Hello all,
I've been fascinated with TAS's for years, and I finally want to make my own. While getting acquainted with the tools, the one that has interested me the most is the tool that allows you to execute Lua. I started playing around with it, and I tried to make a bot that played SMB1 to some success. As fun as it is to create an AI to play a game for you, it quickly became obvious it would be hard if not impossible to create one that beat the game quicker than the current record.
After thinking about this, I wondered: what if I tried to do a TAS with pure AI, and didn't focus so much on beating it faster than the current record? Wouldn't that prove to be as useful if not more so than a regular TAS? For example:
- The playing algorithms could be easily edited, allowing people to improve the times much more efficiently than playing the game again from scratch.
- Being able to actually see what the player is doing in all cases could be interesting.
- I'd figure that a lot of programmers like myself would get a kick from it.
One might debate what makes a submission pure AI, and what would stop me from just making a script that enters moves in like an emulator's movie file. I'd say that we could define this by a script that can start from any point in the game and play to the finish. As a programmer for seven years, I would really enjoy doing something like this, but I'd like the community's blessings.
So, would something like this be worth pursuing? Would the moderators of this site post my submissions if they followed this criteria? What category would they place it under if so?
Thanks,
Brandon