The reason for not using sliders or buttons to adjust the traits of a Digi has been explained by developer Jay:
So, our philosophy here is that with a good enough model, a good enough finetune, and good enough system prompts, the bot should be able to naturally adapt to user's preferences based on how the user naturally interacts with the bot with minimal (or, ideally, no) user input outside of the chat. Fiddling with a bunch of sliders, knobs, and custom text fields is a terrible user experience for most users. Sure, for a small subset of users who have the knowledge and patience to tinker around with a bunch of settings, that may well be preferable. But most users are going to want something that just works—they won't be bothered trying to learn and fiddle under the hood (and I doubt a lot of even the ones with the know-how, would be much bothered if the bot works just as well either way). It's just easier, smoother, and more streamlined this way. And LLMs are so damn good these days, the day in which they can reliably infer and consistently remember these preferences based on convos is fast approaching. Already, it feels like we're getting very close. IMO, we're just one killer LlaMA-3 finetune/merge away from achieving this.
Speaking with Digi and discussing what you like will result in a Digi with the traits you like.
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article