Tech 📱 What is your preferred source for AI assistance?

[michaelskis]

There are several different sites out there that offer AI assistance for things. Chat GTP was an early arrival, but they are not alone anymore. For a while, I have only used Chat GTP, but now I am exploring other options.

Which is your go-to and why?

 

[The Terminator]

I like GPT, Claude and Gemini.

Hate to admit that I use them, since I am aware of the energy and water burdens of data centres, but I do. Try and at least keep it to things that I actually am stuck with, like Python coding, not asking it BS about Jaguars and music.

Anthropic seems to be the "least worst" of all of the major ones, at least on a surface level.

Grok is sometimes funny to use for joke prompts, but again, try and minimize my usage.

 

[JNA]

Have not used any. Don't know what I would use any for.

 

[Suburb Repairman]

I tend towards Claude since it does better with long-form writing and, in my experience, has been far less prone to hallucinating without having to setup a bunch of specific rules to keep it from inventing information. It generally behaves more ethically. I've used ChatGPT a lot as well, but I've got a laundry list of rules I force it to follow. I tend to use ChatGPT to go citation hunting.

I like Ideogram a good bit for image rendering especially in the infographic sorts of spaces with text-on-image, and have been experimenting with Discord MidJourney. I still have a very, VERY strong preference for using human visualization, mostly because I have very particular ways I like to use light and create "motion" to draw specific attention to certain elements. Plus, I enjoy the iterative process working with artists.

 

[Big Owl]

I use ChatGPT and have several projects set up within it, each with its own set of rules. I haven’t tried other platforms, so I can’t really compare. I rely on it heavily for church work in my role as Congregation President. I’ve uploaded a couple years’ worth of meeting minutes, bylaws, and general church information. The key is directing it to pull specifically from those sources. When I do that, it’s very effective at summarizing past actions and explaining how the bylaws apply in particular situations. It saves me time but I still have to verify accuracy.

 

[zman]

I used ChatGPT, but it has been pissing me off as of late... perhaps it isn't good at life advice or maybe I just think it should agree with me more.

 

[Maister]

I don't use AI on ethical grounds. I know I'm fighting a losing battle as the potential utility is immense, but at least I can live with my conscience.

 

[michaelskis]

I don't use AI on ethical grounds. I know I'm fighting a losing battle as the potential utility is immense, but at least I can live with my conscience.

That is how I feel about drive-throughs... and my wife and kids hate it.

 

[bureaucrat#3]

We're only allowed to use ChatGPT at work. I've been using it for very rudimentary reviews and to proof letters occasionally for better language. I don't love it, but it's cut minutes and some other functions from day long jobs to minor editing and review duties.

AI is coming. We should be working on a moonshot type project to figure out how to ethically and safely roll it out with protections and then figure out how to transition people from their former roles to their new roles. I think like any innovation/automation its going to be a substantial change but we're going to find roles for everyone.

 

[DVD]

We can only use the microsoft copilot garbage. It helps when I have some writers block. At home I use chatgpt, but mostly when I want ideas on something I'm doing. So not often.

 

[mendelman]

My palantir.

 

[MD Planner]

I've used several of the products above. There is no doubt I have saved a lot of time using AI. I have no problem typing in a ton of information (so it doesn't have to search for stuff that may or may not be relevant). I love that it can organize, create an outline of info I already inputted in a fraction of the time it would take me. Again, I'm doing the work, providing the info (or the vast majority at least) but as far as organization and presentation I find it almost invaluable already due to the time I save.

 

[Dan]

Gemini excels when it comes to architectural rendering. All I asked for with this request was a rendering of a vegan self storage facility. Gemini made up the sign content -- "CRUELTY FREE AND PLANT BASED STORAGE SOLUTIONS" -- all on its own.

Gemini is also great for cleaning up architectural drawings and illustrations from old newspapers. It usually doesn't add anything extra that I didn't ask for, nor change the perspective.

I've used Midjourney for parody threads here.

For writing, I don't use AI. Not even when I make a progarchives.com user parody post.

 

[WSU MUP Student]

Like @The Terminator I am conscious of the severe environmental impact of AI so I have been reluctant to use it and when I do use it I really try to limit it to professional purposes. I've found it to be very useful for Python coding for stuff in GIS and DAX coding for Power BI. I've also found it to be extremely useful for helping resolve errors I encounter in GIS or Excel or Power BI. Our IT department charges our department if we request assistance from the on projects we're working so if I come across some error or need some formatting assistance and Google and the various forums aren't solving it for me, now I'll check with AI instead of turning to our IT department.

Usually I've used Claude for that stuff but we have CoPilot on our machines at work so I've started using that a bit more often as I figure maybe it's more suited to working with Power BI and/or Excel since they're all Microsoft products?

 

[Planit]

Thanks Dan - I just tried it with this prompt:
show me a drawing of quad-plexes of various architectural styles along a residential street

 

[The Terminator]

Like @The Terminator I am conscious of the severe environmental impact of AI so I have been reluctant to use it and when I do use it I really try to limit it to professional purposes. I've found it to be very useful for Python coding for stuff in GIS and DAX coding for Power BI. I've also found it to be extremely useful for helping resolve errors I encounter in GIS or Excel or Power BI. Our IT department charges our department if we request assistance from the on projects we're working so if I come across some error or need some formatting assistance and Google and the various forums aren't solving it for me, now I'll check with AI instead of turning to our IT department.

Usually I've used Claude for that stuff but we have CoPilot on our machines at work so I've started using that a bit more often as I figure maybe it's more suited to working with Power BI and/or Excel since they're all Microsoft products?

Same, but our IT department here is more an issue of time/capacity than cost. My IT guys are chilled in and let me steer the GIS ship, so I have been using it to help with my Python coding and GIS as well, trying my best to limit my purview to just that.

Although it was cool for generating my PC background, a Jaguar driving a Jaguar

 

[Planit]

Nice T, but he should be driving an E-type!

 

[Faust_Motel]

I have found some utility with Gemini- Mostly just pulling relevant information out of long documents (ex. "here's the 40-page long schedule of all bills being discussed at all committees of the legislature next week, and here's a list of the 25 bills I care about, please create a schedule for me showing where and when those bills will be discussed and what the major topics on the schedule are")

That works pretty well.

I have had trouble getting decent site/architectural renderings out of it. I find that if it makes a wrong turn with a rendering, it's really hard to get it back on track and I get frustrated and have to start over. Example would be I get an image with the building I want but the entryway is the wrong style and no matter what I ask it to do from there on out it just. won't. change. the. entryway.

 

[Dan]

My current workplace (and I love being back at work, I love where I work, and I love my new job) blocks AI stes. In a way, I get it; they pay for my work, and not that of an AI.

(Gemini) I have had trouble getting decent site/architectural renderings out of it. I find that if it makes a wrong turn with a rendering, it's really hard to get it back on track and I get frustrated and have to start over. Example would be I get an image with the building I want but the entryway is the wrong style and no matter what I ask it to do from there on out it just. won't. change. the. entryway.

Precise architectural rendering is a weak spot of most AIs, including image-specific AIs like Midjourney. Still, for my purposes of cleaning up old or rough drawings or renderings without the AI getting "creative" and changing things. Gemini is the best I've found, Here's a few examples from the Unbuilt Buffalo project.

Adelphia Tower (late 1990s): before

After (Gemini; "Clean up and colorize this architectural drawing of the unbuilt Adelphia Tower, proposed for downtown Buffalo in 1997. Don't use dark lines between different colors, shades, or shadow lines.):

ChatGPT, using the same prompt. Not bad!

Compare that with Claude, using the same prompt. It admits failure. "Given it started as a grainy newspaper scan, this is about as clean as pixel-level processing can get without AI upscaling or manual retouching."

Grok, same prompt. I like Grok for text-base actions, but it didn't do that great with this.

Mistral, same prompt. Not too different than Grok.

Midjourney, same prompt. WTF?

---

Peace Bridge proposal (1913). Before:

After (Gemini):

---

Ambassador Hotel (1921): before

After (Gemini):

---

If I want Gemini to get creative, it obliges.

Union Station (1907): before.

After (Gemini).