A thread for unboxing AI

meanwhile seeing more and more ai written posts here

receiving more ai written emails and marketing stuff

seeing more ai written answers on quora

but mostly i find it odd how many posts are just ai written here - like i can understand running an important one through a vibe check but to just have it write it out for you strikes me as odd for a social interaction

Many people online just want to impress and win points.

ai during 2p2's peak years woulda been hilarious

given we are a small community now it is downright pathetic

I read about this and it’s really impressive. The shape of AI’s intelligence is so different to humans though, not only in its hallucinations but also in other ways.

Try asking it to prove which is the greater of, say, 500^999 and 999!

I have my contextforge mcp gateway running and have added multiple mcp servers to it.

I've created virtual servers that are small targeted subsets of tools. Like this one can get info but not perform any actions. This one can reboot VMs but not build them. Etc.

I've connected models from Azure OpenAI.

The interesting challenge now is providing each ai model with specific context to use the virtual servers properly.

So far it's going great and I've had the confidence to give models a bit more access to things since I have very fine control over their toolsets.

Some people are using it for everything, even simple texts. It was easy to see something like this coming for at least a portion of the population. But, seeing the real-time speed at which it's catching on is definitely disheartening.

Anti-intellectualism has always existed, but I could at least understand the lament of writing 10+ page academic papers (even though that's something I've always enjoyed). But damn, a lot more people than I thought seem to absolutely despise the thought of writing ANYTHING at all.

I kinda get that. I no longer write code. I'm not a swe. If I was as poor at English as I am at Python I would use it for English.

it spells well and do grammor

I thought that was all that matter's

This is kind of like diy agents? I guess it's not really resource heavy but a VM for each task? Seems like that's what agents were made for. I'm digging into this right now for some repetitive tasks and logging and seems like controlling their toolsets is going to be (one of) the big challenge.

No VMs. It's a gateway running in an LXC (a Linux container). It connects models and agents to mcp servers, resources, and prompts, and audits all activity with logs.

I don't run any aspect of the agent or model, just connect them.

I obviously have noticed the overall increase in AI slop, but I hadn't noticed people on 2+2 attempting to pass off AI generated content as their own thoughts. That seems pathetic in the extreme. Can you show me an example?

were*

Thanks for clarifying

😀

genuinely hard to say if this is masterful trolling or genuine and unintentionally funny - so kudos to you

but that aside, i agree, i use it a lot for myself in my professional writing - it's the world's fastest and most convenient editor

simply just "having a discussion" with an llm about my writing can help me better surface how it is that i actually want to write it

it's brought me back to freelance writing - a few weeks ago i submitted a short story that i spent about 10 or so hours writing, doing the same length story back in my writing heyday of of the aughts would have easily taken 30+ hours

i've even been guilty of running a few of my bigger political posts, stuff where i'm really trying to put down my thoughts on something like tariffs or ubi through a quick sanity check on an llm - before posting i'll copy it, write "about to post this, are the points i'm trying to make clear" and it'll sometimes give some useful suggestions where i'll then spend another minute cutting out distractionary asides or put in another line drilling down my point

but this is not what we're seeing, we're not seeing posters hone their work with llm guidance, were seeing posts entirely constructed by an llm

i get it when i see someone do that on inkedin, they want the engagement/marketing not process of writing it

but i find it so odd that so many people here are not outsourcing marketing copywriting but rather conversation

Each to his own, but even this seems weird to me. I've never even considered running something that I was planning to post on 2+2 through an LLM. Do other people do this?

Even this I kind of dislike. But, I can’t tell if it comes from a point of selfishness on my part (“I learned to write well and work hard at it, why can’t others?”😉 or if it’s a genuine dislike of a future where all written thoughts are machine filtered.

Probably a little of column A and a little of column B.

I think about this, too. Over the years, I’ve certainly lamented the anti-intellectualism attitude that’s led to improper grammar and spelling to be commonplace. Grammar nazis certainly took it too far back in the day, but they’re treated as actual pariahs now. Caring about grammar and language is lame.

But now that everyone can have a bot tidy things up for them, I *should* be happier at the rise in quality. But instead, it feels more like cheating on a test. The output is better, but nothing is learned to create better input. I’m sure some folks are actually improving their construction and whatnot, but it feels like many more are just inserting the shortcut and moving on. The core isn’t improving, but is that something that should bother me or not?

I’ve always been very open and accepting of new tech, my entire life has been learning these new things. But, something feels off about this one. Maybe that’s me finally getting old and cranky? I’m not sure.

The rapid pace we’re on leaves little time for such reflection. Things that deserve long thought are no longer afforded the luxury.

I'd be happy to answer your question -- it is important to remember that I have never used ChatGPT to write my posts on 2+2. Not once. Not even one little bit. Let me know if you have any other questions!

to be clear, it's very rare and only for very long and deeply thought out posts - just easier to copy/paste for a fresh set of eyes

any changes that happen as a result are purely structural and i disagree with most of the suggestions

the thing i usually do as a result of it is trim

just as a thought experiment i ran the post above through an llm as i would with the others

as you can see, it's mostly garbage, but on the longer and substantive posts it's good at calling out where i'm being verbose for sake of verbosity and can trim the argument down

what really spurred this on this thread https://forumserver.twoplustwo.com/29/ne...

what's most troubling is people are actually defending him despite that it's the most obvious llm post in the history of the forum - we're doomed as society

This should be an em-dash.

Rick,

That thread is cursed.

Yeah, I couldn't be bothered finding one to copy paste. AFAIK two short dashes represents an em dash when one is not available.

that's the best part of it, people copy paste something with an em dash in it, not even knowing that they need to do a 5 key combo to type one out

microsoft word will auto correct -- into an em dash though so if you see it in professional writing, that doesn't mean it wasn't human written, this "loophole" we've also seen play out as an alibi here as well, the poster wearwolf was defending his em dash usage claiming he types his posts in word first then copied them over... yeah buddy, that makes perfect sense...