WEBVTT - AI 101: A Beginner's Guide to Understanding Artificial Intelligence 0:00:01.240 --> 0:00:01.600 Earners. 0:00:01.639 --> 0:00:03.760 What's up. You ever walk into a small business and 0:00:03.800 --> 0:00:07.320 everything just works like The checkout is fast, the receipts 0:00:07.320 --> 0:00:10.840 are digital, tipping is a breeze, and you're out the 0:00:10.880 --> 0:00:15.000 door before the line even builds. Odds are they're using Square. 0:00:15.480 --> 0:00:18.079 We love supporting businesses that run on Square because it 0:00:18.160 --> 0:00:21.119 just feels seamless. Whether it's a local coffee shop, a 0:00:21.239 --> 0:00:23.639 vendor at a pop up market, or even one of 0:00:23.680 --> 0:00:26.799 our merch partners, Square makes it easy for them to 0:00:26.840 --> 0:00:30.680 take payments, manage inventory, and run their business with confidence, 0:00:31.040 --> 0:00:34.320 all from one simple system. If you're a business owner 0:00:34.520 --> 0:00:37.920 or even just thinking about launching something soon, Square is 0:00:38.040 --> 0:00:40.080 hands down one of the best tools out there to 0:00:40.120 --> 0:00:44.080 help you start, run, and grow. It's not just about payments, 0:00:44.479 --> 0:00:47.120 it's about giving you time back so you can focus 0:00:47.159 --> 0:00:49.879 on what matters most Ready. To see how Square can 0:00:49.880 --> 0:00:54.720 transform your business, visit Square dot com backslash go backslash 0:00:54.800 --> 0:00:58.960 eyl to learn more that Square dot com backslash, go 0:00:59.360 --> 0:01:03.560 backslash eyl. Don't wait, don't hesitate. Let's Square handle the 0:01:03.640 --> 0:01:07.360 back end so you can keep pushing your vision forward. 0:01:09.760 --> 0:01:12.000 This episode is brought to you by P and C Bank, 0:01:12.400 --> 0:01:15.160 a lot of people think podcasts about work are boring, 0:01:15.360 --> 0:01:19.200 and sure they definitely can be, but understanding of professionals 0:01:19.319 --> 0:01:22.559 routine shows us how they achieve their success little by little, 0:01:22.920 --> 0:01:26.600 day after day. It's like banking with P and C Bank. 0:01:27.040 --> 0:01:29.839 It might seem boring to save, plan and make calculated 0:01:29.880 --> 0:01:33.000 decisions with your bank, but keeping your money boring is 0:01:33.000 --> 0:01:36.360 what helps you live or more happily fulfilled life. P 0:01:36.440 --> 0:01:41.280 and C Bank Brilliantly Boring since eighteen sixty five. Brilliantly 0:01:41.319 --> 0:01:43.960 Boring since eighteen sixty five is a service mark of 0:01:44.000 --> 0:01:47.240 the PNC Financial Service Group, Inc. P and C Bank 0:01:47.680 --> 0:01:49.840 National Association Member FDIC. 0:01:51.680 --> 0:01:54.240 I want to start off by making and giving y'all 0:01:54.280 --> 0:01:57.560 my framework for how to think about what artificial intelligence is. 0:01:58.080 --> 0:02:02.040 So first, AI is not just one thing, right, just 0:02:02.080 --> 0:02:04.440 like when you think of the word car, car is 0:02:04.480 --> 0:02:07.120 not one thing. You can have a two door car, 0:02:07.240 --> 0:02:09.280 you could have a four door car, you could have 0:02:09.320 --> 0:02:11.840 a sports car, you could have a truck. And even 0:02:11.880 --> 0:02:14.560 though there are all these you know, different types of cars, 0:02:14.560 --> 0:02:17.560 like whether they're electric or whether they're gas, even though 0:02:17.560 --> 0:02:20.480 they literally end up different in the way that they look, 0:02:20.800 --> 0:02:23.440 all cars pretty much do the same thing. Which is 0:02:23.480 --> 0:02:25.840 that they get you from point A to point B 0:02:26.360 --> 0:02:28.839 faster than you would get there if you were walking right, 0:02:29.080 --> 0:02:33.760 unless you're in LA traffic for it card. Similarly, artificial 0:02:33.800 --> 0:02:37.119 intelligence covers a whole bunch of different tools and technology 0:02:37.120 --> 0:02:40.400 across different disciplines. It touches tools that are being built 0:02:40.440 --> 0:02:45.160 in computer science and mathematics and electrical engineering and mechanical engineering. 0:02:45.560 --> 0:02:48.959 But all AI aims to do pretty much the same thing. 0:02:49.600 --> 0:02:53.959 Artificial intelligence is trying to teach machines how to think 0:02:54.360 --> 0:02:57.960 and act like humans. So all those different technologies and 0:02:58.040 --> 0:03:01.120 all those different disciplines are just trying to teach machines 0:03:01.480 --> 0:03:04.400 how to think and act like humans. Well what does 0:03:04.440 --> 0:03:06.760 it mean to think and act like a human? Well, 0:03:06.919 --> 0:03:09.880 humans have five senses. We have hearing, we have sight, 0:03:10.200 --> 0:03:12.640 we have touched, we have smell, and we have taste, 0:03:13.000 --> 0:03:17.600 and artificial intelligence kind of has five domains. So teaching 0:03:17.600 --> 0:03:21.120 an AI, teaching a machine how to hear and how 0:03:21.120 --> 0:03:24.919 to speak is called natural language processing. That's what's behind 0:03:24.960 --> 0:03:27.959 tools like Siri inside of your phone. When you talk 0:03:28.000 --> 0:03:30.000 to her and she writes the text down, that's called 0:03:30.000 --> 0:03:33.080 text to speech. Then when she understands what you just said, 0:03:33.080 --> 0:03:36.080 to her. That's called natural language understanding. Then when she 0:03:36.160 --> 0:03:38.600 comes up with a response to you, that's called natural 0:03:38.680 --> 0:03:41.760 language generation. And then when she reads that response back 0:03:41.800 --> 0:03:45.080 out loud to you, that's called text to speech right 0:03:45.120 --> 0:03:47.880 where she reads it back out loud to you. So NLP, 0:03:48.080 --> 0:03:51.120 or natural language processing, is teaching machines how to hear 0:03:51.400 --> 0:03:54.400 and how to speak. Teaching machines how to see is 0:03:54.440 --> 0:03:57.080 called computer vision. So if you have a ring camera 0:03:57.240 --> 0:03:59.720 in your house and it's like, oh, person detected her, 0:03:59.760 --> 0:04:03.080 and well detect it, that's using computer vision. That's also 0:04:03.200 --> 0:04:06.280 behind stuff like face ID inside of your phone, when 0:04:06.320 --> 0:04:08.800 you go to the airport and TSA asks you to 0:04:08.840 --> 0:04:11.000 stand there and it checks the picture of your face. 0:04:11.480 --> 0:04:15.080 All of that stuff is computer vision. Teaching a machine 0:04:15.080 --> 0:04:18.080 how to touch is called haptics. Now, this is a 0:04:18.120 --> 0:04:21.200 field that you mostly see in robotics, where they're doing 0:04:21.279 --> 0:04:25.960 things like creating artificial skin that's as sensitive as human skin. 0:04:26.240 --> 0:04:28.160 Now why would they need that? Well, think about it. 0:04:28.200 --> 0:04:30.760 If you have like a robot and an Amazon factory, 0:04:31.120 --> 0:04:33.240 you want it to be able to understand things like 0:04:33.360 --> 0:04:36.160 pressure and weight so that when it picks up whatever 0:04:36.160 --> 0:04:38.920 you ordered off an Amazon that might be fragile. It 0:04:39.000 --> 0:04:42.160 doesn't pick it up and shake it around and mishandle it, right, 0:04:42.200 --> 0:04:44.839 So they need that artificial skin to be able to 0:04:44.839 --> 0:04:47.920 teach robots how to understand what it is that they're 0:04:47.920 --> 0:04:51.599 physically touching and grabbing. Then teaching machines how to smell 0:04:51.800 --> 0:04:55.200 is a pretty new field. It's called machine oh factory. Now, 0:04:55.240 --> 0:04:57.560 I'm not gonna lie, I don't really like this field 0:04:57.640 --> 0:05:00.680 because in order to teach the algorithm and what a 0:05:00.800 --> 0:05:03.360 smell is, you got to use like somebody's nose as 0:05:03.360 --> 0:05:05.880 an example. And I don't know about y'all, but some 0:05:05.920 --> 0:05:09.080 people out of the colognes they be picking the perfumings. 0:05:09.120 --> 0:05:11.880 They be picking like I don't trust everybody's nose, right, 0:05:12.240 --> 0:05:15.760 And so MACHINEO factory is a newer field. We've seen 0:05:15.839 --> 0:05:19.000 like maybe ten twenty research papers in this space altogether, 0:05:19.240 --> 0:05:20.920 where they're trying to do stuff like come up with 0:05:20.960 --> 0:05:24.839 new perfumed smells or like detect smoke from a smell, 0:05:24.920 --> 0:05:28.360 et cetera. And then teaching machines how to taste doesn't 0:05:28.400 --> 0:05:30.640 actually have a fancy name, but it has the same 0:05:30.760 --> 0:05:33.520 problem of kind of like teaching a machine how to smell. 0:05:33.600 --> 0:05:36.320 Like I'm not eating at everybody's house, So whose taste 0:05:36.320 --> 0:05:39.559 buds are you using as the example to like teach 0:05:39.600 --> 0:05:40.840 these machines how to taste you? 0:05:40.839 --> 0:05:44.360 Feel me definitely got you there, So that the five 0:05:44.480 --> 0:05:46.880 senses in the five domains makes a lot of sense 0:05:47.200 --> 0:05:51.040 when people start hearing these things like computer vision, Definitely 0:05:51.080 --> 0:05:54.360 that the smell and the taste, it is subjective. It 0:05:54.360 --> 0:05:57.159 starts to lead to uncertainty. And one of the things 0:05:57.200 --> 0:06:00.360 with human nature is that we don't like uncertainty, which 0:06:00.440 --> 0:06:04.479 then leads to fear. So how can people become less 0:06:04.520 --> 0:06:08.120 fearful when they hear about AI and very broken down? 0:06:08.120 --> 0:06:08.840 How you just did it? 0:06:09.760 --> 0:06:11.640 So I feel like a lot of the fear is 0:06:11.680 --> 0:06:14.800 based on the unknown. Like the human brain generally is 0:06:14.960 --> 0:06:18.800 wired to protect you from danger, right, Like the amygdala, 0:06:18.880 --> 0:06:20.960 which we call the fight or flight part of the brain, 0:06:21.560 --> 0:06:24.200 is literally the oldest part of the brain, and it's 0:06:24.240 --> 0:06:26.760 the only part of the brain that can hijack the rest. 0:06:27.040 --> 0:06:29.039 So if you ever had like a panic attack, that's 0:06:29.040 --> 0:06:32.400 literally your amygdala hijacking in your brain and shutting down 0:06:32.480 --> 0:06:34.800 all the other different functions in your brain because it 0:06:34.839 --> 0:06:38.000 literally thinks that you're dying. And so a huge part 0:06:38.040 --> 0:06:41.240 of overcoming fear to AI is learning about it, doing 0:06:41.320 --> 0:06:45.359 things like sitting in this class, taking notes, understanding what 0:06:45.480 --> 0:06:48.839 the difference is between NLP and computer vision, so that 0:06:48.960 --> 0:06:53.360 you're able to not feel like because you don't understand it, 0:06:53.680 --> 0:06:56.200 that it instantly means that it's going to be negative 0:06:56.240 --> 0:06:59.240 against you. Now, there are real challenges with AI, with 0:06:59.440 --> 0:07:02.359 which will probably get into later, but even then, the 0:07:02.400 --> 0:07:05.520 problems inside of these AI systems, with things like bias 0:07:06.000 --> 0:07:09.960 or things like them coming for you know, jobs that 0:07:10.040 --> 0:07:12.960 particularly affect our communities. A lot of that stuff can 0:07:13.000 --> 0:07:15.520 be solved if we step up in other areas, like 0:07:15.600 --> 0:07:18.120 in policy, or if we step up in terms of 0:07:18.160 --> 0:07:20.480 like regulation, or if we step up in terms of 0:07:20.520 --> 0:07:23.760 building and designing our own systems, and so even the 0:07:23.800 --> 0:07:26.520 stuff out there that is actually proven to be scary 0:07:26.600 --> 0:07:29.800 or dangerous to our communities, there are solutions out there 0:07:29.800 --> 0:07:32.880 that require all of us coming together and putting into play. 0:07:33.960 --> 0:07:38.119 Okay, all right, so I know we have the next step, 0:07:38.160 --> 0:07:40.560 which we're going to talk about how to utilize it 0:07:40.560 --> 0:07:42.440 in business. Anything you need you want to talk about 0:07:42.440 --> 0:07:44.480 before that before we go to the business lide. 0:07:44.880 --> 0:07:47.160 Yeah, So before we jump into the business side, I 0:07:47.200 --> 0:07:49.960 want to give a little bit more context about you know, 0:07:50.360 --> 0:07:52.960 how you teach a machine to think at like a human. 0:07:53.040 --> 0:07:56.400 So we know that AI is this group of technologies 0:07:56.400 --> 0:08:00.360 that are trying to do that using this five Census example. Now, 0:08:00.680 --> 0:08:05.960 generative AI specifically is attempting to teach AI teach machines 0:08:06.040 --> 0:08:08.720 how to think and act like the human imagination. Like, 0:08:08.760 --> 0:08:11.040 if I ask everybody on a chat right now to 0:08:11.120 --> 0:08:14.120 imagine a car, you would come up with an image 0:08:14.120 --> 0:08:15.720 in your head of a car. Some of you guys 0:08:15.720 --> 0:08:18.080 would imagine a two door car, a four door car, 0:08:18.160 --> 0:08:20.560 or a sports car. Some of y'all might even imagine 0:08:20.600 --> 0:08:23.120 a semi truck. Now, what I did is I just 0:08:23.240 --> 0:08:26.080 prompted you to create something from the prompt that I 0:08:26.160 --> 0:08:29.320 gave you. You're using your human imagination to create something 0:08:29.360 --> 0:08:32.200 from an idea of it. And so generative AI is 0:08:32.280 --> 0:08:34.640 teaching machines how to think and act more along the 0:08:34.679 --> 0:08:37.160 lines of a human imagination. And I'm gonna get a 0:08:37.200 --> 0:08:40.400 little bit technical just so you guys understand some of 0:08:40.440 --> 0:08:44.120 the differences between the techniques used to build AI and 0:08:44.240 --> 0:08:47.679 AI itself. So, how do you teach a machine how 0:08:47.720 --> 0:08:50.000 to think and act like a human? What else where 0:08:50.000 --> 0:08:52.720 the term machine learning comes in. You've probably heard this 0:08:52.840 --> 0:08:56.280 a lot. Machine learning is the most popular way that 0:08:56.320 --> 0:08:58.880 people teach machines how to think and act like humans, 0:08:58.880 --> 0:09:01.240 but it's not the only way. Are the only method, 0:09:01.520 --> 0:09:03.920 and machine learning is when you teach a machine, which 0:09:03.920 --> 0:09:07.000 they call training, how to complete a task by giving 0:09:07.040 --> 0:09:10.240 it examples which they call training data of that task 0:09:10.520 --> 0:09:13.360 so that it can learn and improve itself on how 0:09:13.400 --> 0:09:16.080 to do that task right. So, for example, if I 0:09:16.120 --> 0:09:19.160 wanted to teach a car how to drive on the road, 0:09:19.240 --> 0:09:22.040 if it was old school coding without machine learning, I'd 0:09:22.120 --> 0:09:25.040 have to code every single step that the car needed 0:09:25.360 --> 0:09:28.920 to do well. A lot of those rules are you know, 0:09:29.000 --> 0:09:31.120 there's like millions of rules of driving, like to have 0:09:31.240 --> 0:09:34.560 the code like if your right tire gets within this 0:09:34.720 --> 0:09:37.600 many inches of the line, then move this steering will 0:09:37.679 --> 0:09:40.560 you know point two centimeters. That's a lot to learn 0:09:40.760 --> 0:09:42.880 and a lot to code. We probably wouldn't even be 0:09:42.920 --> 0:09:44.920 able to make an app that would be able to 0:09:45.000 --> 0:09:47.160 let the car drive like that. It was just take 0:09:47.200 --> 0:09:50.200 too much code. So instead, with machine learning, we can 0:09:50.240 --> 0:09:53.960 take a bunch of examples of cars driving on the road. 0:09:54.000 --> 0:09:57.080 That's why like Tesla's have eight cameras. They're constantly capturing 0:09:57.120 --> 0:10:00.679 that data and then let it learn from those examples 0:10:00.720 --> 0:10:02.840 of things that it should do and things that it 0:10:02.840 --> 0:10:05.600 shouldn't do. So there are two types of machine learning. 0:10:05.679 --> 0:10:09.520 The first is called supervised machine learning, and supervised machine 0:10:09.600 --> 0:10:11.640 learning is when you teach a machine how to complete 0:10:11.679 --> 0:10:15.079 a task by giving it examples of the right answer 0:10:15.400 --> 0:10:18.000 up front. So an example of this is like, let's 0:10:18.040 --> 0:10:21.800 say I wanted to teach a computer vision algorithm to 0:10:22.000 --> 0:10:26.400 be able to tell the difference between a strawberry and avocado. Right, 0:10:27.160 --> 0:10:29.679 I would give it examples of images that I knew 0:10:29.760 --> 0:10:32.600 upfront were a strawberry. I would tell it upfront that 0:10:32.640 --> 0:10:35.400 it's a strawberry. We call that data labeling, and then 0:10:35.440 --> 0:10:38.360 I would check it as it's learning how good it 0:10:38.400 --> 0:10:41.680 gets at telling me that's a strawberry versus telling me 0:10:41.760 --> 0:10:45.600 that's an avocado. Right. So, again, supervised machine learning is 0:10:45.679 --> 0:10:48.000 I already know what the right answer is up front, 0:10:48.200 --> 0:10:50.280 so I'm gonna give you these examples, and then I'm 0:10:50.320 --> 0:10:52.680 gonna test you against how well you're learning how to 0:10:52.720 --> 0:10:54.960 do the task because I already know the right answer. 0:10:55.600 --> 0:10:59.280 The second type of AI machine learning is called unsupervised 0:10:59.360 --> 0:11:02.199 machine learning, And this is when you're training a machine 0:11:02.240 --> 0:11:05.360 how to complete a task by getting given it examples, 0:11:05.600 --> 0:11:08.880 but you're letting the algorithmic self discover what the right 0:11:08.920 --> 0:11:11.560 answer is. Right, So you're giving it a whole bunch 0:11:11.640 --> 0:11:13.920 of data, and then it goes in and finds patterns 0:11:13.920 --> 0:11:16.959 across that data that it uses to give you the 0:11:17.040 --> 0:11:20.280 right answer. So, for example, a lot of AI is 0:11:20.320 --> 0:11:23.679 being used in things like drug discovery, So how can 0:11:23.679 --> 0:11:25.680 we find a new medicine to treat AIDS or to 0:11:25.720 --> 0:11:28.720 treat diabetes, or even when COVID came out, there was 0:11:28.880 --> 0:11:31.400 a whole effort between a bunch of big tech companies 0:11:31.440 --> 0:11:35.240 to figure out medicines that could be used as a vaccine. Right, 0:11:35.600 --> 0:11:38.000 So in that case, they didn't know the right answer 0:11:38.080 --> 0:11:40.240 up front, but they had a whole bunch of data 0:11:40.280 --> 0:11:44.040 and then the machine uncovered patterns that help them understand 0:11:44.040 --> 0:11:47.000 what the right answer should be. Right, all right. So 0:11:47.080 --> 0:11:49.560 then you have one more thing, which is called deep learning. 0:11:49.960 --> 0:11:53.360 So deep learning is basically, without going into all the 0:11:53.400 --> 0:11:58.040 math behind it, it's a more complex way of doing 0:11:58.080 --> 0:12:00.760 machine learning when you have things that are more complicated 0:12:00.760 --> 0:12:03.840 in their patterns, and so it teaches the algorithms in 0:12:03.880 --> 0:12:07.520 a way that mimics how neurons fire in the human brain. Right, 0:12:07.840 --> 0:12:09.800 So machine learning, like if I want to tell the 0:12:09.800 --> 0:12:12.800 difference between the strawberry and avocado, that's not a very 0:12:12.840 --> 0:12:16.640 complex task. But like teaching a plane how to fly itself, 0:12:16.760 --> 0:12:19.439 like the US Air Force just did, is gonna require 0:12:19.480 --> 0:12:22.720 a lot more complex relationships and a lot more complex 0:12:22.880 --> 0:12:26.280 understanding and nuance inside of that data and the patterns 0:12:26.320 --> 0:12:28.400 and how it's supposed to do that task. So they 0:12:28.400 --> 0:12:31.960 would use something like a deep learning algorithm instead to 0:12:32.040 --> 0:12:34.079 be able to capture that. Now, I know that was 0:12:34.120 --> 0:12:36.320 a lot of information. I don't expect y'all to have 0:12:36.400 --> 0:12:38.720 this memorized. So while yeah, before we move on to 0:12:38.760 --> 0:12:41.800 the next part, I made a little TOLDR slide for y'all. 0:12:41.960 --> 0:12:44.160 You know, take a moment, take a screenshot, Make sure 0:12:44.160 --> 0:12:46.520 to share this with your friends. It just recaps what 0:12:46.640 --> 0:12:47.360 I just covered. 0:12:47.440 --> 0:12:51.120 An illegal alien from Guatemala charged with raping a child 0:12:51.120 --> 0:12:54.959 in Massachusetts. An MS thirteen gang member from El Salvador 0:12:55.160 --> 0:12:59.320 accused of murdering a Texas man of Venezuelan charged with 0:12:59.360 --> 0:13:03.280 filming and selling child pornography in Michigan. These are just 0:13:03.360 --> 0:13:07.120 some of the heinous migrant criminals caught because of President 0:13:07.160 --> 0:13:07.600 Donald J. 0:13:07.679 --> 0:13:08.640 Trump's leadership. 0:13:08.920 --> 0:13:10.080 I'm Christy nom the. 0:13:10.160 --> 0:13:14.720 United States Secretary of Homeland Security. Under President Trump, attempted 0:13:14.800 --> 0:13:18.480 illegal border crossings are at the lowest levels ever recorded, 0:13:18.800 --> 0:13:21.840 and over one hundred thousand illegal aliens have been arrested. 0:13:22.120 --> 0:13:25.680 If you are here illegally, your next you will be 0:13:25.720 --> 0:13:29.600 fine nearly one thousand dollars a day, imprisoned, and deported, 0:13:30.080 --> 0:13:33.319 you will never return. But if you register using our 0:13:33.320 --> 0:13:36.520 CBP home app and leave now, you could be allowed 0:13:36.559 --> 0:13:37.559 to return legally. 0:13:37.920 --> 0:13:40.199 Do what's right. Leave now. 0:13:40.480 --> 0:13:44.840 Under President Trump, America's laws, border and families. 0:13:44.440 --> 0:13:46.839 Will be protected. Sponsored by the United States Department of 0:13:46.880 --> 0:13:47.480 Homeland Security.