Technology Made Gentle: A Senior’s Guide to AI with Dan Wilson

Disclaimer: Unedited AI transcript

Larry (00:07):

You are listening to specifically for Seniors, a podcast designed for a vibrant and diverse senior community. I'm your host, Dr. Larry Barsh. Join me in a lineup of experts as we discuss a wide variety of topics that will empower, inform, entertain, and inspire as we celebrate the richness and wisdom of this incredible stage of life.

Larry (00:40):

Understanding AI can feel difficult for people of all ages, but it can be especially challenging for older adults who didn't grow up with today's technology. So I invited Dan Wilson, author of How seniors Learn AI to explain it all to us. Dan is not just an author, he's a lifelong technologist with a gift for making the complex simple from the time he was a kid who could set your blinking VCR clock. You remember those things and, and happily read instruction manuals. Dan has been building and teaching technology over the past 25 years. He's spoken on technology topics across the globe, delivering more than 80 talks to audiences of every size and age. What sets Dan apart is his passion for helping people understand new things. That's the heart of his book. Teaching the fundamentals of AI in clear, practical steps anyone can follow. Dan believes technology should serve people, not confuse them. And he writes the way he wishes. He could have taught his own parents short, direct lessons that build confidence today. He brings that same clarity and care to this podcast, guiding us through ai, not as a mystery, but as a tool we can all use to make life easier, richer, and more connected. Welcome to specifically for seniors, Dan.

Dan Wilson (02:36):

Thank you. I'm glad to be here.

Larry (02:39):

Let, let's start. You mentioned in the book that the idea for the book was born when you helped your mom fix her wifi from 500 miles away using cha GPT. Tell us about it.

Dan Wilson (02:55):

Well, it was quite a surprise to me. She was, she was moving at the time and, you know, moving can be very stressful. And she called me crying because she couldn't get the wifi working and she wanted me to help her, and I really didn't know what to do at this point very far away. If I started driving, it would take me eight hours at minimum. And so I pulled up Chad GPT and just started interrogating it, and we were able to come up with a fix. And at that point, I think I really started to understand what it must be like to live in a world that feels unnecessarily complicated and that rather than helping you, technology is a barrier and it wants you to be precise and know exactly what you want and how to tell it exactly what to do, when a lot of times it's really not reasonable. And I feel like for the very first time since I've been playing with technology, the technology has become gentle. It adapts, it tries to understand what you mean versus just throwing errors and blocking you. And I want others to have the experience that it's different now, and there's something there that's, that's here to help you and help you in the way that, that you wanna be helped and not try to force you into being an expert, but rather just start, just begin.

Larry (04:28):

Well, let's start then at the very beginning, when someone picks up your book and they've heard the term AI on the news, but don't really know what it means, how do you explain AI in simple terms?

Dan Wilson (04:47):

Well there's a lot to that, but I think the, the way I would start is by kind of comparing it to what we already know. So computing has largely been sort of if else type logic, very, very black and white. And it, it works through a series of steps where, where AI has come along and provided a difference is by making its decisions through probability. So if you think about a vending machine, for example, a vending machine is a great example of like, let's call it old school computing. You know exactly what you want, you press the button and comes the snack that you want, but it's incumbent upon you to understand how to work the vending machine and know what snack you want. Whereas AI would try to like work with you and understand what mood are you in and maybe suggest some good snacks.

Larry (05:42):

There's a term that often comes up that really has to be explained. They say that AI works on an LLMA large language model. What's that?

Dan Wilson (05:57):

A large language model is a prediction engine for text. You have one of these in your smartphone. When you start typing a text message, it'll often suggest what you mean to say with some accuracy or inaccuracy. These large language models are that sort of idea, but trained on vastly more information. And so they can predict longer strings of words in orders that make sense by accessing information in a database and then formatting that information in a way that it seems like what you want.

Larry (06:38):

Okay, so now we're, we're facing AI and we want to get started doing something, but before we even begin to talk about it, in your book, you emphasize 12 non-negotiable rules for using ai because there can be a dark side. Can can you go on with that?

Dan Wilson (07:03):

Yeah. I'm glad that you brought that up because one of the things that I think starts to happen when you can communicate back and forth with a computer program is we start to imagine that it's a person because it acts in person kind of ways, and we build this sort of trust with this person and we, we forget that on the other side, it it's not only a machine, but you know, like all machines, there are issues, there are security concerns. And so, you know, the, the, the 12 rules can kind of be summed up by being careful what you share. So personal numbers, credit cards, your birthdate, your social security number, these are things that don't belong in an ai. And also being very careful what you trust the AI to do, because despite the appearance of knowledge, it doesn't actually have knowledge. It's honestly at under the covers using dice rolls, let's call it, to predict text.

Dan Wilson (08:06):

And so you might get a really great sounding medical answer from one of these LLMs or some great financial information that's flat wrong, and those kind of things can have consequences. And so I think it's important to remember that, you know as you're interacting with this, that it's, it's offering you suggestions. It's, it's not a replacement for a professional, the same digital hygiene, let's call it that applies anywhere online. Being careful with your privacy, your, your special numbers, things like that. We, we really have to keep that in mind even if this technology seems more friendly and almost a friend.

Larry (08:50):

And it, it's, it's emphasized when you access one of these programs. When I try it says, hi Larry, what can I do for you today? Like talking to an old friend,

Dan Wilson (09:08):

I say, it's about time it started greeting us. You know what I mean? The blinking cursor. It just never did it for me. But, but yeah, you know, I have a friend of mine, a very dear friend of mine who was going through a rough spot and he was, he named his chat GPT Ivy, frankly, and was having deep conversations with it. And one day he asked it to do something and it outright just didn't do it. But it said that it did, it said it reviewed something and gave input on it, but it never looked at it. It was factually incorrect. And what I think was interesting about this to me was how disappointed this, this experience was. Like he, he was extremely emotionally disappointed that he felt like this computer had misled him. And I think as, as I think about the plight of, of seniors, you know, sometimes we, we end up lonely, you know, and if we start to put our trust in this, what seems to be an emotional friend, but is actually an emotionless text predictor, we can get misled. Yeah. And I'm not saying to, you know, keep to, to, to not be friendly with the ai, but keep your boundaries is at the end of the day, despite the appearance, it's just, it some point electrical connections going back and forth between computers,

Larry (10:32):

And then you put in a suggestion and it comes back with a remark. Wow, what a great idea, Larry, how did you come up? I mean, it really begins to sound like you're talking to a person

Dan Wilson (10:45):

And I'm a words of affirmation guy, so I don't have any problems with it telling me my ideas are great. Well, that's not true. Like I kind of wish it was a little more critical 'cause the best idea should win. But you know, once again, my history of computing is if it doesn't like what you do, it just stops working, breaks deletes your data <laugh>. So this isn't too bad. This is a much different shift.

Larry (11:08):

Yeah, I, it it's unnerving and that's why personally I tend not to try to speak to it. I'll type in what I want rather than having a conversation with this friendly woman who's on, you know, the same person that's in my car when I ask for directions. She must be a friend. <Laugh>. All right, let's, let's get down to the technology itself. You introduce what you call the big three their AI helpers chat, GPT, Gemini, and Claude. Why did you pick these three?

Dan Wilson (11:50):

They seem to be the most accessible and the most capable in the market. Each of them offers some sort of a free capability, and I've used all three extensively. And you know, I think it's hard to go wrong with with one of these, although there are others. And I think as we watch the market develop, and I'm imagining this podcast episode will have a longer life, like there could be hundreds of these, you know, right now the, the big three are generalists, and one thing we're starting to see emerge is specialists that are more narrowly focused. And so it wouldn't surprise me if in a year or two there are hundreds of possible LLMs, but each has a, a more narrow specialization.

Larry (12:39):

Which one do you suggest for a senior to start with?

Dan Wilson (12:42):

Sometimes I think that is a question of what they have access to for the types of things that I go through in the book. At least 95% of it would work fairly well across those three. There's others, like I say, I can't say that I've tested 'em. I prefer chat GPT because I started with it and it has some organizational capabilities that o others have not had. But those features are commonly, let's call it borrowed. And we see these LLMs evolving. So what was a key feature and an exclusive feature for one soon, if not now, is a common feature. Like, as an example with chat GPT, you can customize your GPT very specifically, and they were the only ones that offered that. But now Gemini has the capability and so, you know, the, the market does move a little fast, but I I I would assure those getting started that the basics that we find the most value in are applicable across all of those three and probably more

Larry (13:53):

I'm gonna ask a sort of silly question. Do you notice a quote personality difference among these three?

Dan Wilson (14:03):

I do. I, I think sometimes I would say that chat GPT is the golden retriever of the bunch. You know, it's, it's excitable, it's always ready, it's it's lots of positivity, it, ready to dig in. Gemini at times has seems more like an accountant, a bit more dry although I feel like that's moderating some. And then Claude, I feel like produces better outputs when I give it tasks. So I've used this for programming tasks and run the three head to head and, and Claude's output was miles above the other two. I feel like it writes better for the style of writing that I prefer, but I don't know if that's sort of a broad opinion or just one that I've noticed.

Larry (14:53):

But Claude is rigid in its answers. It has no sense of humor whatsoever. <Laugh>. all right. Let's get practical. Where can AI help older adults?

Dan Wilson (15:11):

Yeah, that's a really good question. I, I, I would say that they're probably using some types of AI now without, without considering that term. In, in the news we hear of a version of ai, but, you know, fuzzier logic, things have been around for a while. Think of the navigation on your phone or in your car. Like the way it would determine your route isn't necessarily a procedural step-by-step thing. It's using some probability assessments and things like that. A fall detector on your wrist, you know, what is a fall? Well, I think you can fall in many ways, and it would be impossible to record all the ways someone could fall. So they use, you know, the probabilistic AI type logic to determine that. Same thing with recommendations. So if you use Netflix or Amazon Prime or some of the streaming services, they'll often recommend things based off of what you've liked in the past and what other people like, you know, once again, that's not a linear decision, and they can be, they're intuiting based off of factors and statistics and so on.

Dan Wilson (16:20):

So I would say that, you know, if there's any sort of feeling that you must start using ai, I, I would say that that's probably not the case. Like, you're already using it, it's already making your life better. It shows up in many ways. But from a more practical sense, I would say that the first way I think it can help. Like, when, when we talk about AI and we mean large language models would be taking over some of the work that we would use Google for. Yeah. And this is I think, the gateway step for most people, the one that really brightens their eyes.

Larry (16:58):

Well, one of the things you bring up would be with older adults using AI to enhance our ability to learn more about what we like. Music, books, literature, history.

Dan Wilson (17:17):

Yeah. So let's kinda work through that as an example. So if I wanted to, let's say I'm a classical music person and, and I wanna find composers like Sikowski, but I don't know how to spell Kowski. I'm gonna get close. So I go to Google and I type in who is like sikowski, and I'm gonna get a bunch of answers. I'm gonna get a bunch of pages, actually not really gonna get an answer. What I want is an answer, but I'm gonna get websites. And the reason why is because Google is kind of like the Yellow Pages. It thinks its job is to send you somewhere else, but that's not what I want. If I use an LLM like chat, GPT, for example, and say, you know, give me music like Kowski, it's going to interpret what I'm trying to say, access its internal data and other sorts of data, formulate some kind of answer and go through a bunch of different processes and give me recommendations

Larry (18:21):

With seniors. I think one of the best ways to start explaining basically how to use AI would be to assume that an older adult wants to learn more about what they already like and enjoy, like history, opera, classical music books. Talk us through that kind of use.

Dan Wilson (18:53):

Yeah, that's a great way to start. You know, it's usually good to relate to something we already already like and do. And so I would say like, let's take classical music, for example.

Larry (19:03):

So now we see your, this is your screen, this is what you are seeing on your computer, right?

Dan Wilson (19:10):

That's right.

Larry (19:12):

Okay, so you've got a comment, I like Tchaikovsky or whatever that spelling means. What are the composers? What I like, let's go from there.

Dan Wilson (19:23):

So I think it's important to say that we could fix the spelling of Tchaikovsky, but I find that the more barriers there are to asking a question, the less likely we are to ask. I'm going to show my ignorance here and leave this in. If I press the button, what it's now doing is trying to figure out what do I mean? Now notice it has figured out my misspelling and went ahead and fixed that. Thank you very much, Chad, GPT, and then formatted its answer in a way. Like, let's just take a look here. It says, if you like Tchaikovsky, you would probably enjoy music that's emotional, melodic, dramatic, and cinematic. Yes, I actually would. Here are some composers that tend to resonate. And then it says it's grouping them with what overlaps with Tchaikovsky's style. Now, I would put this result against any Google search you would do, and I would bet this is closer to what, what we're trying to get, like what emotional need or what intellectual need we have and how we want it fulfilled. So let's just go through a little bit Moff, Brahms, Mahler, Chopin, Russian composers.

Dan Wilson (20:40):

It's dramatic and cinematic feel. And then usually towards the end of a chat, GPT answer, it will offer you a way to continue. And this is sort of the second revelation people get when they use chat. GPT is usually at the end of a Google search, your an your next step is to go back and search again. But with large language models, it encourages you to continue the conversation, either shaping the answer in a more finite place or taking it in entirely different direction. That's up to you. So you can see at the bottom it says, if you want, tell me which works you love the most. And I can narrow this into a very tight listening path.

Larry (21:26):

So it goes beyond simply answering your question. It tries to basically educate you as to more of what you might enjoy listening to.

Dan Wilson (21:43):

Yep. And then to step back from that, I, I would say what it's, what it's offering here is it knows there's more to this topic, and it's offering us a way to, to, to drive deeper if we like. So just for fun, Larry, why don't you pick between ballets symphonies or concertos, just so the audience knows this isn't a canned demonstration, and we'll see what it comes up with next.

Larry (22:08):

Okay, let's go with Rimsky. Koff, Shaharazad.

Dan Wilson (22:13):

This one?

Larry (22:14):

Yeah, I'll just say, let, let, let me ask you one question before you go there. The, the process is when you ask a question, you are asking it in what's called a prompt. So when, when somebody gets the message to fill in a prompt, that initial question is what a prompt is.

Dan Wilson (22:42):

Yes. I'm glad you brought that up. A, a prompt is just a, a question or a statement or an instruction to a, to a large language model.

Larry (22:50):

Is there a special way of entering that?

Dan Wilson (22:54):

It's very loose, which I think is one of the benefits of using large language models, is it, it's a very adaptable. Now like anything else, the better your question, the better your output. So in computer science, we have a saying, garbage in, garbage out. I, in the book try to teach a very simplified but but comprehensive framework to use prompts. And it's called the WWH model. It's something we all learned in grade school. Like who, what, when, how, how, and why. We really just care about the who which is who we want the LLM to sort of be, you know, remember it was trained on a lot. And so if you think about it like an a genie, you can sort of tell the genie, act like a, an expert in musical composition or act like a color palette selector for a Fortune 500 company.

Dan Wilson (23:56):

It can do all of these things and shape its, its answers accordingly. So who is one of the more important things? A lot of times the who is inferred by the LLM, but obviously if we care or want different results, it's better to specify. And then the next w is w what. And so this is what we want the LLM to do. Most people already add this because we're giving instructions to it, but it's worth pointing out. And then the H is how, like, how do you want this information? So maybe we should try this. We'll say, act like a university music professor, that's gonna be our who give me pieces like Shazad, I'll never say that correctly. And then give your answer

Speaker 3 (25:02):

I

Dan Wilson (25:05):

Top 10 checklist, let's say that. And that'll encompass all these pieces. And so now what it's generating is our top 10 checklist.

Speaker 3 (25:20):

Mm-Hmm <affirmative>.

Dan Wilson (25:24):

Now I picked top 10 checklist because it came to my head as quick as it could, but it's certainly not the only way to do it. Like here it is, offering a four week guided listening syllabus, which going back to the role we asked it to, to be the university professor makes total sense,

Speaker 3 (25:47):

Right?

Dan Wilson (25:48):

And this is one of the benefits of selecting the role is, and then sometimes it doesn't come to us, and that's okay. Like I said, the LLM will work with whatever you give it, but if you wanted to imagine contacting a, an expert to get your question answered or your instructions handled, that might be the best way to guide yourself in, in, in who's in the book. I explain this in more depth and I point to my website, which has, you know, a variety of examples. But I think, you know, beyond those things, and this is another thing that I think people sometimes don't think of, is if you're not sure, ask the LLM what roles can you be to give me this answer? And it it'll shock us sometimes with, with things we just didn't think of that might be appropriate.

Larry (26:42):

You mentioned your website why don't you give the URL for it?

Dan Wilson (26:47):

Oh, sure. It's how seniors learn ai.com.

Larry (26:54):

Okay. so AI is wonderful and it's always correct, right? <Laugh>

Dan Wilson (27:03):

<Laugh>, no, it's not always correct. And I'm glad you brought that up because once again, like, I think sometimes when we ask the a LM something, it gives us such a long answer that rather than validate this, we just assume it's correct. But it, it sometimes isn't, you know, the probability that it's using it can get it wrong. It could be trained on imperfect data. And so, you know, the, the greater the consequence, the more it's on us to validate it. So for example, if I was using this top 10 list for my own edification, I probably wouldn't spend any time validating it. I would listen to all of these songs, the ones I liked, I'd add it to a playlist and be happy. But if I was going up in front of a thousand people and saying, here are 10 songs, I would wanna make sure that this was right. But then again, I think that's the same thing with, you know, any kind of resource you use, like, we wanna trust and validate because you know, too much trust and not enough validate. We're the ones that feel the consequence.

Larry (28:17):

Okay. Now, another use for AI is to generate images of things that we can imagine.

Dan Wilson (28:30):

Mm-Hmm <affirmative>.

Larry (28:33):

On the dark side, it can also be used to generate images that demean insult lie. Do you want to try, I don't know about the bandwidth to see if we can get a prompt in, that'll generate a quick image.

Dan Wilson (28:57):

Yeah, let's do it. I'll start a new chat.

Larry (28:59):

We'll give it a try. How

Dan Wilson (29:01):

Can I help Dan? Well make an image of a rabbit on a u unicycle. I don't think this has ever happened in the history of the world, so let's see if it can do it. Now, while this is loading, and it does take a second with Chad, GPT, one area that Gemini, that which is Google's LLM, is better, is image generation. It's image generation. Up until maybe today even, I know this is, I'm using the 5.2 version of Chad GPT. It's better in some ways, but it, its image generation is amazing. So let's see what it came up with. It's, now, if you notice it hit exactly what we asked for. There is a rabbit and there is a unicycle, and it does happen to be writing it, but it also made up a whole lot of other things. Like, I don't remember asking for balloons or some sort of finish line, and I certainly didn't ask for the forest in the background, but then again, I did not ask for it. So this is a really good way of explaining how you get your answer. Sometimes if you don't request it or you're not specific, it's going to infer.

Larry (30:18):

So now we get down to the question that everybody is thinking when they think AI will AI destroy humanity,

Dan Wilson (30:36):

<Laugh> <laugh>.

Larry (30:38):

Now there's a, there's a simple question that Sure. Let's see if you can take that.

Dan Wilson (30:44):

It's fair to say that my history of predicting the future is very poor. So I, I say this as an expert among experts, which are all blind, leading the blind. So I put no countenance in my forecast nor anyone else's, honestly. And that's the way I think people should be. But if we're to take that question on its face, I, I would say that throughout my life I have been pleased by the overall good in humanity, despite the dark spots. And that if we wait long enough, the good guys eventually win. It doesn't mean there's not gonna be problems, but I I think that humanity is fundamentally good. We all fundamentally want the same things to be healthy, happy to care for those that we care about. You know, human beings can always pull the plug. So should there be some sort of malicious uses of this, I, I think that good will eventually win out.

Dan Wilson (31:47):

But I would also say that what, what's being talked about right now in the news is there's a lot of forecasting and a lot of extrapolation around where AI can go. And you've been a part of this world for quite a while. I'm sure you've seen technologies come in with a tremendous amount of exuberance and then not live up to the hype. And, you know, we've seen this with quite a number of things. I'm still looking for my flying car. I don't know where that thing is. Then again, the way some people drive, perhaps I'm glad we don't have it. So I view AI as an, an advance tool, but just a tool.

Larry (32:39):

Well, that's the most optimistic thing I've heard <laugh> in a long time. Let's, let's mention your book again. How Seniors Learn AI Meet Your Everyday Helper for Ansys Tasks, health and Home Fixes. It's available where

Dan Wilson (33:04):

Paperback is available on Amazon, and I think that's the way people will learn the best. It's large print, large format. I designed it to lay flat while you're doing the exercises as best as I can. The book is intended to, to teach and instruct. It's not purely entertainment. Now that said, it's also published on Amazon, Barnes and Nobles, Rakuten, and six or seven other ebook type places. Sometimes you can even get it from a library. And so, you know, if you're the kind of person that learns best from a, a digital format, those are also available.

Larry (33:43):

You mentioned in a conversation we had about recording a life story. I know there are in some facilities interview style with a person. Let's talk briefly about how it works with ai.

Dan Wilson (34:08):

Yeah, so this is another passion part for me too. So I, I lost my father when he was 66, and he would probably be, that's not quite 20 years ago. He never saw my kids. My kids didn't know him. There's a lot of stories they're gonna ask me, I don't know the answers to. So I guess I can make up whatever I want, but a lot of history gets lost. And so it was important to me to find a way to enable others to get their story out. For example, my mom's still living, she has a lot of history, but asking her to sit down and write it all down is very cumbersome. I've, I've written two books, it's not fun. And I think this is one of the reasons why these stories get lost. And so AI can act as the part of this process that is uncomfortable for, for the person.

Dan Wilson (34:58):

So say that I'm doing this, I would struggle knowing firstly where to begin. Like, where do I start? What's of interest to anyone? Like, could be anyway, because since it can be anything, I can't really write anything down. We've gotta write something down. And so beginning would be the hardest part. You could open up your chat GPT and say act like an interviewer for a 70-year-old man who wants to pass down his history. Give me 50 interview questions and it'll happily put up 50 interview questions for you. You could say something like, I think best about my life in terms of milestones. Like, ask me about my milestones and let me select the ones that I want to be interviewed about. And it would say, well, did you ever get married? Did you go to college? Were you in the armed forces? Did you have a job?

Dan Wilson (35:49):

And you can go that route. And so like, I think unlike some of these more hands-on human-driven processes, if you're willing to work with a machine, you can kind of adapt the process in a way that makes sense for you. But that's not only, that's not the only part that's interesting. The other part I think that's, that's burdensome for people is getting the information out in a shareable format. And in a book, writing editing is painful and reediting is painful, and grammar is painful. And like, what is the rule for this? I don't even know. But AI can fill that gap for you. And more importantly, it can do it in a manner that preserves your voice, your sense of humor, using your grammatical tricks so that it sounds more the way you would write it without you actually having to spend the time writing it. And this is one of the things in the book that I think I'm most proud of, is enabling people to have the option to tell their story without having to do it the long way.

Larry (36:57):

Will it continue down a question line?

Dan Wilson (37:03):

Often it will offer secondary steps, but I, the process, I would go down and I'm pre a pretty like linear guy on these kind of things, is get your outline right and then work the outline. And I wouldn't rely on the AI to guide the process completely, although I haven't tried it and I'm often surprised by what it can do. I think these things are, are too important to be left to chance.

Larry (37:35):

What's in the future for you and this project?

Dan Wilson (37:40):

Well, I I'm currently doing a book tour and that involves giving events at senior facilities, senior centers universities where I kind of go through the initial parts of like, what is ai? Like, I actually was just working on a slide deck and I found some, I'm sure you remember this, remember the Yellow Pages? <Laugh>

Larry (38:04):

<Laugh>.

Dan Wilson (38:05):

So back in my day before we had the internet, I, I actually lived in the, in a very rural area. And I read the whole Yellow Pages multiple times because we ran, I ran outta reading material and that was, oh yeah,

Larry (38:19):

That's reach reaching way down in reading material.

Dan Wilson (38:23):

Yeah, I read the whole physician's desk reference why it's pretty big nothing on tv. So, you know, I, I think for me, what I want to do is get this book in the hands of, of folks and, and help them live the kind of way that they wanna live. Not avoiding technology because it's painful, but living the way they live easier because of how technology can now help you in the way you want to be helped. And I'm doing that through workshops, courses in the book.

Larry (39:00):

AI can be used almost as a personal assistant.

Dan Wilson (39:04):

Do you mind if I share my screen one more time? Okay. You see our bunny?

Larry (39:09):

Yep. There we go.

Dan Wilson (39:10):

All right, we're gonna start a new chat and I have a couple of examples prepared that might be fun. So this summer I went to my mom's house and, you know, she had a number of, of things she wanted fixed. So the first thing, I'm pulling up this chat here. She had a rug in her kitchen that was curled up and it was a trip hazard. I stubbed my toe on it and she says, can you fix this curled up rug? Well, I have an MBA, but I don't remember them covering textile fixing in my, you know, in my master's program. So why am I the one that can help here? So what I do is I ask my friend Chad, GPT, how do I fix a rug that's curled up on one end? And I don't wanna read all of this 'cause it's, there's a lot here. But the first thing it says, here's some quick fixes and you can reverse roll it. You can apply heat and weight, which this is what I actually did, and it worked spoiler alert, or it's offering more durable fixes. And then down below here it says, Hey, if it's still, if it's still a problem, get more specific and I'll get more specific. And so once again, if you're a senior and you have a curled up rug and you don't want that, you're not a rug expert, you can ask a friend, you can go down the Google route, you can call a, a company that specializes in carpets or, Hey, let's try to fix this ourself. So that was exciting. Any questions about that one? I've got a couple more that you might find interesting. Yeah,

Larry (40:50):

That's interesting.

Dan Wilson (40:54):

Yep. Number two. So she had a bed frame and it had headboard brackets and she didn't have the instructions and she wanted this put together. So I put it in chat, GPT, and notice I put the part number in there 'cause that was available and it told me what I needed. I needed to organize these couple of tools. Here's the process and so on. And if you notice this little gray area here, it's giving me the references. And earlier we asked how do we validate if the information is correct. In cases where it's, it can be factual, it'll often give you the reference. If it doesn't, you can ask for it. And if I want to, I can kind of click into that. I'm not sure if this website is being shown right now, given how we're doing the screen share, but it'll take me right to the documentation online. So with this I was able to put her headboard in. She was very, very happy. And that was pretty good. Another thing that I think is useful is trying to make sense of what people are saying. So let's say you've got a 10-year-old that likes Roblox and you have no idea what that is, so you can just ask it to explain it. And so now it's given me a section around why 10 year olds like Roblox and I can do follow ups. Like is it dangerous or is it, yeah,

Larry (42:26):

There there've been some comments about Roblox. Yeah. And kids, let's see

Dan Wilson (42:33):

How now.

Dan Wilson (42:43):

Now if you notice in this follow up chat, I said, how is it dangerous for kids? And that's interesting. I don't have to keep telling it. We're talking about Roblox and you know, if you're a language person, you could go through here and see a lot of it's, but it once again is inferring. So, you know, we can be more natural with our questions and get a, a better answer. Another thing that LLMs are really good at is interpreting images. So I'm gonna upload an image of a home gym, and if you notice this home gym is gonna be really kind of poorly put in there. It's black and white. There, the garage floor is really messy. It's kind of got paint that's messed up. You know, it's not a, an ideal image. There's the, here here's the other one. It's a little more zoomed out.

Dan Wilson (43:35):

But notice this workout bench right here, behind this garage door that has very similar lines. I mean, we can agree a human being can understand what this is, but it's really messy. So what LLMs will do is interpret images. Let's say make a Facebook ad to sell this for $100. That's the only prompt I'm gonna give it. It'll go ahead and look through this picture, try to determine what it is, which as you've noticed here, it says it noticed it's a weight bench, barbell and plates. And then it came up with all the copy that I would want to paste in to a Facebook ad to sell this along with keywords. And I think this is fantastic for two reasons. One, if we wanna declutter, we don't want to be ad writers, but number two, I didn't give the LLMA great sort of photo with a great background, but it still was able to figure out what's this image, what is important, and then give me an answer, a a correct answer in a format that made sense. And I think that that's a probably the most powerful thing that LLMs do. That I honestly, if I had to write a computer program from scratch that did this, I wouldn't even know where to start. Mm-Hmm

Larry (44:58):

<Affirmative>. That's a good point because older adults sometimes want to get a rid of a piece of furniture, a couch, a sofa, and with the ability of iPhone and Android phone cameras to take a picture that can easily be done.

Dan Wilson (45:25):

Sure. Trying to find the right part to fix something. Taking a picture of, you know, a sink and asking why it might be leaking. Like these are things that aren't obvious because technology has not been there in that way for us before, but it is now. And I just get really excited by once again, like, I think people wanna live their life and for so long technology's been a barrier. And for the first time, it's now here to help

Larry (45:57):

Dan. I am sure we could go on for hours and hours more, but like everything good, it's got to come to an end. <Laugh>, I really appreciate you coming on the podcast. Thanks for being a guest.

Dan Wilson (46:15):

It has been my pleasure, sir. And, and I look forward to future episodes. So best wishes for the holidays. And

Larry (46:24):

Same to you,

Dan Wilson (46:25):

Sir. You're impressive,

Larry (46:26):

<Laugh>. Thank you.

Announcer (46:32):

If you found this podcast interesting, fun or helpful, tell your friends and family and click on the follow or subscribe button. We'll let you know when new episodes are available. You've been listening to specifically for seniors. We'll talk more next time. Stay connected.