5e - why you think it sucks, and why you're wrong

[Beoric]

A couple more years in the oven, and we might just have something that can pass the Turing test.

It already passes the Turing test. The Turing test has proved insufficient.

 

[DangerousPuhson]

It already passes the Turing test. The Turing test has proved insufficient.

Then it'll pass the "whatever comes after a Turning test" test. You get my point though, yeah?

 

[Beoric]

Then it'll pass the "whatever comes after a Turning test" test. You get my point though, yeah?

I do understand what you are saying.

 

[DangerousPuhson]

Turning test

Ducking autocorrect, garbage piece of short!

As someone whose job is not at risk of being out-automated by AI, and also as a very lazy man, I am excited to see what they'll be able to do. My only real qualm about AI in general is when people try to automate some kind of profit venture via AI (like, writing their adventures for them, or making an art book to sell or whatever). That shit is inexcusable.

 

[robertsconley]

Dumb question possibly: So was ChatGPT generating that dungeon room by room, or working from an already complete dungeon? Can it do the latter?

If the dungeon would fit into its memory around 3,000 words or so.

 

[robertsconley]

I think if it knows the style (as you identify) and can create spontaneous, iterative results (as evidenced by the generated dialogue which cannot be found verbatim elsewhere on the internet), then it's already sitting at a 90% solution. It may not know what it's doing, but it sure knows how to do it. A couple more years in the oven, and we might just have something that can pass the Turing test.

I feel what Generative AI and LLMs illustrate how much of what we do revolves around the recognition and manipulation of patterns, visually, textually, etc. Along with what happens you have a dense set of pattern data to work with.

But there is more to human intelligence and soon the newness of it will wear off in favor of exploring what it can actually do and what that can be applied too.

 

[DangerousPuhson]

I feel what Generative AI and LLMs illustrate how much of what we do revolves around the recognition and manipulation of patterns, visually, textually, etc.

Yes, much of everything has a formula that can be followed, or is rife with enough commonalities that generation is essentially calculated mimicry. But I always divert to the old adage that sufficiently advanced technology is indistinguishable from magic, and in time, I believe AI will surpass what we expect of it. I think as AI develops, it will be able to better pick up on these formulas/commonalities, so much so that it will likely even grasp the ones we don't notice (which is what I argue distinguishes AI materials versus human-made ones; the tough-to-identify intangibles that denote intuitive themes, and evoke real emotions).

I paint a little bit, acrylic on canvas. When I look at a professional painting from afar, I wonder "how the heck does someone get so good at art that they can just paint that?". It's like magic to me. Only when I get in closer do I realize it's just calculated brush strokes and effective use of shapes. That face that looked so detailed to me from afar actually looks like a face-shaped blob with two little blob eyes and blob mouth on it when seen up close. My eyes were tricked into seeing this amazingly-detailed face out of mere face-shapes. AI I think is the same thing - magic from afar, pretty basic up close. We are tricked into seeing intelligence that isn't there. However, I think the big optimistic difference between the two lies in the fact that painting on canvas hasn't changed in a thousand years, but modern AI changes every single day. One day it *will* be indistinguishable from magic, mark my words.

 

[Beoric]

I feel what Generative AI and LLMs illustrate how much of what we do revolves around the recognition and manipulation of patterns, visually, textually, etc. Along with what happens you have a dense set of pattern data to work with.

Sure, an LLM can spit out tropes, because the nature of tropes is that they are repeated, and will be encountered frequently in the dataset. What it can't do is intentionally subvert tropes, choose to avoid tropes, refer to uncommon tropes, or otherwise creatively manipulate tropes. If you start seeing multiple LLM generated gaming material, it will quickly become evident that it is all very similar. It is an amalgam of other peoples' work, and it will always be derivative, because being derivitaive is core to its design.

@DangerousPuhson an artificial general intelligence ("AGI") may eventually be developed, but it won't be an outgrowth of LLMs. LLMs are a dead end on the road to the evolution of an AGI. The only contribution that LLMs make to developing an AGI it to prove the Turing test to be insufficient - which is valuable information - and to demonstrate how certain areas of research might be fruitless.

I'm going on about this a bit because I do a type of work that techbros are constantly trying to suggest could be replaced by LLMs, and in discussions with these guys it is always clear to me that (a) they don't understand my industry, and (b) they don't really understand how LLMs work. And it is really annoying. "But you could do X!" dude, I don't need to do X, nobody needs to do X, X has nothing to do with what I do. "But it could do Y!" no, it really can't do Y, incompetent/lazy people have tried to use it for Y, and it has inevitably ended badly, because doing Y expressly requires you to do something that hasn't been done before, pattern recognition is of no assistance. It can do Z, which you haven't mentioned, and I know this because we have been using generative AI to do Z for at least 15 years, which you might know if you had a clue about what my works actually entails.

The logic seems to be, "You use words, LLMs use words, therefore LLMs can do what you do." And it is no more true of LLMs than it is of parrots.

 

[DangerousPuhson]

LLMs will surely improve, change, and evolve, and probably end up as something wholly alien to how we see them today (they'll also probably be called something else, and have the usual technological branches and dead-ends throughout). I think declaring "it will never do X!" is shortsighted - we didn't expect the modern smartphone all the way back in the '90s either. Or internet before that. Or Windows before that. Or touch-dial telephones before that. etc. Invention means something new, something not yet seen and not yet even conceived.

Once we said we'd never go to the moon, too.

This stuff is still in it's infancy, which means A) it has all the room to grow, B) a lot of people are dumping money into it, and C) we have no good product lineage with which to compare it. We just have to see what shakes out, because with all the investment *something* is going to shake out of there; my philosophy is to adjust to it as it comes.

At any rate I expect LLMs will be only a small adopted part of a proper full AI - a singular processor within an unfathomably complicated matrix of processors. One tool housed in a warehouse-sized toolbox. To declare it the solution to all problems is like declaring a hammer as the solution to all problems.

As to these guys saying LLMs could replace you - they're dumb. Modern LLMs can't do that; if they could, it'd be done already, and you'd be out of a job.

It might be able to one day, though. To any of us, really.

 

[squeen]

Thanks for thinking of me @The1True. Seriously, I do love that my crumudery lives rent free in your head.

Hard agree with all the goofy player nonsense in the video.

Here's a fun aside: last night I started an AD&D game with two old friends that never played D&D together before in LAKE GENEVA. We saw the D&D 50th anniversary exhibit in the Lake Geneva museum and stayed in my brothers condo (where Luke Gygax bar tends).

We used straight 3d6 down the line with two re-roles of 4d6, keeping the highest 3 (no undo). There was a natural 3 in strength followed by a natural 18...wait for it...in Charisma! And you know what... THE STATS DIDN'T MAKE A FLYING F***KS BIT OF DIFFERENCE!

 

[DangerousPuhson]

last night I started an AD&D game with two old friends that never played D&D together before in LAKE GENEVA.

Cool stuff, but I do find it kind of funny that we were are like "who the heck would pay to play D&D in a castle?" just a few days ago, and now it's like "wow guys, I played D&D in WISCONSIN!". Just funny, that's all.

 

[The1True]

last night I started an AD&D game with two old friends that never played D&D together before in LAKE GENEVA.

We must all make the holy haj at least once in this lifetime.

 

[squeen]

Cool stuff, but I do find it kind of funny that we were are like "who the heck would pay to play D&D in a castle?" just a few days ago, and now it's like "wow guys, I played D&D in WISCONSIN!". Just funny, that's all.

convenient half-way point. Grew up north of Chicago that's why my friends started playing OD&D around 1975 from the LBB, which somehow had trickled into our neck of the woods (about an hour away from LG).

Wisconsin is beautiful this time of year!

 

[The1True]

Wisconsin is beautiful this time of year!

Someday man, someday.

 

[DangerousPuhson]

Someday man, someday.

We're already in Canada, my dude - Wisconsin is just Canada Jr.

 

[The1True]

Bro! There's like a solid 10 hrs of Ontario just to get to "South" Detroit (Windsor). Don't even get me started on air travel within North America. I've been living in some far off places and paying hundreds of dollars more and adding many hours and multiple stopovers onto my flight times to go around the TSA in all cases. Literally rather be doing mortal combat with Chinese seasonal workers dressed in 8 layers of clothes in the hot, dusty passport line at Addis then dealing with the savages at Newark.

But yeah, I've been trying to convince the boys to take a bucket list trip to GenCon (or has that moved to the West Coast now?) for years now.

 

[squeen]

But yeah, I've been trying to convince the boys to take a bucket list trip to GenCon (or has that moved to the West Coast now?) for years now.

GaryCon is the one in Lake Geneva now. Much smaller, I believe. I may go to it next March. It would be cool to meet up if you're in town.

What should be the big draw to the faithful was the Dungeon Hobby Shop, but it is shut-down/closed these days. The only D&D site is a small exhibit in the main Lake Geneva museum. The gal selling tickets there knew Gary from pre-D&D days, and chatted with us a bit about him. It was folks-y and quaint, just like the whole "resort" town. Good food!

 

[The1True]

It would be cool to meet up if you're in town.

I've got a couple more years of Africa ahead of me. Malrex' dream of getting the whole forum fam together for D&D is worth pursuing though, I think.