Heye.Earth
Heye.Earth
“Make Machines Learn” Explainer

“Make Machines Learn” Explainer

‣
Menu

How to solve problems

Fiction

Other Projects

DEEPWAVE

You can comment on everything. Or give me (harsh) personal feedback.

Made with 💙 by me (how to).

Dedicated to my family.

How to solve problems

Fiction

Other Projects

DEEPWAVE

You can comment on everything. Or give me (harsh) personal feedback.

Made with 💙 by me (how to).

Dedicated to my family.

“Make Machines Learn” Explainer

Status: Draft, never finished

In the fall of 2022 a few of us at

gathered for an ML Ultralearning sprint. In the end, I wrote down a summary of what I learned.

Using only the 1000 most common used words.

image

Hello.

This is everything I learned during the last 10 days. Three friends and I wanted to learn really fast how machines learn (really fast). I will tell you what we did, but I have to follow three laws:

  1. I am only allowed to use the ten hundred most used words

    2. (like the person who explained things this way and is funny)

  2. I can use define → making up a short word that stands for more words
  3. I believe that you have a new computer and know nothing

    4. (like the guy named Snow)

At the end of the 10 days I build this:

(Insert working Streamlit version)

If you want to do the same, you just have to read a special form of paper. What makes it special? You can play with it!

Let's begin.

super-embed:
<orbit-reviewarea color="blue">
      <orbit-prompt
        question="What does it mean to define something?"
        answer="making up a short word that stands for more words"
      ></orbit-prompt>
      <orbit-prompt
        question="What's Setup?"
        answer="Making your computer do things it couldn't do before"
      ></orbit-prompt>
      <orbit-prompt
        question="What's a Code?"
        answer="Language your computer speaks, so you can tell it do things (like learning!)"
      ></orbit-prompt>
      <orbit-prompt
        question="Visual Studio Code/VS Code"
        answer="A 'doing-things-place' where you can 'see' your 'code'"
      ></orbit-prompt>
      <orbit-prompt
        question="Python"
        answer="A choice of which special language we want to use (there are many)"
      ></orbit-prompt>
      <orbit-prompt
        question="Paperspace"
        answer="A machine that's better then the computer you are using"
      ></orbit-prompt>
      <orbit-prompt
        question="Github"
        answer="A place where we can share what we wrote to our computer (so that others can do the same, and we what they did)"
      ></orbit-prompt>
      <orbit-prompt
        question="Remnote"
        answer="The one you are using right now. It makes you remember!(You can even write on it, if you sign up)"
      ></orbit-prompt>
      <orbit-prompt
        question="Jupyter Notebooks"
        answer="One that you can play with. It makes some parts of the words so that your computer can do stuff with it!"
      ></orbit-prompt>
      <orbit-prompt
        question="Algorithm"
        answer="tells anyone who is interested in which order to do things. Like, when to put food into hot water and how long to wait before it's good."
      ></orbit-prompt>
</orbit-reviewarea>

Setup ↓

  • Making your computer do things it couldn't do before.

So what do you need to make a machine learn? Well, a machine. You are reading this, so you have a machine. Good. You are half way there.

Next, we need some way to make it learn. Why? Because then it can do cooler things. Like this art piece:

I hope this is enough to make you really want to do the same. If not play around with "In the middle of traveling" or "Art maker, but written like a loved clean up machine".

Humans learn really well if they know why they should learn something. So if I say "Machine Learning" your brain should make those fire up-goers we use when the earth went around the sun once. I will spent little time on making you excited, so please do it yourself.

Done? Good. Then lets write some code.

Code→Language your computer speaks, so you can tell it do things (like learning!)

We need:

  1. A "doing-things-place" where you can "see" your "code" <- Visual Studio Code, VS Code
  2. A choice of which special language we want to use (there are many)→Python
  3. A machine that's better then the computer you are using <- Paperspace. (Not really needed, but you will run into a lot of problems if you are using your own machine and don't know much already. You can also just use Paperspace and not VS Code)
  4. A machine that's better then the computer you are using←Paperspace

    5. 1. (Not really needed, but you will run into a lot of problems if you are using your own machine and don't know much already. You can also just use Paperspace and not VS Code)

  5. A place where we can share what we wrote to our computer (so that others can do the same, and we what they did) <- GitHub

Instead of words on boring paper we will also use two special forms of paper:

  • The one you are using right now. It makes you remember! <- RemNote. (You can even write on it, if you sign up)
  • One that you can play with. It makes some parts of the words so that your computer can do stuff with it! <- Jupyter Notebooks, also called the Notebook.

Let's catch them all!

Paperspace

Machines have many small parts that can break, and making them do things is very hard. But many people have found some ways to make machines do some things. They wrote those into books for computers, or rather boxes of paper with code on them. You can get those from a book store and let your computer read them. But these books are confusing! If you open the wrong boxes together everything breaks. And there a new ones coming out all the time. So instead of starting from nothing, we will make a phone call to a computer that has already read a lot. It's number is Paperspace.com. Call it now (don't worry, it's free). You see it?

image
  • Touch the "Sign up free" field. Then create a way that they can remember you. While doing so, they will also ask to call you back, so they know you are real. Next, follow these steps:
    • image
    image
    image
    image
  • Start from nothing, or you will be very confused! Also you can pick how good your computer should be. The free ones are enough.
    • image
    image
  • Finally. It's time to start your Notebook!
    • image

The Notebook

The next step will be to play with this Explainer. For this I wrote a second book, as a Notebook. I can send it to you here:

It will tell you how to get the Notebook on your computer.

Before you leave, a few words about the other helping things:

RemNote

You have been reading this in RemNote. Remember, that I said this special kind of paper helps you remember? It does!

The way this works is that all the things we have defined have all been turned into memory cards! A special algorithm will tell you when to look at them!

  • algorithm→tells anyone who is interested in which order to do things. Like, when to put food into hot water and how long to wait before it's good.

So, sign up, and view this page every day, at least for as long as you are working through the Notebook.

Always remember to do your remembering!

GitHub

You have already seen it! I will write a Git Explainer soon.

VS Code

You don't have to do this, but you can get Visual Studio Code, and bridge it to GitHub and Paperspace. It makes it easier to work with your own computer later. Just touch these Blue Words for a How-To.

Now finally. Let's play, by

Sending our Notebook to Paperspace:

  • Go to Paperspace
  • Touch this
    • image
  • Open the Notebook
    • image

Well. That was easy. See you over there!

  • Words I have defined so far:
  • define
  • Setup
  • Setupc
  • Code
  • Visua lStudio Code, VS Code
  • Python
  • Paperspace
  • GitHub
  • RemNote
  • JupyterNotebooks
  • Notebook

All definitions from the Notebook

Other Explainers

VS Code Explainer