Exploring Ethical AI

Speaker 0: i recently discussed on our silicon scoop podcast, the recent debacle about google gemini telling, some of the people working on it that please die. and that leads me to weird ethical considerations is the business needs to keep in mind when implementing ai and automation when, you know, things can go off the rails quite quickly with ai sometimes. how do you balance that make you know, with the progress in making sure that, ai stays on track?

Speaker 1: so i can speak to this from an engineering perspective. i think that, there's a lot of work that we do to actually put guardrails around ais, not even necessarily from an ethical constraint perspective, although that is one of those pieces. but often just in the process of getting the bot to do it what you want, you have to put a lot of guardrails around it just to get it to work. forget ethics. if you add ethics on top of that, the once you put all of these constraints on just to get it to work, the bot really cannot violate your rules. if your rules and what you want the bot to do is unethical, the bot's gonna do your unethical biddings. and if you have values and integrity as a person and you put rules around the bot to, impart those values, then, i think the bot does what it wants. what we shouldn't do is just let these bots go, like, willy nilly, like, you know, with these recursive self expanding plans. you know, they make their own plan and then execute their own instructions. those things are a recipe for disaster. that's these bots hallucinate too often. you know, if it, you know, one day hallucinates kill brad in the plan and then just, like, matake starts meticulously executing its own plan, yeah, maybe i do get killed. so, i think the but, like, that particular setup, which although it sounds crazy, i'm not so worried about it because it also just doesn't function. like, it as well as it being, like, crazy from an ethical standpoint, it's also crazy from, like, a product design standpoint because it doesn't fulfill the actual purpose that the customers want. you know, customers want bots that they have tight control over, not bots that there are going loose and willy nilly and inventing their own plans and whatnot. you know, maybe that is someday where we're gonna get. and as these models get more accurate, that's what we're gonna get. ultimately, we may be able to do these self recursive expansion loops. but i think right now, our process of just doing basic, you know, ground quality engineering on our bots has the side effect of causing those bots to be imparted with our values and our ethics. and, so i i don't think we should be considering the ethical constraints specifically as related to ai, but rather we should be considering the ethics of the people who are building ai. and are these people of character and integrity? and are they considering ethics at all when they build those systems?

Speaker 2: right. and it's no different than what businesses would have to consider today in terms of, you know, ethics. it's it's the same thing what they would have in place for their human employees, human resources. right? same things. trainings will go into place. there's rules and regulations, dues and, you know, don't dos of the business. and similar way, you would have to make sure your ai agents are working within those, those rules and regulations as well.