What skills should you learn outside of machine learning? | Ask me anything! (live)

YouTube. And I need your help in the chat. I need your help to let me know if we're live. Tell me if the audio is good. I'm testing out this new little microphone here. Can you hear me? Can you see me? I'd love your, love your help. Um, I've got to make sure it's okay here as well. So I'm trying to, I'm trying to get this, this audio set up. I haven't fully worked out how to set up this studio yet. Like this wall is looking, is looking pretty bland. There's a road just over here. Um, so I'm trying to fix this audio. So if you let me know, that'd be amazing. Hello, hello, hello everyone. It's so good to see you. Hello Manuela. Hello Vikram. Hello Subs. Hello Karthik. Ian. Hello Matthew, Leadington Hill. Bring the noise. Ashik, how you going my brother? Shabam, how are you? Poetry. Audio is good. Beautiful. All right, so this is what we're going to do. Oh, actually, if you're new to the channel, there's a fair few of you who might have not been here for a live stream before. We have a lot of fun. We just spend an hour dancing. So wherever you are, please, please get out of your chair. We have to start the stream with dancing. All right, so let's boogie down. Yes, beautiful. What's your guys' favorite dance move? So this is what I'm going to do. I've got a bunch of questions. So boom, we've got the, we've got the comments here and I've also got some stuff on Twitter. So what I'm going to do, I'm going to go back and forth. You guys were so helpful. Everyone was so helpful, um, posting these questions. I'm going to jump in between hitting up the questions on Twitter and I'm going to, um, jump in the chat. Oh, Kenji, you can hear the truck in the background. Hahaha. I don't know whether to trust you or not, my friend. I hope not. Probably can actually, cause the road is literally right there. Um, I always have trouble because I've got like all these screens going on and I'm trying to keep the chat in my field of view as well as the, the questions. We'll figure this out. How are we all? It's a beautiful, it's a spring day in Australia. See, look at this. We've got this and I need Streamlabs. I've got an idea. I've got an idea. Streamlabs over there. Boom. Okay, so if you're new to the channel, welcome to the channel. My name is Daniel Bourke. I live in Australia and that's about as much as you need to know about me because I'm sure you might've found out some other stuff. Um, but I do one of these live Q&As once a month or so just to first and foremost to have fun and hang out with everyone. And second of all, to answer some of the questions that I get most often and to, to just answer or Let me put a disclaimer. This is a big disclaimer. I do this at the start of every stream as well, is that whatever you hear from me, make sure you take it with a grain of salt, right? Because everything I say is going to be biased with my own experience and that goes for anyone else. So be careful of what anyone says. Just because I can talk in front of a camera doesn't mean you should adhere to what I say. Okay? That's a disclaimer for everything said in this stream. But anyway, Now that that's done, let's get into the fun part. We have, oh my goodness, we have a fair few people here. I'm going to just start jumping straight into questions. Um, Ken, of course, Ken G, my boy Ken G is, uh, talking smack. I don't know if I can swear on live streams without being taken down, but anyway, he's talking smack in the chat. I was going to say another word beginning with S there. So let's jump into the questions. I've got to switch over here. I should really get like a little keyboard shortcut. So I'm going to start with my boy Ashik. He's asked in the Discord chat. I've got a Discord chat for the machine learning students of the Zero to Mastery machine learning course. He says, how to study and implement ML/ML papers because most people couldn't get a proper way. Any resources for that? How did you do it? What's your tip for doing the efficient way? I'm going to be straight up with you, Ashik. Um, I don't really have an efficient way of doing it, so I don't really know if there's an efficient way. You know, there's a really good quote from, from one of my favorite scientists called Richard Feynman. Let me look this up for you cause I want to read it out. Um, I've got a little, I've got a little note here with all my favorite quotes. So study hard what interests you the most in the most undisciplined, irreverent, and original manner possible. Let me just get this on the screen for you. We'll go over here. Boom. So these are my favorite quotes. Zoom in here. So this is, I think once you start to, this is a quote I was talking about. So Ashik asked, how do you implement machine learning and deep learning papers efficiently? I think once you start to over-efficient, like make things over-efficient, they start to lose their fun. Like, I like the challenge of, of going through a machine learning paper. Of course, once you get a little bit more practice of trying things out, you can get a little bit better. But if you're over-optimizing things, well then again, it just becomes just Groundhog Day. It's the same thing over and over again. And so I'll give you a little story. When I was at my previous job as a machine learning engineer, we had this, this natural language processing problem and it was, it had, I think, I don't know exactly how many, but it was hundreds of labels. So really extreme multi-classification. And there was a few papers out there. One was using LSTMs, one was using convolutional neural networks. And I spent basically 2 full days trying to implement a paper. Like I'm talking from, I got in in the morning and I read this paper and it said this is the architecture that we used for these state-of-the-art results. And I just spent 2 full days trying to implement it. The first 6 hours were hell because I had no idea what was going on. But after that, because I sort of had read through, I'd written a bunch of code. That's really important. As much as you do thinking, make sure you do doing. And then finally, by the end of the day, day 2, or may have been day 3, I can't even remember, right? That's how poor memory is, I got it working. And it turns out that it ended up getting the state-of-the-art results, but we couldn't end up using it because this is what the caveat of using state-of-the-art architectures is, that it took too long to run. And so that was kind of a long-winded way of saying is that I don't really— if I— I'm very skeptical of anyone who tries to be over-efficient in anything. So if you're trying to implemented machine learning or deep learning paper, go through the journey of figuring it out, even if it, even if it gets hard. And if you don't understand something, that's all right, leave it and come back to it. Okay. So now we're back to the G85. Maybe I should have— doesn't really matter. Um, okay. What's another question? We go back to Notion. Uh, Ashik also asks, David Goggins says suffering is what differentiates top performers and normal people. What's your call on it? I'm a big fan of, of David Goggins, but I think you can flip this coin in two ways. So this is the question that we're on here, right? This one here, I'm going to zoom in a bit so you can see that. Um, so suffering is what differentiates top performers and normal people. What's your call on it? I agree, but I think you can look at it from the 180 as well, is that life shouldn't always be about— you shouldn't have to just continually— what you get out of life shouldn't be decided by how much you can suffer in comparison to others. So there's, there's two caveats to this. So one side is yes, if you can suffer more than others, you might achieve things that others don't. But then the other side, you can just argue as well, it's like, well, I don't want to suffer. I just want to, to live life with it with as little suffering as possible. And that might be perfectly okay for that person. So there's two sides to every story and this is, well, actually, there's more than two sides to every story. So David Goggins has gone through the trouble of figuring out what worked for him. And so yes, in my case, I would agree that there are times where I've gone through periods of suffering and it's made me stronger, but there's also times where I've gone through periods of, of fun and basically zero suffering and it's also been a great time. So this is what I want you to think of is like, not, I personally don't have all the answers, but this is how I approach things is go, Okay, if what David Goggins did by running all those ultramarathons and all those amazing things that he did, if that worked for him, okay, I might try this for a little bit. And so I went running and I did a whole bunch of running and I thought, okay, this is great. I've kind of, I think I've learned about 80% of this lesson. I'm going to move on to the next thing and see what works for me. So that's probably the best thing that I can give you from that is that I really like David Goggins. But I also really like taking little aspects or characteristics from the people I look up to and creating my own style. And you know what's even better than if you don't have someone to look up to is if you have someone who you don't want to be like. So maybe you have an anti-role model, someone who does things that you don't value. Maybe you just avoid those. You can live a really good life by avoiding the things that you don't like what that other people do. So we'll tick that one off and we got the third one from Ashik. Great questions by the way. Anyone who's asking questions, the more specific the better. So thoughts, if you ask me thoughts on different things, uh, I can ramble on for a while, but if you ask me specific questions then I can, I can answer them as best I can. So Ashik says, how to get over addictions, whatever it is. For example, for me it's, I blank this out, uh, your takeaway on how to get over addictions in your life. Hmm. What have I been addicted to? I've been addicted to video games. Um, I think I have a very addictive personality and that's, to me, that's not a bad thing. You see, I see addiction, I see addiction as like energy in the universe. You know how the, the law of thermodynamics says energy cannot be destroyed or created. I see that, that you can use that for addiction. Once addiction is created, it cannot be, uh, it cannot be destroyed. So it can only change form. That's like energy, right? Energy can only change form. For me, once you, cause when you're born as a baby, maybe you're not addicted to anything, but then eventually once you succumb to what society has to offer, and you figure out, I was like, oh, video games, they're real fun. I'll play them for 8 hours a day. Um, or maybe you get into whatever, drugs, gambling, all the, all the things that people call addictions, social media. People are addicted to the internet these days. I'm probably addicted to the internet. So once that's formed, to me, you've kind of got that for life. And what I mean by that is is you've got the essence of addiction for life. And so the way I look at it is that I can, if I'm addicted to video games, I was when I was a teenager, I used to get home from school and play it 8 hours a day and I was very good at that. But, oh, here's the thing. This is back to the David Goggins and suffering thing. To me, playing video games was not suffering. I got very good at video games and it was a whole bunch of fun. But to me, getting good at video games is like getting good at masturbating. It's just, unless you're a big streamer or something like that. It's just, it's not something you want to brag about. Anyway, uh, that's just me personally. If you play video games, enjoy them. Um, but if you want to get over an addiction, you have to, it has to change form into something else. And so for me, my addiction probably went from video games into fitness. I was really addicted to, to health and fitness for a few years. I still am in a sense, but not as much as I used to be. And now it's probably shifted, changed to, to learning and to creating, to writing. So that's what my addiction is now. And so what I'm saying is you don't necessarily have to get over the addiction. It's just targeted in a different way. And now there's a lot of things. And now for me, I know how vulnerable I am to, to being addicted to different things. So video games, porn is another one I used to watch when I first got my smartphone and I'd watch when I was 17, I got the first iPhone. I'd watch porn almost every day, which is just not a great habit because it, it messes with your mind. You don't want that. And so I would have to, the way I would fix these things is even if I had a little inch to sort of be able to access them, then I could not trust myself with, with that addiction, right? I would have to, create a binary switch to get rid of that addiction. So in the case of being addicted to my smartphone and watching porn on the internet, I just had to turn my phone completely off or block websites or whatever. So that's what I do when I'm studying and whatnot. Now it's still like, I still, I'm still addicted to my phone, but I'm getting better at managing it. I put it in a drawer and then I focus my addictive energy towards something else. In this case, it's for me now, it's learning and studying, creating different things. And so yes, I still look at my phone from here to time to time, but I have to, I've told myself, I've put the, the idea in my head that it's a bad thing and to once it goes over a certain amount of time per day to just get rid of it, put that thing down. That's not what I want. I want to use my energy elsewhere. So that's my advice for getting over addictions. Is don't necessarily focus on getting rid of the addiction itself. Just transfer that energy somewhere else. And if you can't trust yourself, uh, create a binary switch. So if it's you're using your phone too much, turn your phone off, put it on the other side of the house or give it to your friend and just live without a phone for a month. I think that might be an experiment I might try. Anyway, whew, great questions, Ashik. Let's now go to the chat. Um, Hello everyone. So happy to see you all. Sorry, I've been ignoring the chat. Uh, actually not sorry. I've been, I had to, this is, someone asked me on one of my videos to shorten them down and I think I like longer form discussions because if you're going too short form and then the information that you need to, to get out doesn't necessarily get there. And I know it's a talent to be able to compress different things, but To me, I like to hear long-form conversations. I like to, to have discussions. So let's go here. Matthew says, shit is a perfectly valid technical term. I totally agree. So Daniel Wang says, does machine learning ever get repetitive or is each day challenging coming from a high school student's perspective? So great question, Daniel. And first of all, I mean, second of all, great name. So let's go in here. So Daniel YouTube says boom, and we'll turn that into a little to-do. And then so I'll go chat and then back to Twitter, chat, Twitter, chat, Twitter. All right. So this is the question that we're on. Does machine learning ever get repetitive or is each day challenging? Coming from a high school student. Of course. So again, each day, some days do get like repetitive, like when you're working on a problem and you're, you've, the things that you've tried so far, you're stuck, you're kind of hitting your head up against the wall. If anyone has tried to learn to code before, I mean, I have this all the time still. I'm still learning to code. Um, you think that everything you're doing should be working, but for some reason it's not working. And then that's when it starts to get repetitive. But again, you can flip it from the— and look at it from the other side. Well, that repetitiveness, instead of seeing it as repetitiveness, you could see it as a challenge. So that's what we're all looking for, right? Where we want to get up in the morning and we want to go about the day. We want to have fun. We want to find a challenge because if you didn't have a challenge to take on, kind of life would just be a bit bleh. So for me, when I find that it's getting repetitive, I take time away from it. Usually, um, if I'm sort of lacking motivation or something, it just because I'm stuck or because I'm at the threshold of where my current skill level is and that's a good place to be. And so I got to ask myself, is it okay? Am I taking a break because this is just getting repetitive or am I wanting to take a break because this is becoming challenging now and I'm trying to avoid doing the hard thing? So that's again, comes back to that looping into the suffering. You could call it suffering, but for me, I don't like that term. I like to call things fun. Like that's why I started this livestream with a bit of a dance. So Daniel Bourke, to really succinctly answer your question, is that the field of machine learning is incredibly broad. I mean, you've got math, you've got code, and you can apply that to basically any other field you could imagine. That's why I'm so interested in it, because I have an eclectic group of interests. And so if it does get repetitive, you can find yourself a challenge in some, some new subspace of the field. So there we go. Great question, Daniel. Now what we're going to do is we're going to go to Twitter. Hello everyone into the chat. Um, let's, uh, where's my Twitter window? Nope. I'll just get a new one up. Look at me struggling to use a computer. So we go here and we're on. So these are some of the questions I— everyone was so kindly to, to offer from Twitter. Oh, here we go. During your 5-day fast, let's do this one. I want to do this one because, um, that's what I just did over the last few days. So Lazy Python says, so turn into to-do list. So this is the question that we're on now. Boom. During your recent 5-day fast, what changes did you see in yourself both mentally and physically? And so, um, when did I start the fast? It is the 5th of October. And I started the fast on the 28th of September. So last year I did a 4-day fast, so just water only. And this year I decided to go 5 days and it was really just because— why did I start? I wanted to just see what it was like not to eat for 5 days and it turns out It's a— I'm going to put a video out on this by the way, so I'm not going to go too in depth, but Lazy Python specifically asked, um, what changes did you see in yourself both mentally and physically? So the hardest part, you probably want to know the hardest part. The hardest part was on probably midday day 3. I was just feeling— so I'd gone already at this point over 60 hours without food. So just water only and a little bit of salt. You'll see this in the documentary coming out. Daniel Bourke, spoiler alert, and my friend was doing it with me and he said that he was going to go the 5 days as well. But then it came to about midday on day 3, we'd just gone for a fairly long walk on the waterfront. It was kind of hot, the sun was out and he got to the point and he was like, you know what? I'm not feeling 100%. I'm going to break the fast. I'm like, no dude, just wait an hour if you're hungry. It goes away. It's like being horny. It goes away after about an hour. So that's— don't make decisions when you're hungry or horny. That's probably the best advice you're going to get from this whole stream. It goes away after an hour. Trust me. And he's like, "Nup, I'm going." And so he goes and he has lunch and then my little brother was doing it as well and he decided that he was going to end his fast as well. And so I was just like, well, Now it's just me running solo and you know what I did? I had a nap and after the nap I woke up and I found great. I felt great. So after that, the hardest part was just missing the ritual of eating food. So that's one of the funnest activities that I can do in life is sit down with others and eat food. I usually do that with my family every night. I look forward to it. And that was probably the biggest thing that I missed because very counterintuitive as well. The longer you go without food, the less hungry you get. Um, but Lazy Python asked more specifically is what changes did I see in yourself both mentally and physically? So without giving, um, too much away from the upcoming documentary, I would say Physically, I, the lowest I weighed in the fast was 70.5 kilos. I'm not sure what that is in pounds. 70.5 kilos. So 153 pounds, 155 pounds. Um, and I started at 74.9 kilos. So I lost 10 pounds in 5 days. Uh, I did notice that I lost a lot of water weight. So right now I'm, I'm pretty damn lean. Like my abs are fairly showing. Look at that. That's a flex. Um, legit a flex. And if you're, if you're coding Python, you gotta be rocking pythons. Let's be real. So that was probably the major changes physically. Uh, mentally I became a lot more calm. And it's very hard for me to describe this, but I felt like I was just in tune with the universe. I felt as if I had nothing to do. I could just sit here. Whatever I'm doing in this existence was just perfect. And it's very, very hard to explain that. But that was day 5. Probably the first 2 days was kind of just breaking the habit of eating. And then by day 4, day 5, I was just like, I felt like a Jedi. Like I was very chilled, was not hungry at all, but I felt as if I needed to go into battle or hunt a wild animal or something like that for food, that I could definitely do that. So that's kind of how I felt mentally. Uh, did it make you more mentally fatigued while coding and studying? I will say that I did have like naps during the day. But when I actually sat down to study or code, I felt, I felt great. Like I seriously felt great. I didn't notice any sort of, like my energy was like this, like it was just a solid line. It wasn't like spiked up, spiked down. It was just chilled out. But yeah, so that's, uh, that's some details on the fast. There'll be a longer video coming out on that. Let's go to the chat. Where are we at? Um, Hi Tiao says, how many Pomodoro you use to study a day? Um, so I, um, I used to do up to 10 Pomodoros a day, but now I've kind of gone in longer stretches of blocks, so I won't do a 25-minute timer. Now I'm finding I do 2 to 4 hour blocks, so as in like I'll put I'll put a timer on for 2 hours, nothing but one task for that 2 hours. I'll put a timer on for 3 hours, nothing but that one task for 3 hours or 4 hours. Probably the longest I usually go for it at the moment until I get into like a deep focus stage. Maybe I could go for longer, but that takes some practice. Uh, but I find if I do at least 4 hours minimum per day, that usually ends up enough. Some days I do more, some days I do less, but I try to aim for 4 hours of unbroken work per day. Per day, and that might be 2 2-hour blocks or something like that. Um, here we go. Where are we at? All right. Path says, please give a shout out to me. Shout out to Path. How are you, my friend? And now let's go back to the Twittersphere. So what do we have? Is school worth it? Uh, depends what you mean. So this is, this is what I mean by, by questions. So there's so many different ways to, to answer this question. For me, school was worth it. Um, but it depends how you look at it, of course. So Nimesh says, there are ample number of machine learning tools and open source code out there. I want you to focus on data creations of tasks can be monetized to be used by such models. I'm not sure what question that's asking. I like this one actually. If you have a venture investor in front of you, who would you like to invest in your idea? What kind of problem would you like to tackle? Boom. Who's this from actually? Andre. Andre, good name. I know an Andre. Okay. Andre says, this is the question that we're on. I'm highlighting this so I can go back over it later and, uh, put timestamps in this video for the people who are watching this in the future. Sue, if you have a venture investor in front of you, who would you like to invest in your idea? What kind of problem would you like to tackle? Now, if you're not sure what a venture investor is, a venture investor is, say for example, I wanted to start a business and I needed, say, $1 million, maybe more. Uh, let's just say $1 million because easy number. If I needed $1 million to get that business up and running, like I needed to buy 100 computers, I needed to hire, not 100, I needed to buy 10 laptops and I needed to hire 10 people and I needed to pay them $100,000 for a year so we could build this And so a venture investor will go, you know what, I like your idea. I like your team. I'm going to give you this $1 million, but in return I'm going to own 30% of your business. So if you make $10 million next year because your thing is so great, I'm going to take $3 million of that. So that's a very basic equation. You can look that up. Anyway, who would I like to invest in my idea? I have one specific person in mind actually. Let's go here. Chamath. I don't actually know how to spell his name. Papantelaya. Hopefully it shows up. Chamath. There we go. Did I get close? No. Anyway, so this is Chamath. You should check out his stuff. He's a CEO at Social Capital. And social capital is solving the world's hardest problems. Uh, chairman of Virgin Galactic and owner at the Warriors. And so Chamath, social capital. I would want Chamath and social capital. I would be like, let's go back to the question. Who would I like to invest in my idea? Well, Chamath, because I'm not going to talk about him. I'm going to let you research him. I just like his vision of how the world would go. And the kind of problem that I would like to tackle, it would be twofold. It would be, or potentially, so the problems I would like to tackle or most interest me, uh, number one is education and number two is health. And so kind of the, the way I would like to see education is I think the, the current model of education is kind of broken. So there's a lot of free resources out there. I'd like a way to, to build a kind of create your own curriculum style thing. And if you want this, if you want to build this, take this idea and run with it, please do. I'll share it out if you do make something. But as in like, I'll grab this YouTube video and, or this YouTube series of education and put it in my track. And then I'll also grab this Coursera course and put it in my track. And then you can go to myowncurriculum.com and see the things that I've done. Right? And then if someone else wants to study something similar to me, so say I was learning, you were learning machine learning and you created your own curriculum, they could see your curriculum and go, "Oh, that's pretty cool. I want to join in." And then they could contact you and go, "You know what? Do you want to be a study buddy?" or something like that. I'm not sure exactly how it would work. But I feel like there is, especially in the world we live in today with what's happened in the last 9 months over 2020, the idea of being in a certain place at a certain time is kind of getting shifted. So just as remote work is becoming normalized, remote education should definitely become normalized. Um, so that would be one problem, just education and having verifiable skills and whatnot. Uh, the second one would be health. So again, the events that have unfolded this year, there's been a lot of talk about like vaccines and, and, uh, and ways to, to combat the coronavirus, like social distancing, all that sort of stuff. But there isn't— there, the voice in the crowd is not loud enough for just doing the basics, like getting enough sleep, Taking care of your nutrition, taking care of, of, uh, your movement. So moving every day, these simple first principles of health. Um, so I would probably do something in that space, specifically the nutrition because that's what interests me most. Now I've, I've just done an experiment with the 5-day fast. I didn't actually eat anything for 5 days and I felt phenomenal. So I think there's a lot more research that should be done into fasting. There's a great app called Zero Fasting. You should check that out. That's what I use to track my fast. But yeah, I feel like we're, we're still only scratching the surface with what our bodies are capable of. And I think again, this is another point, science and technology kind of gets in the way of things that, that we already do know. You see with fasting, right, there's not much scientific evidence behind it or it's, it's amounting. But if I, when I told someone that I was going to fast for 5 days, I told multiple people, or when I was fasting, I went to BJJ training. I said, I haven't eaten in 4 days and I'm still going to fight you. They thought I was crazy. Right? But there is, why is there fasting components built into every major religion around the world? And so the way I filter the quality of an idea is how long has that idea been around for? That's called the Lindy effect. That's something you should look up as well. Let's go. So yeah, that was answering Andre's question. Who would I invest in my idea? Chamath Papantia, Papantia, that's his last name. Paliapantia. That's my, that's my very poor pronunciation, Australian pronunciation of Chamath's last name. That's who would invest in my idea and it would be something in the health or education space. So let's go to the YouTube chat. Boom, boom. But I also like the fact that instead of a venture capitalist, I'm trying to build my own business here and so that I can self-fund myself and just build these things. And then once I've got like some sort of 70% working prototype working and be like, hey, Show the world, look at this. So that's, that's where the ball is rolling for me. Now let's go back to the chat. Um, Deadwood says, so you really didn't eat anything for the whole 5 days? Yep. Water only, my friend. Water and salt. Well, I'm going to bring this over here so I'm not looking the entire other direction. So Daniel Wang says, I'm a high school student. Daniel, another great question. A lot of students are really suffering from the change in education. Everything is over Zoom now. The education in the US really needs to change, like you said. Yes, totally agree. Q Cole says, hi, this might be a vague question. I'm torn between Python and JavaScript as a backend engineer. Which language should I Learn, uh, choose one and stick with it. Like, just, just learn one. Whenever— if anyone asks me, like, should I do PyTorch or TensorFlow, or should I do Python or R, or Python or this or this or that, or math or code— just choose one, stick with it, do it. When you need the other thing that you, you run into problems, use that if you need to. But the curse of the engineer is trying to learn too many tools. So just stick with one, get really good at that. Chances are if it's been around for a while, it'll stay around for a while. Remember the Lindy effect, the longer something has been around, the longer it will potentially be around. Python's been around for 20 years. JavaScript's been around for longer maybe. So if they're gonna, if they've been around for that amount of time, they'll probably stay around for that amount of time. But if you're stuck between two things, you're going to paralyze yourself trying to pick two different things. Choose one, stick with it and go with it. Um, so Carlos says, what other field outside of machine learning is worth studying that might not be directly related? Oh yes, this is great. Great question, Carlos. Let's go into here. These are the type of questions I want to do. Nope. See, Notion, why don't you— All right, let's go. We want the screen. So this is the question that we're on now. I'm going to put that back over here and boom, just, just rearranging about 17 different windows. Okay. What other fields outside of machine learning is worth studying that might not be directly related? So machine learning is pretty broad, right? Like you're learning math and code. These things can be applied to at a fundamental level. You're learning math and you're learning code. These things can be applied to lots of different things. But I would also argue is that I would learn to communicate your ideas. So, as in, can you, can you write about something that you've built? Or can you talk about something that you've built? Because at the end of the day, you could build something amazing, but if people don't understand what you're doing, and I mean, you might have some nerds, if you're doing things with math, that'll understand what you're doing. But you also need to be able to describe it to, to someone who doesn't really understand what you're doing. And so there's a great article I want you to look up is, um, All the Greatest Innovators Are Hucksters. Where is it? Nope. What's his name? Oh, I really want to show you this article. Pocket. Pocket. I am a huckster. Yes. All right. So Google this title, Why Our Greatest Innovators Are Supreme Hucksters. And so what's a huckster? A huckster is someone who can sell something, get an idea over the, over the line. Well, look at this. The importance of generating hype. Can you see that? I'm going to zoom in. Actually, I'll just open this in Safari in the browser. So what I want to show you, this is another field of, uh, important things. So look, Elon Musk, great huckster, very good engineer, but can also get on stage and tell everyone and get them excited about what he's building. So if we come down here, this is one of my favorite points. Um, yes, the importance. So why am I showing you this is because oftentimes, um, the person who builds a thing will like, or talks about the thing doesn't get as much credit as the person who built the thing. But if the person who talked about it didn't talk about it, would the thing have as much, uh, traction? As it actually does. So you could say this with Apple, right? It's, it's, it's Steve Wozniak and Steve Jobs. Would have Apple have been the company that it is if Steve Jobs hadn't been such a great presenter, even though he didn't actually build many things of the technology? So the importance of generating hype alongside meaningful improvements may explain why many of our greatest innovators, Steve Jobs, Elon Musk, and Thomas Alva, are supreme hucksters. The ability to create the hype alongside the invention may be decisive. Perhaps that was Nikola Tesla's failing. And I love this line. As I say to my vicar wife, a vicar is like a priest. If she, if, if ever she criticizes my working in advertising. So Rory Sutherland is the chairman of Ogilvy Group UK. Fair enough, but no one would have heard of Jesus if it hadn't been for St. Paul. Because what did St. Paul do? He helped get Jesus's word around. So Jesus is the teacher, but Saint Paul helped him to get it to more people. So that is why I think that the field outside of machine learning that is worth studying is communication. So just general, general communication, being able to, to write about your ideas or write about your inventions or, or just talk about your inventions, to be able to orate. So if you want some books that I recommend, if we go back to here, if you look up David Ogilvy, David Ogilvy advertising books, these are some of my favorite books. So there you go. That's a great book. David Ogilvy on advertising. Now you might think, oh, advertising, that has such a, like a, that has such a negative connotation to it. I don't want to be an advertisement person. Well, yeah, Facebook has kind of ruined advertising because it just fills it up with your feed and whatnot. But advertising, the heart of it is, is just really communicating what you have to offer clearly so that someone understands it so that potentially they will buy into it. So yeah, that's an answer to Carlos's question. Great, great question, Carlos. Um, so now we'll come back to, We'll come back to the chat. What do we got here? Thank you all for joining in, by the way. I really appreciate it. Um, Ken G recommends Alchemy by Rory Sutherland. Yes, I do have that book. Another great book. Alchemy, Rory Sutherland. So this is another book you should check out. The Alchemy by Rory Sutherland. Highly, highly recommend that book. Um, where are we at? So we'll come to the chat. Muhammad says, did you work out through your 5 days of fasting? If so, what did your routine look like? Um, I did, I did work out. You'll see this in the upcoming video, but, um, my routine was weightlifting and jiu-jitsu. So I think I, in the 5 days I did 3 sessions of weightlifting that was to preserve muscle mass. And I did jiu-jitsu also 3 times. That was just for fun. And because who doesn't like fighting and wrestling? I think every man deep down loves fighting. And so here we go. Um, well, back to Nefesimi says, I was wondering if you could talk more about the maths and machine learning book you posted earlier last month on your Twitter feed by Mark, uh, I'm terrible with pronouncing names, aren't I? Mark Deason-Roth, and whether you would recommend the Coursera offered by Imperial— the Coursera course offered by Imperial College of London now that the book has been published. Between courses and books, I would just go for both, like use both if you need to. Like for me, I'm a— I like to read and I like to watch videos. If I'm learning a new concept, like a brand new concept, I usually watch a video first because I like hearing someone talk about something really helps to sort of jumpstart my knowledge. But then once you get down into it, like there's no substitute for being able to read. So, I would say yes, try the videos, see if they work for you. But at the end of the day, if you want to learn something, you're going to have to read. That's where all of the knowledge is in books. All the knowledge that you need is probably in books and then the stuff that isn't is for you to discover. Um, all right, so where are we at? Atan says, morning from Bulgaria. Sajan says, yo bro, how are you? Um, okay, let's go back to, let's go back to Twitter. Um, so Oz says machine learning specifically when it comes to high variance data. Sam says, how do I improve my focus? I'm going to give this one a real quick answer. Sam, So how do I improve my focus? Practice. No, I can say something more than that. So it's seriously, for me, I had terrible focus, but for now it's like we said before, addiction cannot be gotten rid of. It just has to be, you have to change forms of it. So the focus is the same thing. You inherently have focus inside you already. So what you have to do is train your brain. So this is the practice to enjoy focusing on things. So phone, just a pure dopamine hit. Just cut yourself off from everything legit. Just, just make it a binary switch that you cannot, if you wanted to, you cannot access something else. You just have to sit down, set a timer and go, you know what, I'm going to do this one thing for 10 minutes. I'm going to do this one thing for 15 minutes. Once you get really good at doing it for 15 minutes, do it for half an hour. Once you get good at doing one thing for half an hour, do it for an hour. Then so on and so on. But always remember that you are human as well, so you might need some external stimulus. So if you're sort of in a funk, go for a walk or something. Leave your phone at home. This is what you're training your brain to do. You're training your brain to be in the present and being here and now. That's all focus is. So there we go. Sam, how do I improve my focus? Great question. Let's go back to Twitter. Um, Lakshan says video length under 10 minutes please. Well, that's not going to happen. And Connor, no, Oliver says he looks like a person that I know called Connor. Coming up with good projects to learn deep learning once you've gotten past the fundamentals. So Oliver, I'll put it here. So I get asked this a lot, like, what's a good project? And again, these are— there's no definitive answer here. Like, a good project is something that makes you excited and challenges you at the same time, whatever that may be. So there's no like, oh, this is the best project that you should do. You can, for me, if someone has already done it, like someone's got like, it's a pre-made thing. Like if you could go to Kaggle or whatnot and download a dataset and just work on that. Yes. Okay. That's, that's one way to do it. But for me, I personally, I like the end to end. Actually, there's an article that's in this month's Machine Learning Monthly Machine Learning Monthly, which is Data Scientists Should Be End-to-End. By the way, if you haven't checked out Machine Learning Monthly, go and check it out. It's a, it's a great collection of resources. I'm biased because look who writes that. But I'm going to do a video on this this week. But Where is it? That was September. There we go. This article. I'll talk more about this in a, in a video this week, but I like this article. So check it out. Data sciences should be more end-to-end. It just basically argues that rather than just dealing with a dataset and building a model, which is often like one of the smallest parts of a deep learning project, is go end-to-end. Now, of course, it depends We're on this question, by the way. It depends on what your goal is. For me, I like to, I like to do the full spectrum of things. I'm not interested in being a researcher, but that being said, even if you are building things, you end up researching things and vice versa. If you're a researcher and you find that there's something that doesn't exist that you want to research, you build something to help with your research. And so Oliver says, how do I come up with good projects to learn deep learning? I would, follow your curiosity and to stop like just having like 17 different options in your head. Choose one thing and try it out for like 6 weeks. That's why I created the article. Look at all these resources. So I think it's, uh, mrdburke.com/42days. Try it for 6 weeks. Explore something for 6 weeks and see how far you get. If you had the next 6 weeks to work on whatever you wanted, what would you make? Look at that, 42 days. So say for example, you've, you've done some basic deep learning courses. This is what I always recommend to someone is give yourself like a 6 weeks is a great amount of time because it's long enough to do something significant, but it's also not long enough to go if the whole thing turns out to be a complete failure, which spoiler alert, it won't be. Um, it means that you've, you've, it's not that much time. So 6 weeks, even if, even if it's a complete write-off, it's only 6 weeks. So you could just start another 6-week cycle. That's legit how I, I run things in, in my life is like I go, okay, what do I need to do for this next month or 6 weeks? And then I'll work on that and then reflect on the past 6 weeks and see if I can go, okay, what do I need to do differently next? That's how I'll run my business. That's how I run my, my startup. That's how I build things. So give that a try. Works for me. Might not work for you, but if you're looking for a good project, first of all, follow your curiosity. And then second of all, just choose something and see how far you can get. So Oliver, great question. Now we're going to go back to the chat. Okey dokey. Hello everyone. If you've just joined in, we're, uh, we are answering questions. We haven't done much of a dance yet. I am, I did a great workout this morning and I'm feeling pumped after this livestream. I'm going to go, going to go eat. So what question? Daniel Bourke, it is an insult to people who are watching your live and you're busy reading Twitter comments. Change your title to answer my Twitter questions. Um, time to unsubscribe. See you later. Thank you for the feedback. Gaurav says, hi Daniel, just started to learn machine learning. I'm happy that I took your ZTM course and I'm rocking it. So my question is what next? What should I do after you've completed the course? I would say work on something of your own. So, so if you've done the course, go on to either another course or create your own project based on the things that you've learned or create a resource. On your own blog or something that you've, you've shared. So for sharing what you've learned in the course. So what were your favorite things? What do you plan on doing next? So that'll help you really figure it out because I could tell you all these things, but they're going to be biased between my direction. So I would say the best thing you can do is test out your skills that you've learned. Now you can either do that through a project or by sharing what you've learned. So where we go here? Um, Indra Tej, ready? What's up bro? Why are you mad? This is what we do every stream. If you, if you're new here, well then at the start of the stream I said we're going to go back between the live chat and we're going to go to Twitter. That's how it goes. So I have to, I get asked a lot of questions and I'm trying to help out as many people as possible. So I answer the questions that people have been so kind to leave me on other places and I answer the questions that people ask in the chat. That's why we're here for. We've got plenty of time. I'm going to go back and forth. It's all right. Um, Easy says, what should I do when AGI takes over? I don't know. I don't even know what I'm going to do. Is AGI a real thing? Um, Swashstick says, what do you say about veganism, vegan and vegetarianism? I'd say there's a place, right? So I like this one. I like some nutrition questions. Swashstick, Swashstick says to do Where are we? I'll show you the question that we're answering. So this one is from the chat. So what do you say about vegan and vegetarianism? I would say there's a place for it. For me, I tried it and it wasn't, it wasn't, it wasn't enjoyable for me. And so I think, and I actually like this, the concept of random nutrition. So what I mean by this is that the kind of the world of nutrition right now is like just full of different dogmas. It's like, it's like worse than religion. So if you ask someone what they eat and they tell you this, they say, oh, you should eat this. And if you say, oh no, I eat like this, and go, oh, and just, just have all these different fights. You've got vegans, you've got people who eat meat only, you've got keto, yada yada ya. You've got fasters. I did a 5-day fast. I would argue that fasting is probably the most robust out of all of them. Because it's just, that's, fasting's like the ultimate like, fuck you to consumerism because you just, you're not eating anything. But my whole, I like the idea of random nutrition. So if you think back to when we were hunter-gatherers, the way we evolved, we would go days without food. We would go like 24 hours longer without food and then we would find some berries. And so maybe that day we would eat like a vegan. And then we'd go another stretch without food and then we would hunt an ox or something like that. And then that day we would eat only meat and then we might have plenty of food for a week and we're just, we're just eating as much as we can. And then the next week, oh no, all of our food stocks have run out. So, so now we have to go find some food again. We find an apple tree and I'm making this up as I go, of course, and eat it like that. So that's how I think we would have spent most of our time. And so where this comes back into random nutrition is like some days maybe eat like a vegan, some days eat nothing at all, some days eat meat only. And of course this is— there's not much science behind this, but I think if you were to look back on history and reflect back to how we would have eaten for the longest time, because we didn't always— supermarkets are pretty new, like they've only been around for like maybe 100 years, if that. And so this is where I was talking about before, the Lindy effect, is judge an idea's quality over how long it's been around. So there's many different diets, many different ways to eat, but I think probably the longest one that we adhered to was eating randomly. So some days nothing, some days meat only, some days vegan, and see how that goes. Now again, And don't, don't quote me on any of this. It's just a thought that I've been having. So yeah, vegan and vegetarianism. I've tried it, wasn't the biggest fan of doing it full time. But then again, I do have meals that are vegan. I do have meals that are vegetarian. I just throw them in wherever I feel like. So, boom. That's a great question. Love the— Love the question on nutrition. Let's go back. Here we go. Here. Um, here we go. Where are we at? Peter says, hey man, love your stuff and way of teaching. I'm new to ML and I'm in the geospatial/remote sensing field. I'm doing my honors. USC in drone imagery and ML for vegetation. Whoa, that's amazing. Hey Peter, can you send me an email with your work? Do you have any studies or whatnot? I would love to hear about what drone, drone, what's it said, drone imagery says for vegetation. So here's my email. Anyone can have this. Email me your questions if I haven't answered them in stream. We're not ending just yet. We'll probably go for another 3 or so questions, but Peter, I want to hear about your work. So there you go. You can email me. Anyone can actually email me. So I do take some time to get back to questions, but I promise I will do my best to get back to you eventually. If I haven't, email me again and say, Daniel Bourke, you're taking too long. 'Can you please answer my question?' And I'm more than likely to do it then. So that's just a heads up, but don't abuse that because if you do, I'll just block you. So yeah. Um, where are we at? We're going back to Twitter. Oh, okay. It doesn't matter, Peter. You can just email. If you don't have any work, that's fine. You can just email me and say hi or just, uh, just send me a paper that you've read recently because I'd love to learn more about drone imagery for vegetation. Um, let's go back to Twitter. This is what it's about. This is why we do these things because we help each other out. Um, okay. Heinrich. Oh, this is a great question from Heinrich actually. Heinrich says, Oh, there's another good one there. We've got so many good questions. This is why I feel bad sometimes because I can't get to all of them, but I promise we'll keep doing these streams and we'll keep answering as many questions. Remember, I don't have all the answers, but I'm happy to discuss things, uh, with, with what I've learned. Heinrich, if I could type, says to do Question is, you've made a great tool. Can you see that? Yes. You've made a great tool using machine learning, but now how to sell your product and make it a business. So see, this is what we kind of discussed before was, um, who was it? Carlos says, what other field outside of machine learning is worth studying that might not be directly related? So maybe you're really skilled at machine learning and you've created a tool using machine learning. How do you sell your product? So this is where that communication piece is always a valuable skill to learn because it's like math, like math and code you can apply to many different things, but communication you can also apply to many different things. And one part of communication is being able to sell. So sure, you might be a real big nerd like, like we all are if we're into machine learning and we're building these great tools using machine learning, but now how do we sell it? So that's why I think it's important to, to at least learn to sell in some aspect, right? And people might argue, oh, the product should sell itself. And I would agree with that. Build a product that's good enough to sell itself, but you should also be able to sell it too. So balance it out. So I want to give you a great example of a company that I really like that have done this. So RoboFlow. So I actually paid for their services the other day, so I bought it. So I became a customer. This is a machine learning tool for computer vision. So I paid, I think it was $158 or something to, to process some images, like 5,000 or something images. But this is their machine learning tool. Transform raw images into a trained computer vision model in minutes. So basically I'm going to butcher what they actually offer, but you can check it out here, roboflow.com. Basically it was I uploaded my raw images and some labels and then it preprocessed them into a format that I wanted. In my case, I wanted in them TF records. And so you're like, well, where's the machine learning in here? Well, this is a dedicated tool for preprocessing machine learning data. Now, you might be asking, it's like, oh, okay, well, what does it do? Well, sorry, what problem does it solve? And that's the main question you have to ask is if you want to sell something, the question that your customers are going to be ask you the most is what problem does this solve for me? And so the reason why I started to use this was because I landed on their blog. And what do they have on their blog? The ultimate guide to object detection. How to train YOLOv5 on a custom dataset. What the heck is an annotation group? So if I searched here, I think I must have searched is how to prepare images for TFRecord dataset, something like that. Who knows? RoboFlow might come up, might not. Anyway, I searched something like that and then RoboFlow's guides came up and taught me how to do it. And then I ended up going, you know what, their product makes it, this preprocessing data so easy. I might as well just upload and just have it ready and then I can just plug the dataset into where I need to use it. And so you might be going, okay Daniel, I'm hearing you. So what RoboFlow have done is they built a great tool for for ML, something to help you with your computer vision images. But now they're selling their product by not necessarily selling their product. They're helping me out. They're offering me a service by uploading or creating these guides on how to use their product. And so when I search for something, I stumbled upon it. You could call this search engine optimization, whatever you want. I stumbled upon their tutorials. Learned something and then go and then went, you know what, I can just use RoboFlow for this. So I would say is that you have to be the biggest advocate for your own product. And RoboFlow can back this up is because the founders are the people who write the articles of how to use their tool. So if you've built a great tool using ML, I would go on a relentless I would just relentlessly build things using your tool and sharing them online. So whatever the tool is, just make things, show people how they can use your tool because that's going to be the biggest barrier to entry, isn't it? If someone wants to purchase something new, they want to know, does it work? So show them how it works. That's my best advice for that. Great question, Heinrich. I'm going to like that on Twitter. Um, so now we're going, uh, back to the chat and all this one has 6 lines, so we might go up to here. Um, let's go back to the chat. We'll do 2 more questions. Chat. Um, Harshil says, hey Daniel, how did you conquer the maths, especially things like convex optimization and et cetera, the imposter syndrome? Um, this is a great question from Nefisimi. So this is maths and imposter syndrome. I'm— I haven't covered— conquered them, right? So it's— I think this is, this is what a lot of people think is like, there's going to be some magic moment where you're like, I know all the math, or some magic moment where you're like, I'm over imposter syndrome. And sure, you could switch it. Like, for imposter syndrome, I think you can switch it like that. But for math, it's like an ongoing, People have been studying math for decades. It's like this ongoing journey. You're never going to know it all. And so, whenever I come up to something that I don't know, I do what you'd think I'd do. I'd figure it out. I'd learn it. I'm not trying to— don't take, as I said at the start, don't take the fact that I can talk to this camera as though I know all of the math for machine learning, all of the code and whatnot. I still— I'm still learning these things as well. So if I come up to something that I don't know, I spend time figuring it out. There's no, no real magic potion for it. And the same thing for imposter syndrome. Imposter syndrome is either a lack of confidence, um, and usually a lack of confidence comes from a lack of skill. So imposter syndrome starts to fade away when you spend enough time improving your skill. That's what I've found. And so what that means is that once you've put in, and of course you're always going to feel a little bit of imposter syndrome. The reason being is because if you're feeling that, that's actually a good thing because you know the potential that you have to be better. So I'm confident in my abilities, but I also know how much that I don't know. So that's what you have to, if you want to develop that confidence. Put in the time, put in the practice at improving your abilities, improving your skill level. That's how you get over imposter syndrome. Okay, so we've got one more question. I'm going to put that one in there. That was a good question actually. I'm going to put that in the Notion so we know what what question we took after. Did this one about 2 minutes ago. Okay, so now there's one on Twitter that had a bunch of likes, so we'll kind of use that as a, as a metric to judge how good things are, which is terrible to say, but anyway. Boom. So we're going to copy this from Chris. Chris, and I think I might know how I'm going to answer this one already. Any advice for identifying a subspecialist area to go into? There are so many exciting areas in machine learning, it's hard to decide what to deep dive into and where the biggest future impact will be held. My particular interest is applications in healthcare. Hmm, this is a tough one. I'm not going to pretend that I know the answer to this, but let's just discuss it out. So Chris, I can relate to this because I have, I also have an interest in healthcare. Um, this is a question that we're on for everyone who's watching is you can, you can never really know. So if you're trying to know where the biggest future impact will be, Um, you can't— how good is humankind's track record of predicting? So most of the most important inventions that have come from the last century have been by accident. So if you imagine the internet, kind of, who would have known that the internet could be used for the things that we're using it for? Like, it was just to start off and being some sort of military communication system and now we're using the internet to talk from across the world. I'm recording a video and posting it live. Like this is just, when you really think about it, that's just kind of unfathomable. The laser, the laser was just someone trying to, to figure out if they could, I can't remember the exact reason, but the laser, um, when the person who invented the laser first showed their colleagues, that they'd invented a laser, um, they said their colleagues kind of laughed at them like, oh, why are you building— why are you making that? That's not going to have any use. And now look how many use cases the laser has. And so I think, of course, you could, you could try to go, you know what, what are, what are the biggest impacts? What are the biggest potential for, for impacts in the future? But at the end of the day trying to predict that is probably going to hold you back. And so my, the way I sort of go about these is again, I come back to first principles. This is why I really relate to healthcare is that what's gonna, what's required for human existence over the long term? Health, done. Energy, done. We need those. We need health, we need energy, we need food, we need, What else do we need? Help me out here. But those things, those are the things that I would say if you think real long, long term is like space travel, maybe that's something that we're going to do in the future. So who knows? So if you just think about that, it's like what are the basic requirements? And then of course if you try to, that's not really, you're not going to really specialize if you're just accessing that thing as a whole. Like you're like health or education, we're going to need that. This is how I think of it. And I'm not going to, I want to answer this question in a problem-solving manner rather than just a, here's my straight answer. Okay, here are the things. It's going to be, we're going to, here are some big fields. Education. Yep. We're going to need that. Health. We're going to need that. Yep. We're going to need energy. So they're probably my big 3. I can't think of any others. But if you wanted to subspecialize in something, I would say the best way for me to do it is to figure out what I'm already curious, already interested in, and then just follow that interest. Why? Who knows if it's going to be the biggest thing, but you never know if it's going to be the biggest thing. But if you're following your interest, chances are you will develop some skill in that particular thing. And then you will be even more interested in it. So if you're a bit of a machine learning generalist, that's the beauty of machine learning is that it can be applied to many different fields. I would say just try to apply it to something that you're already interested in. For me, I'm interested in— I love health. So I want to see how machine learning can be used in nutrition or why not. And then that being said is that don't be afraid. What's the number one rule of machine learning? Is to don't be afraid to not use machine learning. So you might go into a field and figure out, it's like, okay, you come from the, 'cause remember with a person with a hammer, everything looks like a nail. That's what happens when you learn some machine learning. Every problem looks like, I can solve this with machine learning. There are a lot of problems that don't require machine learning. So I would say is that yes, go into some sort of subspecialist area, but just be aware that even though you're trying to go in there with machine learning, the problem that you're trying to solve may not require machine learning. I'll give you an example. When I was at my last, last, uh, machine learning engineering role, we were working on machine learning in healthcare to speed up doctors' appointments, but no one ever asked. So the way we would speed it up was that we would use machine learning to transcribe the doctor's notes. So voice recognition, and then say the doctor was talking to a patient, would transcribe those notes automatically. That's what we do in a, in a, in a doctor's surgery. So one-on-one appointment, microphone there, it's getting transcribed using machine learning. Okay, that's well and good. And what was the goal there? The goal was to save around 5 minutes per doctor's appointment so that a doctor at the end wouldn't have to type up the notes that they discussed with the patient. But what problem were we actually solving? Yes, the doctor has to type up the notes. Yes, they could solve, save the extra 5 minutes. But the goal there, or the reason why we're trying to solve that, is so a doctor could see up to 1 to 2 more patients per day. And now, is that what we really wanted to do? Like, do we just really wanted to— is that the real problem there? The fact that doctors have to write notes? Of course you could use machine learning to help that. But do we really just want to keep feeding more people through the doctor's surgery? The fact that— is that what a doctor is? Just someone who's on a conveyor belt and just like, yep, patient. Yep. Boom. And again, I don't have any practical knowledge in a doctor's surgery, but that's just how I was thinking. So yes, you can use machine learning for that subset of health, but really think about things in like a, what is the actual problem that we're trying to solve here? Why does someone go to a doctor in a, in the first place? No one asks that question. Why does someone, seriously, it sounds like a dumb question, but to me, I ask that question. Why does, why does someone even go to a doctor? And I think you'll find most of the time is they want someone to be compassionate. So if you're just, you're a doctor and you're just trying to shuffle people through, is that really providing like the service that you spent 8 years studying for? And so that's how I think about these things. So if you want to go into a subspecialist area, just of course have some machine learning skills. Yes. See if they can be applied to some sort of problem that's in, in something that interests you, but then also be aware, don't be afraid to step back and look at the bigger picture. What's the actual problem you're trying to solve? Of course, machine learning can be used on little, on, on not, I'm not saying little, but like niche problem subsets. But again, what's the larger picture here? That's probably how you could look at something that will have the biggest future impact. All right, that one went on for a little bit, but let's, uh, it's time for your boy to eat. So thank you so much for tuning in. And Matthew says, when's the free pizza back on at the precinct? Haha. I remember that back at the UQ days, University of Queensland. Let's bring this back up here. So yes, Matthew, you're completely right. That 5 minutes of typing was also the 5 minutes the doctor could be invested into their patient. So Sai says your machine learning roadmap 2020 is the best. What drives you to be great in this field? What are your motives? Can you please share with us? My motives are just machine learning interests me and I get excited by it and I'm following that interest at the moment. That's, And if I can share what I've learned and share my excitement, share my energy with, with you guys, well then I'm, I'm happy to do so. But we're going to wrap this up. So thank you all so much for tuning in. If you have any other questions, feel free to email me. My email is daniel@mrdbourke.com. You can find that on my website fairly easily, um, or leave a comment below and I'll get back to as many of them as possible. But as always, I hope you had a fun time. Keep dancing, keep learning, keep creating. We'll do this again next month. I'll see you later.