The Back Garden wants to be a “more experiemental” place for IF, but refuses any works with genAI. Probably 90% of today’s IF experiemental stuff IS from using genAI. Really makes the Back Garden seem quite lame. If there was one place in the IF world open to experiemental LLMs, the BG would have been the right place.
This discussion has been done to death and the upshot is this: no-one with a significant stake in Spring Thing is interested in engaging with AI-generated works. You can claim all you want that it would be “right” to allow AI works in the Back Garden, but at the moment, there is no-one from this community willing to run such a competition and very few people interested in reviewing or judging the games. So if you’d like that to change, you’ll likely need to take the initiative in organising a more AI-friendly competition yourself.
27 Likes
Yes, this has been discussed here and in the threads linked there:
TLDR: Some people want to enter AI games in comps, but few to no people are interested in playing and judging AI games, so the best thing to do is probably for the AI people to make their own comp for AI games. So far nobody has bothered to do that, which probably says something in and of itself.
22 Likes
Yeah, as much as it would make sense philosophically for the Back Garden to be open to LLM use, judges have outright said the presence of LLM-generated entries turns them off the comp entirely, and without judges the comp can’t happen.
In the end, it seems like people enjoy generating games with LLMs a lot more than people enjoy playing LLM-generated games.
21 Likes
Heck, short of a competition or releasing actual good games people want to play (which to bend over backwards to be generous, I suppose can be hard with a newer technology), I think it would really behoove the pro-LLM folks to like have conversations about design and theory and critically evaluate their stuff. If 90% of experimentation in IF is happening with LLMs, it’s very striking to me that many of them recapitulate dead-ends in design that the mainline IF community noticed, discussed, and discarded decades ago. But while I’ve seen tons of discussions where pro-LLM folks argue about getting their stuff into the existing events, I think the only actual critical discussion I’ve seen of them is when reviewers of those events point out the problems, issues, and errors they’ve encountered, which universally leads to crickets rather than constructive engagement.
It’s maybe a bit too on the nose to argue that people who are very into LLMs aren’t willing to put in the actual work required to make good games and build a community, and just want to get praise and support from the existing community regardless of merit or effort. And I dunno, maybe there are these kinds of conversations happening elsewhere, with emotionally-engaging, well-written games with compelling things for the player to do are thick on the ground outside our tiny corner of the Internet. But I sure haven’t seen it!
23 Likes
I suppose those of us making experimental games with our own hands were wasting our time.
10 Likes
For what it’s worth, the anti-AI sentiment isn’t limited to this community. It’s become the dominant position among mainstream gamers.
7 Likes
It’s a major excepton on my “staying aloof of the gen AI debate”, but I must thank Ivan for the link to Polygon, because in the same news outlet is an article directly involving one of the most complex element of IF coding, the NPCs:
as everyone know, my major WIP involves a pair of rather sophisticate NPC, ideally exploiting to the hilt the powerful NPC handling of TADS3/Adv3Lite, and generally speaking, IMO a good NPC must have a complex expert system behind his/her/it/they reacting & behaving, so as I noted when proclaiming my aloof stance, generative AI is a thing, coded AI is another (is not easy as seems insuring that a NPC don’t refer to earlier state of world (e.g. referring to an unlocked door as still “locked”), or, definitively worst, referring to future state of world (up to major spoilers…)
and in the major mess people forget the third AI: the assistive one: I fooled with blender for a little while, but in the end, I came to the conclusion that in a complex creative work as drawing 3D scenes there’s the need of automated check (e.g. that a polygon is actually closed) and ease of laying out (my ideal 3D editor should have a commandline and parser capable of handling instructions like “move the table3 object 3 meters left, then rotate it 20 degrees right”, and warning if the table ends violating the basic Newtonian law of the impenetrability of bodies; in other words, I want inform 7/10 in lieu of autoLisp 
)
Hope this led to a constructive debate, and
Best regards from Italy,
dott. Piergiorgio.
2 Likes
IFComp has so far always allowed this, and attracted criticism for doing so. If the LLM people are just going to ignore it in favour of whining about Spring Thing’s rules, I don’t see the point in IFComp continuing to stick its neck out at this point.
12 Likes
I, for one, will be spending more time on Spring Thing than IFcomp for exactly the reason that it doesn’t include genAI
8 Likes