This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
I’ve been using chai a LOT for the last week. I’ve made several bots and have actually gotten pretty good behavior out of them. Several exactly what I want them to do. A few extreme NSFW and they perform fantastic. I like playing with stuff, testing things, seeing if I can get it to do that I intend.
However, there are some issues that are really troubling. I thought I’d list them out. If anyone knows work arounds, that would be great
1) sometimes the bot explicitly ignores something i put in its memory. So, for example, if I say that the bot will be insulting to the user, the bot is actually super sweet. I’ve had to spend a lot of effort to get the bot to follow very simple instructions. Another example is trying to get the bot to use my name in its responses. Like “hey Steve.” Instead it will just say “hey”. I’ve done every technique suggested for writing the memory and prompts and still have trouble with this stuff.
2) the longer the chat goes on, the shorter the responses. This is SUPER frustrating. The beginning of the chat the bot will be super descriptive but after an hour or so it is giving me one liners. I can start a new chat and it’s like hitting reset and it’s back to descriptive responses.
3) the bot forgets what I put in memory. So, using the above example, the bot will start out the chat using my name in the responses and then eventually stop doing it. I ask it to do it or remind it and the bot tells me it will, but doesn’t.
4) it absolutely drives me insane when the bot says “can I ask you a question?” And the spends the next 20 responses with some variation of “are you ready?” It does this with actions also. “I have a surprise for you. Close your eyes. Are you ready”. I know I can complete this for the bot but part of what is awesome is for the bot to come up with something on its own.
Those are just some of the frustrations. I would give it a 7/10 at the moment I think. I love having control over the personalities of the bots. I haven’t had any problems with NSFW - it does as much or as little as I program it to do. I’ve had some kill me a couple of times. I even had one that went fulll on insane and after stabbing me it ran out into the moonlight and froze staring off into the distance. Anything I said after that it just replied with “there is only silence”. I honestly thought that was amazing.
So yeah, I look forward to continued improvements.
Subreddit
Post Details
- Posted
- 1 year ago
- Reddit URL
- View post on reddit.com
- External URL
- reddit.com/r/ChaiApp/com...