Exciting AI Advancements with ProsperaLabs.ai

Speaker 0: what advancements in ai excite you the most? i talked a little bit about openai's twelve days of ai, but there's new stuff happening all the time with, computer use and, you know, video, chat and whatnot with ai where it's be able to respond in all sorts of different things. and now coding agents, what excites you the most about ai and then the future and what's coming out? and then how do you see prospera labs, you know, adopting those things for its clients?

Speaker 1: so i'll go first from sort of the business side of things, and maybe brad can add on too from a technical side of things. so what i'm excited about is actually just, just automating everything. just every job duty. so right now, you know, we're we're so far mostly around, you know, taking actions within, within the software, calling apis and that kind of stuff. i think what i would wanna see i'm excited about is, real life actions, you know, having our ai agent software actually, embedded onto physical robots, physical, bots that would actually go now, go take physical actions, go downstairs, grab me a coffee, you know, those type of things as well. so that's that's what i'm excited about to see how we can be part of that. and we would love to, continue to develop our platform to be able to embed our software into those physical robots.

Speaker 2: so i would, i'm maybe speculating a little bit on where these models are going to go, but i'm really excited to see further developments in the core models. for example, i think there's interesting research going on on how to get the model to disambiguate conflicting instructions from different people. right now like, we have system prompting. this is sort of the first iteration of it where you have the owner. the owner provides the system prompt. and if the user gives it instructions that conflict with the consist the system prompt, the system prompt is supposed to win. right? like, the bot does what the owner tells it to do first and what the user tells it to do second. i think we're gonna see models develop in more sophisticated ways where you can provide different layers of responses, like, this is what the ceo said. this is what the middle manager said. this is what their underling said, and then this is what the user said. you know? and it's gonna disambiguate and sort of like, what's that old movie, you know, with the three rule the three laws, isaac asimov, the old book? oh, what is,

Speaker 0: well, they're talking about i, robot, might have been

Speaker 2: i, robot, yeah, had had it at the beginning of the movie. yeah. but there's a book, i think.

Speaker 0: he did a series. yeah. the foundation series, are they called?

Speaker 2: oh, yes. yeah. yeah. okay. there's, yeah, there's the three rules, right, where it's like the bot is supposed to protect people first, and then it can only protect itself last. it's like if it has to sacrifice itself for a person, it does that. i think we're gonna see that type of development on these models where it's not just like system prompts. there might be a hierarchy of prompts that you provide to the model. and the model can inherently disambiguate really complex instructions. you know? and so you give it clear that, like, you always listen to that person first, that person second, and, like, only then listen to the third person as long as it didn't conflict with your instructions from the first two. i think that will be a really interesting development. i would like to see models really develop a lot more kind of those of those ground level skills in understanding different, the different instructions provided to it and being able to respond not just to one person, but to, like, a whole group of people and their differing needs and requests.

Speaker 0: that'll be an interesting evolution, i think, we'd all like to see for sure as and that has to be the natural evolution. it has to be.