# STATURDAY CHAT LIVE with Brandon Foltz, Ep. 63

## Метаданные

- **Канал:** Brandon Foltz
- **YouTube:** https://www.youtube.com/watch?v=pMxPw1_uQhM
- **Дата:** 18.02.2024
- **Длительность:** 1:02:42
- **Просмотры:** 523
- **Источник:** https://ekstraktznaniy.ru/video/52870

## Описание

Very informal, casual chat and hang out! If you have a question, you can submit it by messaging me here or on my website. Stop by for a bit as I talk about things I am working on, answer questions, and have you tell me how I can serve you better. And, of course, I want to check in on everyone and see how you are doing. See you... STATURDAY CHAT LIVE!

#statistics #datascience #machinelearning #livestream #qa

## Транскрипт

### Segment 1 (00:00 - 05:00) []

a hello everybody and welcome to the next edition of Saturday chat live here on February 17th 2024 it is a snowy day where I am at here in Ohio but it is gorgeous outside cold but gorgeous I hope you are well so um I'm going to do a few Channel updates and then kind of open it up to whatever you want to talk about and we'll go from there so first if you are new to the stream this is the way this works I usually go 30 minutes to an hour depending on how good the conversation is and what it is you want to talk about and we kind of see where it goes I don't run it long just to keep it going so if it we run out of stuff talk about I'll just go ahead and stop that's kind of the way it works I also love to know where you all are watching from it lets me know sort of where in the world you are obviously watching from and how far the reach of the channel is and all that kind of cool stuff and it gives me ideas on places to go someday so I'm planning my next trip probably um Portugal or Spain we'll see but I always look forward to like going places that you all recommend or where you're from and you know exploring the world that way so you let me know where you're uh watching from and please say at least hello in the chat and I will do my best to respond to whatever questions or comments you may have so some Channel updates and we'll get into some news and whatever else you want to talk about so please uh fill up that chat so we have stuff to uh to discuss some Channel updates so you've probably noticed that I have not been uh live or have new videos for a few weeks or so and that's because unfortunately my grandmother passed away on January 24th so she passed away on at the end of January uh which of course is um sad and traumatic and everything wasn't unexpected she had been ill for a while but obviously it takes a toll on our family and we had her services and funeral and stuff in the time after that so the reason I've been gone is largely dealing with my grandmother's passing and I'm trying to get back into the groove of everything you know to get making some new stuff the next thing I released on the channel is going to be a little bit different so before my grandmother passed away I had written it's basically a video essay honestly it didn't start out like that but the more I wrote it and ended up being a video essay on different aspects of AI and machine learning that how I see it impacting different things and as I got into it I just kept having ideas and it basically turned into a video essay so I will need to record it's already written I just need to record it and edit it and stuff like that and then get back into sort of the more nuts and bolts of the playlist I'm working on with machine learning and stuff so I have written new content I do have new content planned but life sometimes gets in the way and you just have to adapt and adjust and that's what I've sort of been doing so I never you know a lot of people were quitting i't say but you probably noticed that a lot of big YouTubers have been

### Segment 2 (05:00 - 10:00) [5:00]

quitting YouTube recently and I completely understand sort of where they're coming from you know as those of you on YouTube have probably noticed YouTube is changing and that's because YouTube isn't really going to be YouTube anymore YouTube is trying to be a streaming service so it can compete with other streaming services therefore you know the algorithm changes to behave more like a traditional streaming service than YouTube's original philosophy of being YouTube you know just just YouTube now of course that's just their business model right you know they are a corporation that is um that answers to their shareholders and they need to do what to increase revenues and ultimately profits and the way they see doing that is creating a streaming service out of YouTube instead of just a fun video site where people can upload goofy videos or whatever else um it's also the case that the production value in videos is easier now because of the tools that are available so it can do that you know there are many shows and things on YouTube that I watch that are just as good or not better than things I watch on traditional streaming services and maybe you would agree with that so YouTube is changing um I say that because I'm not quitting YouTube however it's the lowest on my priority list because YouTube no longer like for me the channel was sort of flat flattened out I do get new subscribers but not nearly as many as I used to and um I don't hardly make any money on it so I have to do the things that actually pay the bills first and the reality is that's not YouTube because the way it's changed and you know i' have a career that's my first priority obviously I have another side job that actually pays me a certain amount per hour to do work that's guaranteed and YouTube I might make some money I might not and it's a lot of work you know so I'm not quitting YouTube or anything like that but it is sort of the lowest priority in terms of the time just because the juice isn't worth the squeeze like it used to be but I'm not quitting or anything but that's just the nature of the Beast so I will have new content out but it comes out whenever I have time to do it write it edit it and so on and so forth that's just the reality so look out for this so keep looking for new content as it's released um but it's a bit irregular because I do it when I have time you know to do it but I will keep doing it so goes hi how you doing it's good to see you thanks for stopping by appre appreciate you saying hello in the chat very kind of you yeah please uh you know say hello in the chat let me know where you're watching from and put any comments or questions or things you want to talk about related to statistics machine learning um AI whatever if I can't answer it I will tell you if I don't have the knowledge I'll tell you but sometimes I can get back to you later either on LinkedIn or it's probably the best place is LinkedIn to get a hold of me or my website which I can show you so um yeah so new content coming out I can't tell you exactly when it will be it's just that's just the reality um other channel update is I'm not exactly sure how much longer the live streams will be going so like the CH like channels like mine like the channels themselves um the live streams aren't getting pushed as much either so the number of people that attend and or watch the live streams that I do has decreased in parallel with the way the channel goes and I'm not sure if it's all that useful to people you to have the live streams because if we're not talking about anything then what's the point so I don't know exactly how much longer the live streams will go plus the software I use called streamyard is almost $500 a year that's very expensive especially at a time when YouTube revenue is decreasing so you know if your Revenue goes down you have to adjust the costs and streamyard is extremely expensive um or I might learn how to use OBS so I don't I cannot guarantee or I do not know how much longer the live stream will go I will look into that but if I do it would have to be using like OBS or something lower cost or free because the cost of streamyard is astronomical and the return on the investment just isn't there because the way YouTube is changing that's just the reality

### Segment 3 (10:00 - 15:00) [10:00]

hello uh demu hope I'm saying that correctly watching from Ethiopia that's amazing thanks appreciate your viewership and for stopping by the live stream appreciate it very much uh Krishna hello good to see you thanks for stopping by awesome so this is so yeah so those are some Channel updates new content will be coming I can't tell you exactly when it will be it's just when I have the time live streams may or may not continue past um you know April or so I got to figure out so yeah good question here uh rajik says hi from Indiana hello so you may not know I was in Bloomington for many years one of my favorite place in the world I love going to Bloomington hope I get to back get back there soon ask why uh do I use streamyard is it not possible to live stream directly from YouTube so it is not possible to stream directly from YouTube you have to have some intermediary sort of service so that has to be like streamyard OBS um what's the other one I can't think of it's fairly common wirecast I think is still around so you have to connect some piece of software that does soft Ware SL website that does the streaming into YouTube and then it distributes it out so that's the way um live streaming on YouTube works there are free things like OBS so OBS can do SC uh screen recordings it can do live streaming and so on and so forth the reason there are things like streamyard because it has many additional features that make doing all of this infinitely easier so using OBS you have to set up a bunch of scenes you have to do your own like this lower third um Riata like this little bottom thing you see here that's just a built-in feature of streamyard and it pulls that directly from YouTube using the API it it'll record up to 50 hours for you um it's a very good service I'm not knocking streamyard is a great service it has uh an amazing set of features for it but if you're just like me like a single YouTuber who runs an educational Channel and stuff like that the return on that investment of the cost of it is just not there it's great for other people might be great for businesses or somebody else that have you know more Revenue coming in for that but for a single YouTuber like me that has an education Channel where YouTube is decreasing in terms of how it um recommends videos like mine and similar channels the ad revenue is way down so when that happens I can't people like me can't justify the cost of streamyard even though it's a great service it's a really good service so in the tradeoff is that if I were to use obs lot of the stuff that streamyard does automatically in terms of you know pointand click and all kinds of stuff all the things like um let me share so like sharing screens is a whole lot easier so let me go back to here I got to bring this up anyway so I need to share my screen but yeah things like screen sharing and all that kind of stuff are just built in to streamyard okay it's like this like doing this in OBS would be a really a big pain in the ass because you have to like set up you have to manually Place everything where you sort of want it and it's just a pain so we'll see um I may change my mind but I don't know the status of future live streams just because the number of people they reach and the costs involveed in both software and time so we'll see we shall see but I'm glad you're here I'm glad you know and I'm glad several people stopped by to see the stream and the original point of the stream everybody was just to make myself available to people so it's you know it's one thing for someone to put videos up on their Channel and they just kind of put it there and it just that's it but you don't have any like personal connection to the Creator the main reason I started doing the live streams was to have a personal connection with you all to make myself available to you all who wanted to

### Segment 4 (15:00 - 20:00) [15:00]

stop by and say hello or drop in a question or you know start a conversation or whatever else that's why I did it and maybe that just has run its course as all things in life do run their course and maybe this is just the live stream it's running its course and we'll see so I don't know I haven't fully decided yet but um I can't keep spending all the money to do this stuff you know so all right um uh I try to say names correctly so in UK hope I'm saying something relatively close is it okay to ask stats yes that's the point really the main point of the live stream is to ask questions about stats or whatever else and I'll try to answer them if I can't off the top of my head answer it I can follow up with you later but yes you can ask questions about stats or whatever else and again if I can answer it I will if I can't I won't and I can follow up with you later so feel free to ask away or make a comment or throw something out into the chat you want to talk about that is on your mind okay all right so while we're waiting for any questions or comments um go in the chat as you probably saw this has been the biggest week in AI since the release of GPT 4 it has been absolutely in insane what I what I've seen I'm sure some of you have seen this week with the uh release of Gemini 1. 5 which is just assuming it's all true which I since Google got burned last time for kind of Faking things I assume that they're releasing stuff that's actually true and capable now if the capabilities of Gemini 1. 5 are true absolute just Bonkers Game Changer and in terms of well in many things it's crazy and I can go over some of that if you really want to know about more about it some of the highlights I have the page up I can show you if you want to know and of course open AI Sora uh video creation model again just incredible Bonkers type quality and the length of the videos it can create and I think that the hard thing about the AI space right now is if you're a startup that wants to create some sort of AI service it's just a matter of time before open AI themselves or Google or Microsoft or anyone else just comes in and just destroys you because they create the service themselves in their own platform you know so it's not exactly the same but you know will mid Journey continue when you have open AI Dolly and you know what was Google's I can't remember the name of Googles actually but there are there it's able a fusion of course so there are several image generation platforms that are built into the bigger ones but if you're an AI startup it's really hard right now because the big companies will come in and just ruin your business so could you maybe explain the one factor model what do you mean by one factor model you'll have to be a bit more specific do you mean like a oneway an NOA or um so if it's if you're asking about a one-way Anova yes I can definitely tell you what that is so let me know um yeah so I really encourage everyone to keep on top of and I've been an EV evangelist about this you have to stay up on top of the AI stuff otherwise you just get blown away out of the water so the old adages AI may not replace your job but people who do use it will and I think and I've seen that I think that's true so I've been trying to integrate all that into my day-to-day work also in my personal you know work I pretty much learn it non-stop I do my day job during the day and then in the evenings I'll watch videos on YouTube or I'll do unem courses or I'll do whatever on the weekends of course because it just have to keep going right now it's exhausting it is totally exhausting to have to keep up with all of it but I feel like me personally I have to you know I have to keep up with it all because it's just it all changes so fast you naturally and all these companies are in races with each other to get their latest models out or their latest feature sets out or whatever else and it just accelerates the whole

### Segment 5 (20:00 - 25:00) [20:00]

thing and there's no real regulation around it at least here in the US so they're it's free rain it can just go as fast as it these AI services and models and stuff will grow at the speed of Chip production so as Nvidia right here and other companies make more chips these things will just keep going okay so I really encourage everyone to stay up stay on top of this stuff uh learn how to integrate it into your daily um career or daily life or whatever I'm not saying you have to be an expert on AI and stuff but at least having an ongoing knowledge of what's in the news what these things can do and then learn a bit about how to use them to sort of amplify your own skills because I use the metaphor it's not the best metaphor because it's destructive but I see these AI tools as like putting your skills on steroids right so if you are if you have some natural you know muscle mass muscle mass kind of like me I'm a bigger guy if you have that starting base but then you inject yourself with antibotics anabolic steroids you will get ginormous right so you already have the base knowledge there it's just an En enhancer in that case where you can just get you know huge if you have the body mass in the build to do it to begin with you know AI is sort of that too if you already have a good base skill set using the AI tools and software can be that sort of you know that steroid perfor performance enhancing drug I hate using this metaphor but it's the best one I can think of um to supercharge your skills right so think of it that way but you have to like be but it's not like an anabolic steroid it's something that you can freely use and pay for or whatever um so anyway yeah that's the AI stuff uh some of the AI stuff of course so just in case you sort of haven't seen you know these sorts of things um let's do this one yeah so this is the this is the Sora image generation model from open AI um I think the prompt for this is a flock of paper airplanes for Stuff and this was the prompt a stylish woman walks down a Tokyo Street filled with warm glowing neon animated Sydney signage she wears a black leather jacket a long dress and black boots and yes that is AI created you can see a little bit of weirdness in her boots the way she walks the physics aren't quite right there but still I mean my goodness that is just insane all right um historical footage of California during the Gold Rush again this is AI created this is not you that's not a drone there were no drones back then there was nothing flying back then which is that brings up a good point everybody so here in this what we basically have is a drone shot of the 1840s in California so think about that this is an experience that no one ever saw or never existed or could have existed because obviously there weren't flying things in the 1840s however by using AI we can now imagine what that would look like or recreate what something like I mean it's B it's Bonkers it is absolutely crazy you know so I'm not going to go through this whole thing um I would encourage you to look at this if you haven't already I'm sure many of you are up with this kind of stuff um but it really is just it is really impressive I mean look at this is a this is a fictional Art Gallery look how stable everything is look at the permanence of all the images on the walls and the floor that's a complex pattern in that

### Segment 6 (25:00 - 30:00) [25:00]

floor with the wood grain that has to remain consistent as it goes through and that's completely generated it just it really is well this is one of the weird ones that you get some odd stuff but anyway so really check that out um the technical paper is actually very interesting too on these so yeah so definitely check that out so you know like what the new reality is with these sorts of things okay I'll get to the Google thing here in a second let's see here um so I might have to get back to you on this because um well first of all I'm really impressed that the math a lot of the math showed here so factor analysis um I haven't done a whole lot on factor analysis on the channel because it's uh it's used a bit more in marketing um marketing segmentation um product characteristics you can use a lot of things but I haven't really done that but um the basic gist of a factor analysis and I'd have to go back and really refresh my memory but take this with a little grain of salt here but the basic gist of factor analysis is finding sets of variables uh among many variables that are correlated sort of correlated with each other they tend to go in the same direction and therefore they create sort of a meta a meta measure that's sort of that's a factor okay and those variables can be grouped together into a factor and then you might have other variables that do the opposite and that might be a second factor and the basic idea is for those factors creating those factors and such a way that they explain away the most variance possible in the dependent variable it's kind of all the same thing but basically it's a way of dimen dimensionality reduction where you shrink the number of variables into smaller variables and the factors that all tend to move sort of in the same way or have that relationship they all they can go in different ways but those group of variables tend to go in the same way and the idea is to use those factors to shrink the number of variables and explain as much variation in the dependent variable as possible so um that is that's the gist of factor analysis which is very which is a com which is the same thing we do in many other things in statistics and in machine learning and stuff we're trying to explain way the variance in the dependent variable trying to like gobble up sort of the sum of squares is what the way I explain it and factor analysis is just a way to group variables to do the same thing okay um and sometimes the way it works is like you you'll have two factors and you have to rotate them a certain way to get the final to get the most VAR variation explained but that's the basic gist that's the basic sort of idea and we see this like every day you know there is um you know one of my favorite magazines is Consumer Reports because they do all the independent reviews of products and services and different you know sorts of stuff and they'll do a lot of this sort of work where they'll look at different products and look at um common factors among all the products they review and stuff like that that's a that's one example Market segmentation is another so obviously in the company I work in and many other companies do Market seg Market segmentation so they might go out and look at all their customers and try to find um you collect all the data about their customers and then reduce all that information into certain factors or personas you know personas or a common parlament in business so that's another um way you can apply factor analysis it's very big in marketing for that reason okay so I hope that explains a bit I'm not going to get into the the math notation um because that's a bit more I'd have to go one have to go look that up because you know this dropping math notation in the chat is not all that helpful to me but that's the basic gist okay a fact a factor analysis at least that I can do off the top my head hello um why manic hello good to

### Segment 7 (30:00 - 35:00) [30:00]

see you how can someone be good at maths cannot become good at Stats so I think it depends on the level of stats you want to go into but I will say that I think if you really want to have a basic understanding of Statistics like an intro level statistics background I think all you really need is like algebra you don't need anything more than algebra to do intro stats for two reasons one intro stats tends to stop at like linear regression or multiple regression and those are just linear equations you've seen those before in algebra okay that's the main reason you don't get into more complicated stuff in an intro level sort of stats class now if you want to go higher in stats you will have to have some and this is where it's problematic I think you will have to have some experience working with matrices because a lot of multivariable statistics and machine learning and stuff and AI for that matter is all matrices it's all working with matrices of numbers so you have to have some understanding of Matrix operations if you want to go a little bit higher you in in statistics that being said though nowadays it's good enough to kind of know what's going on without doing all the math you're not doing the math by hand anymore this isn't the 1960s we're not doing math by hand anymore or probably they were using computers by then but you know 1930s we're not doing math by hand we have computers to do the math for us but I think it's important to know enough of the math to know what it's doing how to interpret what it provides back how to read like an equation and know which variables have the most weight in that equation um or something like that I think nowadays unless you're doing like a PhD or Masters in statistics or something you know maybe then you have to know all the advanced math but nowadays we have computers to do that for us that's why we made them to do this stuff for us so I wouldn't worry too much you know about the math unless you want to go higher but for intro stat I say you just need to kind of know you know algebra and then if you go into machine learning and stuff not only Matrix operations but then some calculus as well because things like gradient descent and other um and other types of algorithms that try to that have like loss functions and stuff where you're minimizing loss functions are calculus and if you look at some of the um other courses on machine learning you'll see like partial differentiation and stuff like that but again it just depends on where you want to end up so your level of math knowledge should sort of match where you want to go and of course you can always get better at math so nowadays there are many tools to get a little bit better at math if you feel like you need to there are many online resources of course there are free books out there you can get there are used courses on corsera udemy edex um YouTube that you can do you can just get better at math and or you can get better at math through the learning of Statistics by doing the statistics and following the examples okay so I don't think having a master's level knowledge of math is a prerequisite for being good at statistics at all I don't I think nowadays the actual solving of the problem that's done by computers is the minor thing the really important thing in statistics now is identifying problems formulating problems knowing what questions to ask uh matching the questions you're asking with a problem that you then need to use statistics or whatever else to solve and then doing that process and then more importantly interpreting and communicating the results that you get back to others that's the real juice the statistic themselves are an increasingly minor part of that because of uh computers and that's just sort of my view okay but good question that's a very good question all right I'm not sure this will all fit well it does can you see me above it see this is why streamyard is good because it'll do put all this in here that's one that's another benefit of it uh Nema

### Segment 8 (35:00 - 40:00) [35:00]

asks I've recently decided to brush uh brush my knowledge on statistics hypothesis testing AB testing sampling computational do you have any suggestion on where I can find real world problems on the uh those topics with answers so that I can try them there are many resources on machine learning but it seems there are not uh that many for statistics I think one reason can be a fact it requires someone has deep know really analyze the outcome and interpret them okay um well like any I'm going to remove this because it's so large I may keep I'll just sit up taller how about that so um finding real world problems for statistics uh depends on for the type of Statistics you want for the topics you listed like hypothesis testing and um two sample testing and stuff like that just any basic intro stats book will have tons and tons of problems in there that are based in the real world most like especially got a business statistics book so like um so like the book obviously I work if you don't know I work for sage group used to be S Sage learning and I work on the product team that publishes this book right so this is our stats for b& stats for business and economics book it's ginormous okay um this is the 15th Edition but you can find an older edition used somewhere probably all the problems are business and economics based you know it's not like oh here's a math problem no it's like you know something like um if I can find an example here yeah so Delta Airlines quotes a flight time of 2 hours 5 minutes for flights from Cincinnati to Tampa suppose we believe that the actual flight times are uniformly distributed between two and uh two hours and two hours and 20 minutes and then you're ask several problems about the uniform distribution based on the data they provided in the problem about Delta Airlines so you can find a lot of like simple busy type problems just in intro business stats books you know is a great place and there's a there's thousands of them in here if you wanted to do that way okay I'll put that away later because my bookshelf is a bit far away so that's sort of the easiest thing honestly just um just buy an old stats book that has some of the answers in the back and work them just work them okay um you can also try I have a set of study cards upstairs that I think they're from Barons they're like AP statistic study cards that have like the problem on the front and then the explanation on the back you can work through stuff like that you can go out and collect your own data if you want you can um create your own problems if you want and of course there are many courses like on a udemy or edex or corsera or just YouTube for that matter like mine if you notice almost all my videos try to have some sort of real world business example as the initial problem you know so you can see the statistics in context you in context but there's no substitute for just doing you know working on problems you know just sitting down and working through the examples and it's good to work through the examples if you have like a textbook because the examples in the textbook are like a prototype a prototypical scenario or problem that can fit many different types of specifics but you see that prototype in the in the chapter kind of how it's worked step by step and then you apply that General problem solving technique to many other specific problems that's why textbooks are really good they're really good at that I mean that's their point right so just uh just find you a good a textbook like um stats for b& that I held up um the author's cam mm you can probably find the 13th 14th whatever editions floating around out there um that you could maybe pick up by the way that book I just held up is the market leader by far in the United States so it's good it's been around for a long time okay good question let's see the math doesn't show in the LinkedIn live stream yeah it's probably a little bit different you know I can't I don't control how that works obviously so the API for LinkedIn live may be different than the API for what is YouTube in terms of showing some math symbols like

### Segment 9 (40:00 - 45:00) [40:00]

this you know um it's not perfect but it does show some of the Greeks the Greeks in there all right so good question awesome thanks yeah sometimes it's not all that um complicated like in this case if you want to do real world based stats problems just find a business you know statistics for business economics book and it's it has thousands of them literally thousands all right and usually with explanations examples and all the other pedagogy that comes with just having a good book on the Shelf you know I know we shy away from books now but you know over to my right I have three bookshelves of the most important books and I still keep physical copies of books because I use them all the time as reference they're my reference books now it helps that I work for a company that makes books that's why I have probably more than most people but yeah I keep them over there some of them I don't probably need all that much but uh I still keep them you know because they're harder to find so just pick up a good business for stats for business economics book is my recommendation there sometimes it isn't more any more complicated than just finding a good old book all right yeah keep questions coming thoughts questions comments put them in the chat we'll get to those I'm going to bring this up while I wait for a few more thoughts comments and questions in the chat give you some time to populate that um the other thing I'm just going to show you if you haven't seen it already is Google's Gemini 1. 5 is absolutely insane it's hard to get me shook like just like dear and headlights like watching stuff and get shook by stuff I'm pretty hard to impress honestly Thursday was one of those days I was shook between Sora like between Sora and Gemini 1. 5 I was shook so what are how do how did this change things I think that's what everyone should understand of how this new model by Google changed things because we're used to chat GPT at this point you might have messed around with Google was barred but now it's um Gemini what's the big deal and why are people crapping their pants um over what this sort of thing can do one of the first things is what's called it's context window so if you use chat GPT you have that box you can type in and paste stuff in or if you're using the API you can you do it uh that way it's basically how much stuff you can give the geni model to work on at one time that's the context window and usually that includes what you give it and its response back that's the whole context window and it deals with tokens so a token is a part of a word so like the word stopping is probably two tokens stop and in that's what a token is so you can think of it as one word is about 7 or 075 um no one token is about 75 words seven words okay what this can do this bottle can handle a million tokens a million so as of right now I think GPT 4 is 128,000 I think this is a million another way to think about that is that you can put into a prompt or a context window almost the entire Harry Potter series of books at once all of them that's about 700,000 words so a million tokens is about 700,000 words and it can work on that much context in what one time which is bat bleep crazy okay it's really hard or was really hard for llms to do something like that so it's the amount of information it can process and of course it's not just text because Gemini is fully multimodal meaning it can handle text it can handle images it can handle video it can handle

### Segment 10 (45:00 - 50:00) [45:00]

audio or any combination thereof so you can feed it like an entire video and it can turn that into tokens that it can understand and do all kinds of crazy stuff in there it is absolutely bunkers and um so for example let's see this is a 44 minute video right here it's I forgot the the frame rate but anyway so this 44 minute video when turned into tokens is about 700,000 tokens okay now what they did here I'm not going to play the video or anything because you can go watch it I'll put the link to this in the chat what they did was they handrew like something from one of the scenes in this case it's like it's a water tower with some water coming out of it they drew it by hand they gave it the image they drew by hand and said go find the place in the video where this thing I drew by hand is and it can it does it uh almost perfectly perfectly so but that's crazy of a 44 minute video it can find something similar to a handrawn image that you provide it and be perfect every time it's just crazy um this is a long block of code a 3js so it was you know over 800,000 tokens and they asked it to find the place in the code that creates this thing and it did it okay um the other thing that is really amazing is the try to find it here needle and a hay stack I can find the graph here maybe it's not in this post okay not in this original post it can it's what's called needle on a Hy stack so you can give it this huge massive context window and ask it a very specific detail in that huge context window and it can find it and the way they tested that was they would go into this long context window of text like hundreds and hundreds of thousands of words and they would put some sort of secret thing in there like a sentence like the secret word is banana they would just drop that into this huge context window of text like the secret word is banana then they would ingest all of that and then they ask it what is the secret word and what page is it on it will reply the secret word is banana and it's on page 8117 you can find that in all that text it can find that which is crazy um so the the and it does it fast it's very very efficient so the things that this can do it's not public yet but the things that this stuff can do now is absolutely just incredible it's incredible and again I'm not saying that you have to be an AI expert but it's good to stay on top of what these developments are because if you're not already you will be impacted by it in some way either through if you're a student you'll be impacted by it if you are a um work in a career you'll be impacted by it there's no Escape from it so the only way to deal with it is to understand it by keeping on top of the news what is happening and what's going on and hopefully learning how to ultimately use it to give you your own superpowers because that does that's what this can do this stuff over the past year has given me superpowers that before it existed I wouldn't have I can do now that I couldn't even imagine doing like a year and a half ago because I didn't have the skills or the knowledge now I do so stay on top of this stuff for your own Survival in terms of career or whatever else um because it's extremely important animportant last thing I want to maybe show you is um you might want to try it

### Segment 11 (50:00 - 55:00) [50:00]

out it's also from Google me pull it up here there we go if you haven't seen this is related this is notebook LM right now it's an experimental tool created by Google and it is like an AI powered note taking uh platform okay so it looks um it looks kind of like this it kind of looks like a GPT interface because it actually is kind of based on it but the way this works is you can upload let's do test but what you can do with this is you can upload source documents okay so if you've taken your own notes or have notes from somewhere you can actually connect it to certain Services you can connect it there are ways to connect it to like Kindle and other type of services that'll pull the highlighted things that you've done but you can add sources from your Google Drive via a PDF or just via some sort of copied text you can add up to 20 Source documents okay so let's see let me um let me add something real fast I'm going to add a PDF so you can kind of see this okay so I'm G to bring this back I want sure how long this will take so what I've done is add an intro to statistics PDF this is just an open Stacks PDF so it is open um I can just I can add this in here as some sort of source of text I could also have added an entire Drive full of uh full of documents okay so I should have tested this out I'm not sure how long this will take but the way this works everybody is that in this sort of AI powered not taking system you can add the sort of source documents that you have or whatever else it might be and then you can combine your notes with the AI system to have it like interact with you and explain things to you that you know whatever it is you sort of need to know right um and we've seen this in the past where you can you talk with certain documents but because of the way this is let me um I can't type in here yet because it's not it's still processing I probably should have um done this first but this is called notebook LM and you should definitely try it out and kind of see how it works there are many demos um on YouTube let me find the one this all at this process but I'm going to find the YouTube video that was very good at explaining this I'll put this in the chat me find it there it is so this is the video most of what we consider no taking and from Thiago Forte that's a bad place to pause that's better so this is the video I recommend watching I'll put this in the chat that explains how you can use notbook LM to organize your course notes to analyze your business notes at work um or whatever else obviously I would not put anything in there that is proprietary or sensitive don't put like your company's financial statements in there or whatever else but again this is just another example of how we're finding tools how's this doing it's so processing we are finding tools that will interface with the llms as like another layer to give us more superpowers now the Notebook LM is an experimental tool so it might have you know little bugs kind of weird stuff you know here and there um yeah I think my PDF has taken it's taken a bit too long so I won't mess with that I'll check back maybe one more time but the notbook LM by Google is sort of that tool that you can use your own information with and Google's Gemini to keep track or learn or whatever it is you're doing right so just another example that I encourage you all maybe to check out give it a whirl since we're all here in the education space it is pretty it's pretty crazy me that'll probably keep going so what is this

### Segment 12 (55:00 - 60:00) [55:00]

um yeah so here you can see what this is one of their um example ones so once you upload a source like this okay you upload this so this is the text that was uploaded this is a good example so in this demo they uploaded this text here it's a big chunk of text okay look at that now what happened as soon as that text is added as a source notebook LM then gives you this summary right here does it automatically and then gives you key topics and obviously those are links okay if you click on the key topic it's going to look through here and actually find where in that text Block in this case incandescent light lighting is mentioned and then give you like a summarized version like expanded summary version with all the citations you can click right here look at it you click on one it shows you on the left hand side where it got the information click on number four it shows you at the bottom work out that information see Bonkers okay crazy and then and then you can ask it a question and I think they put um they pre-load three questions uh what are the key factors that contribute to the success and widespread adoption of incandescent light bulbs built by th Thomas Edson and a second question so you can ask it a question oh I need to close this let's go over to here okay um and then you can say save your notes see this is a study guide that was created uh different um the challenges that were part of the process of doing light bulbs you can create all these little notes with the citations to your stuff it is just it is crazy okay that was a preone question um let's see um cost of incandescent I'm not going to WR spelling lighting let's see what it says cost of incandescent lighting of course I can't spell incandescent copy paste see what it says so you can see it down here working so it's all right so it says incandescent lighting bulbs have very high manufacturing costs see and then looked over there here again here are the 10 citations that it found and then I can copy this um I could pin it or whatever else I want to do it is just crazy okay so again this is another example of that let's see if this is done I think it timed out okay I think the PDF was probably too big that I uploaded on the fly so anyway notebook LM something I highly suggest checking out given in a whirl um you can look at the video I put in the chat about it it's just one example of how generative AI is being used as a base for other useful tools and platforms especially For Us in education where we can keep track and interact with our own learning our own notes our own res sources and so on and so forth okay it's not the only one you know there are many others so um notebook LM I'll put that in the chat as well there you go so give that a check that out yeah just play around with this stuff all right it all right last call for any comments or questions in the chat and I'm going to go up to the coffee shop and do some work so if you have any other uh thoughts or questions or comments ments on this anything else I've said in the Stream anything I've not now is the time to ask I'll wait like another minute or two and see what you all say and if not we'll uh we'll wrap up this live stream so I'm gonna take a drink of water while I wait for anything you might want to add in the okay just a reminder go back here yeah just a

### Segment 13 (60:00 - 62:00) [1:00:00]

reminder that the channel is always here uh the new videos will always be under the video section here so um check those out the new ones will look like this kind of they have like a superhero theme so you can always check those out that's how you'll know that they are there and uh I look forward to showing you some more things about orange and stuff like that but again it's just I do it when I have the time and unfortunately you know I have to do things in a priority order with uh what pays the bills and right now that's not YouTube for many people so you know uh YouTube is just changed you know for better or worse it's changed all right so nice info thanks you are very welcome I'm very passionate about this stuff because I think it have to be and you have to be passionate about this stuff nowadays or you're just going to fall behind further behind and you know I'm not saying you have to be a AI accelerationist I'm just up to dat on it you have to know what's going on to be part of the conversation and then hopefully learn as much as you can about it as it relates to how it can be useful to you know and bring new ideas to your company if you work at a company you know bring some new ideas to them as you learn I've been doing that for the past year you know so oh I have this idea and it may go nowhere but at least you're the one proposing ideas right okay everybody I'm going to go ahead and wrap up the stream and I think we'll be back in two weeks so March 2nd but I must always post the live stream the Wednesday before the Saturday so like this live stream was posted Wednesday so you can always look out like on Wednesdays usually for the live stream announcement to go out that way you can plan ahead if you want to join so we should be here on March 2nd unless anything comes up um I don't think anything's going to come up but you never know so kind around March 2nd two weeks from now kind of look if you want to be part of that live stream and I'm sure we'll have more to talk about then so bring your comments bring your questions bring your thoughts because that's what makes the live stream enjoyable and beneficial for everyone okay so I'm going to go ahead and wrap this up and I thank you all for some of your valuable time and I hope you have a very good rest of your day take care everybody bye-bye