Because it's mostly a how-to story, you really need to be at your own computer and try to have a look at how to do things. So try not to read all the verbiage. I'll try to cover it in my talk. So anyway, everybody wants to know what it is and what you do with it.
So basically, the power and potential and limitations of a large language model is a generative AI assistant. What it does is it looks at a huge database of information that it sweeps up off of the web, other parts of the internet, company databases, whatever you tell it to access, and it looks for patterns, and it puts words together one at a time based on the patterns it sees in the data. That's about as much as I know about what makes it work. In fact, most of the AI people don't know exactly what goes on in the black box.
So, I'm just going to tell you what we're going to do. We're going to have a live demonstration. I'm going to give you a guide to how to choose, how to prompt, and use the AI bots to your advantage. So, as I said, lots of information. Just listen up and if you will bear with me. There's a lot of information on some of these slides. Better to look at individually as you're trying to experiment a little bit with the information. And I'll come back to any slide as we need. So please hold your questions until we get to the end. There'll be a time to ask questions then.
So, it's more than just talk. People think, 'Oh, it's going to be my friend.' And it has a very friendly interface. It wants to please you. But it's got a powerful background of information that can give you a lot of insight into the things you're wondering about or want to search for, etc. You can use it either by typing in questions which are called prompts, or you can use a voice connection to speak to it and most of them will speak back to you.
You can also do things like graphics. It can generate a picture of something for you. You can use different voices. My son-in-law and I were playing around with it a few minutes ago and he said, 'Well, I'm interested in Irish pub. Speak to me in an Irish brogue.' Bang, there she was, a lassie speaking back to you. So, you can use it to do things like help you with spreadsheets, even do computer coding if anybody here is into that sort of thing. Searches are made much quicker and to the point when you use AI.
You know what happens when you go to Google? You get line after line of connections to other websites. You click on them and it turns out that website doesn't really have what you were interested in. With AI, as I'll show you in a demonstration, you can just tell it what you're looking for and it'll do the searching for you and give you the precise answer of what it is you are looking for. No more endless Google clicking.
It's a new way to interact and give you a creative engine to generate content that is new and unique to you. It helps you to stimulate your thought, express yourself, and you can have some fun with it like the lady with the Irish brogue we brought up.
So, a little something about the potential of AI. Everybody's seen lots of stuff in the press about it. If you haven't, you're not paying attention because it's all over everything you hear and see on the media today. I won't go into the statistics, but even though a lot of people have looked at it, only 5% according to Pew have actually used it in any depth so far. The World Economic Forum says that AI will displace 9 million jobs by 2030, but it says it'll create 11 million in place of those jobs. So, certain people are going to have to change and adjust to what they do or what their jobs are. But there will be plenty of work for everyone.
It's important because what? Well, it is the next big thing. Wasn't that many years ago that the internet was the next big thing. Look where it's gone. Some good and some bad. Lots of billions of dollars are being invested in AI because it creates, it needs a lot of power and a lot of computing heavyweight investment. It really works. You'll see it's pretty simple to use. Can help you with ordinary everyday problems. You can learn and improve your creativity using it. And it's free to use and readily available. I say free. Everyone of the ones that I know of have a free version. If you want to pay 20 bucks a month, you can get the upgraded version. I've never done that because I'm not into kind of some sort of in-depth research, for instance, that a doctor might do or a scientist or someone that's really into researching subjects.
Excuse me. I pushed the wrong button. Yeah, I think so. Here's a few examples of what bots can do for you. You can create something in a let's say a text of something. I've used it to create a eulogy for a dear friend, something I never did before. I'll tell you a little bit how I did that. I've used it to do a wedding speech. You can do an article. It can even give you spoken simultaneous translation of a foreign language. You can speak to your friend in France on say Zoom in English, she will hear your question in French. She will respond in French and you can hear her answer immediately in English. Pretty amazing.
You can learn more about a subject. Going on vacation. You want to know something special about a national park and whether you can get around with your walker or not? What's to see there? Your doctor says you've got a new ailment. You want to know something about it more than perhaps he was able to tell you in the 15 minutes that he's allotted to talk to you. You can do some shopping. A new car perhaps, clothing that matches your style or your coloration, whatever your complexion, or you can solve a problem you run into on your computer or your phone.
So, let's get right into it. I'm going to do a little live demonstration. Here I'm going to use a voice prompt using Microsoft Copilot, which any of you that use Microsoft Word or Excel have seen the little popups. They want you to try it. So, give me a second here to pull up Copilot. There she is. So, it says, 'Good afternoon, Gary.' But that's in text. I want to speak to it, and I want to ask it a question, a prompt. Let me get my prompt. So I say it correctly.
Hello Copilot. Are you listening loud and clear? Copilot, I have something for you. We are residents of a retirement community and want to learn more about AI and how it can be used to enrich our lives. List five ideas for our next meeting. Use a jovial voice.
All right. Well, now you got to understand that I'm not very good at this. First of all, you understand I can get everything except what I want on my screen. But anyway, there we go. Understand that was generated on the fly. There's nothing pre-programmed. I've asked it that question before and it's given me somewhat different answers before. In fact, if you ask that same question of two or three bots, you'll get two or three somewhat different answers. And they're all generated over the Ethernet per your specific question.
Let's try another one. So, I'm going to use text this time. So, I'm going to copy. Excuse me. Well, it won't work in this mode. Let me do it another way. I should be able to... Excuse me. Well, now... All right. It's going to cause me trouble, I can see. Here I go. Maybe I can do it this way. Yeah, see Copilot keeps popping up when you do a Microsoft program. So, here's the question I want to ask. I'm going to copy that. And then I'm going to paste it into... I'm going to paste it into something called Perplexity. Perplexity. So, it's saying, 'Ask me something.' There's my prompt. I want to buy a pair of brown men's shoes that are comfortable to wear, casual but smart looking, have rubber soles and leather tops, and cost no more than $150. Give me five choices stating brand, price, and links to the vendors.
Now for reasons unknown to me at the moment it's showing you the slides and not so I am going to cut the slide. Oh, come on. I knew this was going to happen. All right. Here is Perplexity's answer, which it came up with about two minutes ago, but I wasn't able to get the screen up so you could see it. So, it's giving me... There's number one, $104. You can buy it on this website. I click that, I would go straight to where I could buy that shoe. It tells me a little bit about it. It says, 'Okay, if you don't like Cole Haan, you can go to Clarks.' Remember I said rubber soles, leather brown tops, and shoelaces. Here's another one from Banana Republic. Another one from ECCO. Another one from Stacy Adams. And I can click on these items here and get more information, related information.
I'm going to see if I can ask a follow-up here by voice. I've changed my mind. I want to have black shoes without laces. Give me five of those, please. Well, try again. I've changed my mind. I want black shoes with no laces. Give me five of those, please. They want specifics. I have changed my mind. I am getting very frustrated with you. Just type. I have changed my mind. Can you hear me? Well, I'm sorry, folks. The fact is if you don't like the first answer you get, ask it another question and you don't have to repeat all about the rubber soles and everything else. It remembers, it knows what you're talking about and it gives you an alternative answer without all that rigmarole.
So, let's go back to the slides and the slideshow and we'll get to where we were. So, I don't know if you're impressed or not. It impresses me that can you imagine looking for that many brown shoes with rubber soles and laces on Google and how long it would take you to get somewhere where you could actually buy what you wanted for the price you wanted. You get the idea. So I'd like to get into a little bit of ideas of how you use AI. And we're going to cover several things. I'm going to skip over here. This is where I tell you what I'm going to tell you. So, I'm going to skip that.
So, they say different bots have different personalities. Well, they have different personalities or voice. But listen, they're tools. There's no sentient human being sitting somewhere in Baltimore giving you these answers. They are, it's not real folks. It's a computer simulation. But nevertheless, some excel at creativity, others at precision or safety. They have strengths and weaknesses. You want one that you can find that meets your specific needs. One that's easy to use. You can see I was having some trouble with that. And my recommendation is to try different ones. In fact, ask them the same question.
When I was putting together that eulogy, I asked two or three different bots to write something for me. I gave them who he was and so forth, a bit of facts about his history. I got those three or four responses. Parts of some I liked, parts of others I didn't like. And some of it didn't sound like me. So I did a little cutting and pasting on Word to come up with what I thought was what I wanted to say. But it helped me collect my thoughts and it helped me phrase things that might sound better in the particular circumstance. And it certainly gave me sympathetic answers, answers that were suitable to something like a eulogy. So that's what I would suggest. Explore different ones and compare the results.
I'm going to skip this slide. It'll be in the deck if you want to see it. It just gives you some of the attributes of the personalities of different bots. But there's a wider world about AI. They're sometimes called agentic AI, agents. It's things that a company or a nonprofit or anyone might use to accomplish certain tasks in an automated fashion. We've all listened to the guy in Hindustan explaining what our problem was on Microsoft and about 20 minutes into that conversation, we gave up because we couldn't understand half of what he was saying. If you ask a bot, which will soon be prevalent on most customer interfaces that you call, you will get an answer that's researched throughout the entire web. So hopefully it'll be correct, but it'll also be spoken to you or typed to you in plain old American English.
If in a company you need to handle some employee PR items or some other departmental task, AI assistants are starting to help companies automate that sort of thing. Bots can also produce images or translate your voice or write computer codes. There are medical bots that are in use and certainly being developed that can help diagnose illnesses or design new drugs. These should be used by professionals, not you and I going on and trying to diagnose our own illness. But they've been shown to be able to pick up lung cancer and other items that the human doctor looking at that X-ray has not been able to detect. And they can go through design of new drugs at internet speeds and try many different combinations before they find one that works. So, the pharmaceutical companies are starting to use them.
If you want to go on a trip, pretty soon you'll be able to talk to a bot, tell it where you want to go, when you want to go, how you want to go, and it'll set up an itinerary for you. And if you tell it to, it'll make the bookings for you. Maybe that's still coming along, but it's going to be here very quickly. You use most of them for free. Some of them 20 bucks a month roughly for the enhanced versions, which I don't think any of us really need. I certainly don't use them. So, experiment and find the ones that fit your needs.
Now, everybody says, 'Well, I know this thing gave me a wrong answer. It was what they call in the trade hallucinating or fabricating information.' And that has happened and will happen. They are constantly learning and improving, but you got to be careful. They sound very confident. You heard some of what we did in demonstration there. Gee, they know you. They know exactly what you want. They give you an answer. Bang. Don't take any time at all. But please, do we need that on? Somebody turn the lights out. Are you done with the slides or you have more slides? Yes, I have more. Okay. Is everybody getting tired? I can cut this short anytime anybody's thinking that it's getting too long. We're about 30, 40 minutes in. So don't believe everything you see or hear from a bot. Critical thinking and verification are still something you need to do. Trust but verify. And you can get the bot, if you ask it the right question in the right way, to give you citations for its information. And you can find out whether in fact what it's telling you matches what the citations are.
So it's all about how you prompt. You got to give specific instructions or questions. It's the old GIGO, garbage in, garbage out. You'll get back from a prompt exactly the quality that you put into it. So, I'm going to cover some of the ways that you can do that. You got to provide enough detail. Don't be vague. Let it understand what the background is, what the context of what you're asking is. Like I did with the group here of retired people who want to know something about, give it some background, give it some information. You can have it reply to you in a voice, I think I said musical jovial voice, in other words friendly, nice, easygoing. But you can say things like explain it like I'm a fifth grader, which sometimes we have to do, or write in a certain style. But dream up something that doesn't exist if you want existing original, excuse me, stories or poems. It writes very nice poems, by the way, and jokes. So tailor the responses to your requirements by giving it enough information.
Give it some structure. I want a list. I want a bullet list. I want an outline of a PowerPoint. Limit the number of words or paragraphs. Avoid unrelated questions. In other words, don't mix it up in the same prompt. And you can hit the deep think button if you want it to give you a more nuanced detailed answer than the quick answer it gives you. It may take a few seconds more, but it may be worth it in certain circumstances.
So, if you're not sure of what the bot's telling you or just as a good practice anyway, ask it to explain its conclusions or elaborate or just rewrite the whole thing altogether. If you don't like what you heard the first time, it knows the context you're talking in and it'll come right back with a rewritten section. You can do the deep research button. You can ask it to cite sources. You can challenge it like I did about, I tried to do and didn't, the black shoes and the no laces. It remembers where you were in the conversation so you don't have to go back and repeat a bunch of stuff. Usually a few seconds, maybe a half hour if you're asking a really detailed research question. Regardless of how polished the answer seems, always verify. It's a simulation, not a sentient human sitting somewhere.
So the best way to learn is to experiment and have fun exploring what AI can do. Different bots give you slightly different answers, sometimes radically different answers. You can then decide for yourself what is relevant. It can help you enhance your productivity or your creativity. The other day I had a problem on a spreadsheet. I wanted it to do something. I had no idea what the formula was to do that. I asked the simple question of it. How do I get this out of this data set? And in a second, it gave me the answer, including the actual Excel formula to plug into my spreadsheet with the columns and the lines that I needed to cite and bang, I got the answer I wanted. I could have spent hours trying to figure out the formula for myself.
However, be careful about personal information. Especially medical and financial information. Some of these bots, probably most of them, store your information somewhere for some period of time. So, be careful what you say. You shouldn't try to diagnose an illness or get financial advice without help from a professional. Be aware that the platform data usage and storage policies, what they are, and whether you want to give it any particular information and always fact check. It's a tool, not an authoritative source.
So, I thank you for listening. What we talked about is that prompts unlock the potential. You need to pay attention to how you ask the question and give the bot some background information context. They're information troves, but they're not your buddy, even though they may sound like it because they want to please you. They want to tell you things that are going to make you happy. That's the way they're designed. Try different ones. Remember, garbage in, garbage out. Do not be afraid to challenge the bot with follow-up questions and always trust but verify. Thank you. Any questions? I see one back here.
