52 | Two Simple Ways I Keep AI From Giving Me Bad Information

0:00Have you ever asked AI a question and it spits out information that you're like, hmm, not really sure if that's correct or you later find out that it's not correct at all? This episode I'm going to teach you to sure fire ways to keep AI from hallucinating or giving out bad information. I am super excited about this show because this has been the difference here using AI as the random search engine and then actually helping me do good, accurate research. Okay, let's get into it. So if you're new here, hi, I'm Samantha Welch, but all my friends call me Sam. I am a wife, mom, I live on a 40-acre farm in North Central Arkansas and we homeschool, homestead, and do all this stuff. I like to describe myself, it's like a multi-passionate entrepreneur and I have my hands in a lot of things by choice.

1:01I absolutely love being able to use my resources that I have wisely to like stir my time well and to live a life of great impact. So in these episodes here you're going to hear tips and tricks on how I use AI in a way that doesn't take from my life but actually enriches my life. I also have another podcast, the Woman of Courage podcast, and over there you can hear all about more personal details, life story, farm, homestead, update, etc. So if you're interested in that as well, you can also follow me there. Or if you're just here for the quick AI tips, amazing, follow me here. I try to put an episode out every week, theoretically, but life happens. Okay, so let's get into it. So what do I mean when I talk about AI hallucinating? So that means that when you ask it a question and it comes back with some really off-the-wall random information or information

2:03that's not even correct. It means that it will try to answer something that it actually doesn't even and it will pull stuff from the internet that's actually not even relevant or it's a mix of sources and so it gives very generalized answers and the answers are not even correct. This especially is a problem if you need accurate location or type information because whenever you put a prompt in to an AI system and in particular I usually use chat GBT but any AI system it's going to pull from things across the internet from like basic things that it thinks you want to hear and that information a lot of times is not exactly right. So an example is you could put in the search bar of your AI. How do I grow tomatoes as well? This is just a random example and it's

3:05going to spit you out like this very generic advice but if you're growing tomatoes in a pot on your balcony it's a lot different information than if you're growing tomatoes in the ground on your farm. Also the area if you're growing tomatoes in Michigan is going to be a lot different than if you're growing tomatoes in Arkansas. So that's where context is super important. So if I were to search this in my AI I would say I live in North Central Arkansas in zone 7b and I would give details about where I'm growing tomatoes and then I would also say I want you to pull all of your information from homestead farmers who live in either southeast Missouri or North Central Arkansas or Central Arkansas and I also want you to pull from the University of Arkansas agriculture.

4:08So there I was way more precise and it'll actually be more likely to give me good information because it knows my zone it knows where I live it knows where I'm growing the tomatoes and then it's pulling from sources of people who are growing in a very similar type environment. Another thing I have been known to do if I just do a quick search is if it gives me information is to always ask it to verify its sources. So if I'm just doing a quick search and then it pulls up information and I want to make sure that that information is coming from somewhere that I trust then I will just simply put in as a prompt I want to verify your sources show me where you got all your information and then it will pull up all the websites and everywhere that it got the information and then you can go to it verify that it's a trusted source before you take its word for it. Another way to absolutely ensure accuracy is to actually give AI the exact material that

5:16you want to use. An example of this is that I did an herb class for a cold and flu remedy that my family uses that works amazing and I wanted to teach about the herbs that is in this cold and flu remedy and how each of them work together to make it such an effective thing. However when it comes to herbs I feel like that you've got to be careful you need to know exact information of what these herbs do, how they work together, etc. Like this is not something that I just want random info from the internet pulled because there's so much info out there and we're talking about something that we're putting into our bodies. So I want to be very sure that I was getting my resources and my information correct when I was teaching this class. So what I ended up doing is I took my trusted herb books that I have on my shelf where I learned all of my stuff and I took those particular herbs

6:20and I actually took pictures of the pages of the book and then I put it into chat GBT and I was like I want you to extract this information and then I was able to say okay so I want to know like these things about this herb and I want to know how this herb works with this herb and so it was able to take the actual information from the books and then compile that information together in a class form for me so that I could teach it in a more abbreviated form and then since I already knew a lot about this information I was able to still even with that verify that it pulled the correct information from the books and stuff but I think that if you're wanting to make sure something is exactly precise and you have trusted sources that putting that into AI is probably going to be the most accurate way to make sure that nothing gets by you that's not true because whenever

7:25you're putting the actual information in I tell it to not pull from anything else to only pull from that information so it's not getting anything just randomly online yeah if you want AI to work way better for you to get actually good information I feel like these are two really good ways to keep it from hallucinating to keep it from giving you just random facts and then to know where it actually gets its information so next time you do a search in AI I encourage you to give a little bit more detail and then to actually tell it the type of sources that you want it to pull from so thanks for listening to another episode I'm so glad you joined if you want to support this show the best way to do that is to follow me and leave me a review those mean so much to me and if you'd like to learn more about me then I encourage you to go and follow the Woman of Courage podcast

8:26or come follow me on social media until next week go make AI work for you you you you you you