

And where do I find such an engineer? I'm just hoping one of the condo owners might have such skills.

I'm not sure what your point is. In my first contact with AI, I didn't send it the manual I had, I just gave it the model number.

And your point is? Sounds like your making a case FOR AI. Just be sure to give it a good prompt?

My point is always, you can't trust the content it gives you. If one element turns out to be demonstrably wrong, you can't trust anything else to be correct.

You gave it the correct model number, it produced the wrong manual. That's not a useful search tool. Apply the same principle to every question you give it, is all I'm suggesting you do, but don't trust any of it.

I picked up a tip from @Bramblethorn the other day - put -ai at the end of your search query, and your search return will list the sources and omit the AI distillation. Now that's useful information.

 

[EmmAgain]

I used Topaz’s AI image enhancer on a whole bunch of old, low-res personal pictures as a batch job and throughout regretted it. It wrecked the original images and left my thousands of often deeply personal photos with irreversible Insta filter effects.

Use sparingly and selectively! Not every personal souvenir needs a patina of Tropical Beach Daiquiri Glam or droopy bunny ears.

 

[XerXesXu]

What's a diffusion model? I know, I know. I could Google it and let AI tell me, but I'd rather continue in a human conversation.

It's explained here:

 

[Bramblethorn]

You are quite wrong. Several of my colleagues use it for things they used to just do themselves. It's very pervasive.

Yep, thinking is work, sometimes fatiguing work.

One thing I always find unreasonably challenging is when I need to write a farewell message in a card for a co-worker, especially if I don't know them well. Everything I can think of feels so trite and generic, especially when a dozen people have already written in the card and covered the bases on stuff like "wishing you well for your next job!" and "keep in touch!" I can't help overthinking that kind of stuff.

We usually do that kind of thing via an e-card since it's hard to get a physical card to everybody who needs to sign. Last time I did it, they'd added a new option: you can just click a button and have an AI generate a farewell message for you.

I could've saved five minutes out of my day clicking that button, and nobody would ever have known. But the idea just felt repulsive. No, I'm not good at interpersonal stuff with people I don't know, but even if I can't write an original or memorable message I can at least affirm that they were worth five minutes of my time, and get in the practice for saying things that are important for other humans to hear.

 

[Aviastroitel]

A widely misunderstood group of people, the Luddites.

I used to moderate a Luddite web forum.

 

[ChuckWolf]

I've used AI as a search engine and for recreational use for image, voice, movie and music generation.

I wish it were more accepted, but I thoroughly get why it isn't, especially on a site like this.

 

[AngieGZC]

As a tool and not as something to do work for me...

 

[ronde]

The search engine I use has an AI component that will give me relevant details about my search terms, but I use it as a starting point for further research, not as an absolute truth. It's pretty good for questions for which there is published data, like the demographics of a particular area, but if has been known to be wrong when compared to further searches. It's a lot like Wikipedia in that respect. The validity of the answer depends upon which sites it searched and the author of the data on those sites.

 

[AG31]

I've had a mini-course in how AI can get things slightly wrong over and over (different errors), but how it can be used to finally get some useful info. As I described above, I've been wrestling with an uber complicated, super non-intuitive thermostat. I progressed from asking for manuals to asking specific questions. How do I change the set temperature? How do I turn off the fan? What does "auto" mean if the thermostat turns the fan off without "auto" showing? Blah, blah. The pattern was that I'd ask the question, and the answer would not reflect my thermostat experience, and I'd tell the bot so. It would refine its answer. I did get good, useful info. One bit was that one of my two thermostats had been installed as if its temperature were under the control of the building manager. It matched my experience well enough for me to get a visit on Monday from someone who promises to show me how it worked.

Bottom line. With a lot of "arguing" it did give me useful information.

 

[AG31]

Well, finally another resident of the building, who has a working thermostat and the correct manual, came and spent half an hour to get mine in a workable state. He finally had to do a factory reset (which involved pressing 3 icons simultaneously over and over till he got "heard")

I learned through that that almost everything ChatGPT had told me was untrue. It would just make stuff up, based on my confused questions.