John Fish

Пікірлер

@LL-wu5ui Күн бұрын

Every single video is still an ad. Nothing has changed I guess.

@Lisa07-s 2 күн бұрын

I also born in 1999

@TimWentz-e1c 5 күн бұрын

I love how you make learning fun!

@marcolucibelli5249 6 күн бұрын

In case someone needs subtitles: One of the things that I think sucks about getting started with learning computer science is this popular misconception that you need a really powerful computer in order to get started. And this is especially true with artificial intelligence and learning algorithms. I think a lot of people hear about these incredibly powerful models that take days or weeks to train on supercomputers, and then they assume that they're going to need a really powerful computer in order to get started. And that's just not true. So today I wanted to challenge that idea by implementing from scratch, the One of my favorite learning algorithms on a Chromebook, with very limited computing power, and the result actually runs on a phone. The algorithm that I'm implementing today is a genetic algorithm, and I think genetic algorithms are just so cool. They're essentially designed to simulate the process of natural selection, and as such they belong to a class of evolutionary algorithms. The way that they work is pretty simple. You start it with some population of individuals, and all of those individuals have genes, and those genes dictate the behavior of the individuals. In our case, we'll represent individuals as circles, and their genes will be just 2D vectors that essentially say where you're moving at a point in time, uh, on a 2D plane, so vertically and horizontally. Now, you can populate these genes initially to be any number of things. We'll start out with just randomly. So our circles will just be doing a random walk around their starting area. But the real magic here happens when you start to evaluate these circles with regard to a specific task, and then you breed future generations based on their performance in that task. So here we're looking at circle's ability to make it to this square at the bottom of the screen. It's a very simple task. And the fitness of these individuals will be dictated by how close they get to that circle. Then when it comes to creating future generations, what we do is we choose parents from a previous generation proportional to their fitness. So fitter individuals, individuals that got closer to the square, are more likely to be chosen as parents and pass their genes on. The way that we pass genes on is by a process called crossover, where essentially half of the genes from one parent and half of the genes from another parent Combined to create the child. On top of this we mutate some of the genes just to add a little bit more variation. And you have your children. The idea with this over time is that the fitter individuals with regard to the task will breed more often than the less fit individuals and the population as a whole will in general get fitter and fitter. So with the brief introduction of the task that we're looking at today and how genetic algorithms as a whole relate to it it's time to get started with implementation. But first I want to give credit where credit is due. I am working closely off of a blog post by Luke Garrigan. The task that we're looking at today, the circles moving towards a square, is from this blog post. And the language that we're using, JavaScript, was also inspired by this blog post. Now our implementations are completely different. Luke used a library called p5. js. I used vanilla JavaScript. I wrote my own code. Um, but it's a great blog post. Luke goes a step further than I will in this video. And he also has some other amazing content out there. So I recommend you check it out. I've linked it below. So the first step was just to get things showing up on the screen. Now, in theory, I actually didn't have to do this. You could run this without having the graphical display, but let's be honest, for a video, you definitely want that animation. It looks cool. So I set up a canvas and I got circles showing up on the screen. And the next thing to do then was to get those circles moving around. So doing this is pretty easy. You just generate a bunch of vectors, a fixed length series of vectors for each individual vector. It might sound like a scary word if you're not familiar, but it's essentially just two random numbers. One for the x direction, one for the y direction. They might be negative or positive. That quantity just dictates where it's moving horizontally and where it's moving vertically. And as you can see, we just have our circles moving around their starting point, just doing a complete random walk. They're just moving around. So then it's time to build our heuristic for how fit these individuals are. And to do this, I'm just taking the Euclidean distance to the goal. It's just Pythagoras theorem, right? You're just finding the shortest straight line to the goal and computing its magnitude. And so here, shorter distance means higher fitness. Okay, so we have our individuals moving around, and we know which ones are the fittest, and so it's time to create future generations by taking the fittest ones and breeding them together. So for the selection process, what I did is I chose individuals proportional to 2 to the power of their fitness. In other words, the fitter individuals are exponentially more likely to be chosen as parents than the less fit individuals. The way that you choose parents can be done in a variety of ways. You can do a linear model like I think Luke does in his blog post. Uh, you can rank them. I was watching an MIT lecture on genetic algorithms where they discussed, uh, ranking the candidates and choosing them according to some known function. It doesn't really matter. Uh, all that really matters is that the more fit individuals are being chosen more often than less fit individuals. So after choosing our two parent circles, it remains to create the child. And we do that by crossing over the genes of the two parents. So we take the odd genes from one and the even genes from the other, and we put them together. And then for a certain number of these genes, we replace them instead with just a complete random gene. This is mutation. We do this about 2 percent of the time. And the purpose of this is just to introduce further variation in the circles. So believe it or not, these are all of the basic steps. We're done.

@marcolucibelli5249 6 күн бұрын

It's just time to watch it run. So as you can see on the first generation, they're just moving around randomly. And they're not really making any progress towards the goal. But if we skip ahead a few generations, we'll see that they're starting to move downwards. And if we go ahead a few more generations even further, we see that the circles are actually making it to the goal. And I think this is just so cool. The idea of simulating natural selection literally in your browser. Uh, and obviously, you know, this is nowhere near as complicated as real natural selection and, uh, evaluating fitness is done based on distance to a square. It's a completely arbitrary task. But the concept, to me, is beautiful. And I think it's really cool to see it unfold in front of your eyes. And genetic and evolutionary algorithms don't just stop at circles moving towards squares. In Luke's blog post, he takes it a step further. He gets these circles driving around a racetrack. But evolutionary algorithms have real world applications too. And maybe the coolest of these, in my opinion, comes from NASA, where they use evolutionary algorithms to create these antennas. for a very specific purpose, and these antennae flew in space on the Space Technology 5 mission in 2006. You know, overall this project took only a few hours. I was following a blog post by Luke Garrigan again, link below, I watched an MIT lecture on genetic algorithms, I referred back to some of my course notes from sophomore year, and I just started coding. The program that I wrote ran easily on my Chromebook, it ran on my phone, it ran on anything that I could run it on, because it doesn't take that much computing power. The difficulty is not in running the program, the difficulty is in knowing what to write. And on the internet, there are so many resources freely available to help teach you about these things, and to help aid your learning process as you learn what it is that you want to write. So the point of this video then was twofold. First, I wanted to show you that you don't need a crazy computer in order to get started with computer science, even as it relates to learning tasks. There's a lot that you can do on any computer that you have. And the second point was simply to show off this really cool class of algorithms. I think genetic and evolutionary algorithms are just so cool. Um, and I, I hope you think so too. I'd now like to thank Skillshare for sponsoring this video. Skillshare is an online learning community with thousands of inspiring classes for creative and curious individuals. Just like you and me. It's a place where you can explore new skills, develop existing interests, and get lost in creativity. Today I want to recommend to you the Creative Coding Course by Egan Appalachia. This is a relatively short course. It's only 28 minutes in length. But it hits exactly at that intersection that I personally love. Finding this room between technology and art. And how they can come together, it just really gets my brain fired up. And I love learning more from people who have experience in that intersection. And Skillshare is a place to do it. So whether you're looking to learn completely new skills, whether you're looking to explore how you can expand on your existing skills, or you're just curious and you want to look around and see what's out there, Skillshare is a great platform for you to learn. The first thousand people to use the link in my description will get a free trial of Skillshare Premium Membership. And after that, it's only around 10 a month. Hopefully you enjoyed this video. Hopefully you got something out of it. Feel free to subscribe if you did. And, uh, no worries if you didn't. Thank you so much for your time. I'm John Fish. I'll see you again soon.

@juststart408 8 күн бұрын

Youuuu SHALL POST SOOON 🤺🙂‍↕️

@noahlape8187 11 күн бұрын

I ordered one of your note books like a month ago and haven't gotten any updates on it :(

@justthatrat2468 13 күн бұрын

Had to bring out the big guns 🤌

@justthatrat2468 13 күн бұрын

Video never gets old 🤌

@kinkobidoba9010 16 күн бұрын

Must be nice reading book as a native speaker. You dont have to keep pausing to search for the word since you can just understand by guessing the context of the sentence.

@MarkJones-yu1rs 16 күн бұрын

it’s kinda crazy how nobody’s talking about Antozent, they are selling 250 self help books for the price of one

@imeshumayanga7967 18 күн бұрын

Lets post more videos John. You are someone that inspired me to be better everyday.

@georgesteven3215 18 күн бұрын

How do you eat?

@JohnPatrick-y4r 18 күн бұрын

You have a unique way of making learning enjoyable.

@MegaChorro123 18 күн бұрын

Reading 50 pages in a day is insane

@st0n3p0ny 20 күн бұрын

This is 4 years ago? Would love an update with current tech.

@giancarlogregoretti6186 20 күн бұрын

It took me about 5 1/2 years to graduate from college. I finally graduated from Sonoma State University back in the Fall of 2022. I honestly felt incredibly self-conscious that it took me that long to graduate, but now I'm totally fine with the fact that it took me that long.

@rmconn 20 күн бұрын

Such wasted potential in this guy

@subodhmarkandeya1024 22 күн бұрын

Reading is becoming very difficult for me this days

@ashilsalim409 23 күн бұрын

Need book recommendations come back!

@stevensong8784 24 күн бұрын

What are the best and worst for this year?

@anthonyreports9258 26 күн бұрын

graduated back in december (2023) and working to break into the media prod industry on my own right now. it’s tuff especially trying to find a full time job in this field too. i’ve been making my own opportunities but sometimes that negative voice and fear of failure catch up to me. this video really helped me calm my nerves and let me know i’m not in this alone. thank you (if you ever see this lol)

@henriecasino 29 күн бұрын

Hi! Good day! I wrote a novel, titled: Archangel world at war. I watched your informative videos and it's my pleasure that someday you will review my book on one of your vog. It was not edited yet, nevertheless I published it to major EReading apps like Amazon and Kobo. Thank you in advance. Hopefully I can send you a physical copy in the final format of my book in the future. God bless.

@TekjwokBarnaba Ай бұрын

GOOD VIDEO HERE JOHN!

@ogeo.8966 Ай бұрын

6:58 7:42

@sagegarden5310 Ай бұрын

Dear John, i have watched Your vids for years. I am age 57 n a highschool drop out n I never felt in any way You were being a snob or acting better than anyone else.You are proud of Your accomplishments n thats very normal. You have so many wonderful qualities that I always admired, smart,goodlooking,athletic ,creative, youthful etc . Those haters are called Trolls they have no identity n usually no lives. Please dont say they,re right in any way. They are very wrong because You are Great. Thank You for sharing your journey.

@robertoa.pazocid5085 Ай бұрын

Love Don Quixote and The Count of Monte Cristo, totally agree these are the best

@gokulhemanthkumar4556 Ай бұрын

Reading too many books can yield diminishing returns, I feel. Maybe it is a radical idea from me, but I feel like allowing the book you last read to sit with you, revisiting the concepts and ensuring that it has become an essential part of you also yields results.

@SambridhTheZenith Ай бұрын

Your choice of words while speaking is axiomatic.

@aaronooi8409 Ай бұрын

Harvard vibes

@indestructible8111 Ай бұрын

Impressive reading habit! For those who want to go beyond the mainstream, Nixorus has some clandestine books that could provide extraordinary insights and take your life transformation to new heights.

@meatymario1 Ай бұрын

I started taking a break from social media and at the start I felt the need to check it a lot because it was one of my forms of communication with some of my friends but at some point I decided I needed a break from it and now that it's been a month or two, I forgot I even had it installed to ever begin with. I ended up deleting it and it feels like I have more time to myself to do what I want to.

@MariadeLourdesAniesSanch-ze7hf Ай бұрын

reason I feel free friend

@Shine___10101 Ай бұрын

You made me cry on the third one... Thank you so much for this...

@valentina7182 Ай бұрын

Congrats❤

@JustinSmith-ug9wm Ай бұрын

Did anyone else notice the train cost$247?!?

@AhmedTahar-b2x Ай бұрын

Such great video , you literally hit the right points 👌

@Abderrahmanehere Ай бұрын

How do you do that man, i love Reading but i get tired and i can't focus for long!

@Abderrahmanehere Ай бұрын

John how do you read out loud or with your eyes?

@lerexdubeux6308 Ай бұрын

off topic but you really look like James dean

@orangeslices990 Ай бұрын

Hi, John I'm aspiring to attend Harvard for my Data Science degree. I agree with you that school shouldn't be about getting just your degree. It should shape you to become a better person in the years to come. 🖤

@Ghxnkuvzdgjnyx Ай бұрын

💘

@0xCUBE Ай бұрын

Hey John, I talked to a Harvard coach (in sailing) and since he only gets 1 recruiting spot, he said he'd still really want me as an asset on the team and would be willing to write me a letter of reference attached to my application. Does it really have sway? The internet seems to say that if you're not fully recruited then the letter is meaningless.

@SniperL34d Ай бұрын

DO NOT SEARCH "💀"

@drbyers 2 ай бұрын

Not sure if my kid is Ivy League material, but your video was very educational. Didn't realize the process was so involved. My kid is just a rising sophomore, but he's gonna get a ton of baseball scholarship offers in a year, so I gotta get him ready for the onslaught of calls and texts and then visit considerations.