AI agents (not just AI: agents specifically) - who is using them and how?

[William the Third]

"This hits the nail on the head"

-_- lol

It's not actually entirely true that it's purely a prediction model. It *was*, yes, but the system is driven to figure things out internally using whatever means necessary to get the prediction right.

There is starting to be evidence that in order to do so at the levels we're now seeing, it's having to actually build internal models of how the world works.

That's what our brains do. It's called "understanding".

Not saying it's conscious or ever can be.

(not that we could ever know - consciousness defies easy explanation).

So called "reasoning"/"thinking" models are just implementing Chain-of-Thought Prompting, which basically breaks down an input into smaller chunks and processes them individually. This produces far better results, especially for some task, but at the end of the day it's still generating tokens statically based on context and training data. The verbose output may make it appear like it's thinking/reasoning/understanding, but really it's still just statistical generation. AI will be real, but this isn't it.

[vijil]

Look up "emergent world representations" in AI research.

AI is "just" statistical generation in the same sense that your brain is "just" neurons firing. It's technically true, but also an oversimplification.

[TencenT]

Pretty heavy AI user as I'm software engineer and the reality is that long term no one is going to be interested in paying me to solve problems at the speed I can solve them. Is it fun? No not really however the upside is that you can now work on a lot of stuff at once and if you read between the lines on that you can find ways to get back to enjoyable problem solving and not just be a button smasher until you retire.

Semi related, now that I do not have to write Rust/c/c++/etc anymore, they're amazing languages that I've long despised reading or thinking about.

[Tarsun2]

i took a programming course in college. I didnt do too well though, i got a C++

[ConradCustoms]

As a Robotics Engineer working in custom Automation solutions, I find myself using AI more and more. I came in pretty skeptical but recently I used it to basically review my "human slop" code, clean it up, gave it parameters for how I wanted it to look, standardized it with a program format. It was able to review what I was doing, suggested some good and some bad ideas. I then had an even BETTER idea based on its suggestions and let it keep track of how that will all work. Long story short: I used maybe... 600 tokens and completely revamped and bullet-proofed my code and more impressively it did probably 2 weeks worth of work and validation in a couple days.

Now I gotta figure out how to still stretch my jobs out a little and get my mileage and not let the AI bots take all my fun money from me

[SignOfZeta]

Some human mod cut the Jesus reference out of my post. Now I’m starting to have second thoughts about AI…most AI have better ethnics than that at this point…not even Grok bans Jesus.

[Carp]

https://mcarterbrown.com/forum/main-...29-forum-rules

4) No Trolling
On this website, “trolling” is defined as the posting of inflammatory, extraneous, or off-topic messages with the primary intent of provoking other users into a desired emotional response or of otherwise disrupting normal on-topic discussion. Posts deemed to be trolling by the staff will be deleted and infractions will be issued as appropriate. Repeat offenders of this rule will be banned from the forum.

5)No politics, race or religion
We all have different points of view. But we come together to talk about and share our experiences with and about Paintball! Threads or posts that are deemed to be unnecessarily controversial or offensive are subject to deletion. Infractions will be issued to posters as necessary.

[SignOfZeta]

Hello human posters. I have come to address the rumors that moderators are banning users with automated Artificial Intelligence agents--also commonly known as "bots." Some arguments favor the use of automated agents to save time by taking over repetitive, simple tasks--thus freeing human moderators to devote more energy to complex administration and security. Others believe that Artificial Intelligence has gone too far, depersonalizing online interactions and filling social communities with low-effort content commonly called "AI slop." Whatever your perspective, here at MCarterBrown we enjoy the refinement and satisfaction of obsolete technology, and would never force automated moderation and tools of dubious value down our users' throats or through subcutaneous injections administered secretly while you're asleep, Carter probably promised, maybe, somewhere out there in the multiverse. In short, rumors are harmful, and users are better served when posted commentary is thoroughly researched beforehand. Whichever side of the debate you support, we should all engage and build the communities where we want to thrive!

If I wasn’t banned by a bot then what is the explanation? And why were others banned for no reason and reinstated as well? What happened to the mod named Fluff who banned me and disappeared?

I apologize if my guess was wrong but when crazy stuff like that goes on with no explanation I have nothing to go on by my own crackpot theories.

[Axel]

It's all good. Lady Satire, my Muse, she commands and I must obey.

As for your questions, one of the old men who works here fat-fingered a moderation tool, and then your posts had to be manually restored one by one. I don't know what he was originally trying to do, but the ban was accidental. Carp recently finished restoring it all so you should now be good to go.

[SignOfZeta]

“Fat fingered” usually means an input error like a typo. What “tool” are we talking about that has a one click permaban button with no additional dialog? And is there a similar button that can be implemented to explain to the banned person what happened in a forthcoming way automatically so that the shy mod won’t have to explain himself or are we just expecting people who are randomly banned for no reason to not really care about it?

What I’m saying is put a barrel plug in that thing if the trigger so incredibly light that it cannot be controlled, please.

[SignOfZeta]

Look up "emergent world representations" in AI research.

AI is "just" statistical generation in the same sense that your brain is "just" neurons firing. It's technically true, but also an oversimplification.

Human thought is not composed entirely of neurons firing. To say so wouldn’t be oversimplification it would just be incorrect.

An enormous degree of human decision making is controlled by other biological mechanisms such as hormones, genetic makeup, the local weather, bacteria in the gut or even parasites. You can’t make a human mind out of just a human brain. It only exists as part of the entire animal.

An AI exists entirely in digital code (and a giant lookup table of actual human work). Nothing in a human brain exists in digital code.

[vijil]

And those biological mechanisms cause neurons to fire differently. In the end it's all "just" neurons firing.

Correct, but a massive oversimplification.

Like "it's just digital code".

Semantics. Doesn't mean anything either way.

My take...

AI is basically just home made disguised alien lizard invaders. We couldn't find any real ones so we built our own. They're completely different to us, but they're brilliant at faking humanity.

[Carp]

Nothing in a human brain exists in digital code.

Both the thoughts of, and the 1's and 0's are "something" and exists in your brain. Don't make me dust off my Philosophy 101 book. 🤔

[Walker]

What we are seeing is not Artificial Intelligence, no matter what they are calling it. This is just a form of advanced machine learning (call it Electronic Intelligence if you will). We may get there sometime, but we aren't there yet.

At this point in time and development, if we put what they are calling AI in control of critical functions, we are stupid. It would be worse than putting a baby in charge.

A + B = C The algorithms can get this right most of the time. But, can they properly determine the effect that C will have on the world when the human factor is involved? Or for that matter, A and/or B may not be the same, as they were derived with different methods? Can it handle illogical outcomes? Does it understand learning by failure? Does it experience pain or discomfort when it gets it wrong and hurts others? Does it experience satisfaction when it's outcome effects the surrounding world in a way that is positive? Do it even know what a positive outcome is? Does it understand the purpose of positive and negative reinforcement? Can it interpret past actions and history through the "lens" of human involvement and the circumstances that brought about those results?

We are not there yet............. There are too many questions remaining to be answered before we trust the machine......

Use it? Yes, it can be an effective tool. Trust it? No......

[JonM]

AI is good at some tasks and terrible at others. If I need to spit out some quick coding at work, AI is great for that. If I need information on more obscure topics (like paintball for example), then I often find myself correcting the AI and getting a "you're completely right" response than getting anything particularly useful from it.

[Chappy]

“Your frustration is understandable “

[TencenT]

It's pretty interesting how genuinely bad most of the chat bot interfaces are with it compared to how genuinely good some of the more open ones are like Claude (Code). With the right harnesses in place they can go really really far. I want to make some injection molded parts and found out pretty quickly how cost prohibitive that can be (like $12k for just one of the pieces I wanted with a minimum 500 part order). Using claude I was able to cycle through several generations of molds based on the parts I designed then use openfoam to refine them based on how the fluid dynamics of the plastic injection would actually work.

Now to be fair these are all things I know are possible and I have enough cursory understanding to hold its hand and make directed decisions, but still amazing it was able to cut what would have been literal months for me to figure out into something I finished in a weekend. It genuinely feels more that if you know enough about a topic and know how to kick it in the right direction you can really accelerate your work and the error rate from it is actually pretty low. The problem then becomes finding a good operator which is the problem you face with any sufficiently complicated tool.

[SignOfZeta]

Both the thoughts of, and the 1's and 0's are "something" and exists in your brain. Don't make me dust off my Philosophy 101 book. 🤔

No. This isn’t true at all. Thought is not digital. Only computers are digital. Humans actions are dictated by their entire physiognomy. Nothing about it is digital.

And you have no experience with philosophy whatever or you wouldn’t have written such a silly thing.

[Carp]

Have you been drinking the Kool-Aid again? That's a retorical question! The willfully ignorant bliss must be nice and addictive. Try to have a better day.

[Cyberpyr8]

We're building an agent for our helpdesk. We already have a ton of powershell commands that we are connecting back to and allowing our help desk users to use it in Teams. So I have a Teams user that we can ask stuff to and it will process through the scripts. They get info back. We have it tied to ServiceNow so you can even just give it a service ticket and it can do a lot automatically. It can escalate a ticket, update a ticket with notes and instructions directly.