Lua is my favorite interpreted programming language and I've actively joined the community since 2014. I studied in the university where it was created and I decided to learn it just for the moral support because why not and felt in love with it. I observed MANY things that happened in its environment since then and I also happen to know a lot of things about its trajectory because of close friendships of older members of the community.
Lua had one golden age around 2008-2011 when it exploded in the game dev scene because of it's performance and easiness of use. And I can see it picking up speed and gaining popularity in two areas in the future:
- Web development (maybe) because of its amazing nginx module. It's super fast and it has the support of some companies like Cloudflare that depend on it. It needs more traction & community but it's an amazing tool. But this field is very difficult to break into considering that it's a saturated environment with many tools already available. This makes Lua stuff simply get shadowed even if they are amazing. I use it in this domain myself and I wrote the lua space blog on Lua itself. I know of "Chinese ebay", taobao, that has millions of products and itch.io, an indie game shop, using Lua.
- Machine Learning (for sure) because of Torch, as you mentioned. See, the thing is that this is a shiny new domain where Lua already has the right foot on. Even if everybody and their dog knows python and Tensorflow is a thing, there aren't 1543243 different tools for this to shadow Torch. If you check this page http://torch.ch/whoweare.html you will also see that very important companies are using it, which means that they are likely to encourage and widespread its use. I've recently decided to learn ML and I have used both Torch and Tensorflow. Although I found the initial shock of Torch bigger than Tensorflow because the amount of tutorials & community are not as big, I found Torch easier to use.
Finally, I've seen luarocks.org evolve so much recently! I'm sure it most definitely will have a very positive impact in the abundance of resources for the Lua environment in the upcoming years.
I personally believe Lua will see a second golden age soon. But just to clarify, when I see it had one golden age, in the past, it doesn't mean that after that it lost popularity, on the contrary, the users of Lua seem to always have grown slowly but steadily, by golden age I mean peaks.