SDS 232: Sleep on it

Welcome back to the SuperDataScience Podcast, ladies and gentlemen, super excited to have you back here on the show. And today I wanted to share a story, an interesting story that happened to me a few weeks ago. So this was the end of December. And what I was doing was recording intuition tutorials for a section of the Artificial Intelligence Masterclass course, which we released not long before that. And the section is called Mixture Density Networks combined with Recurrent Neural Nets. So MDN RNN for short.

And what I found was that I was actually struggling with this section with this specific, with these specific intuition tutorials. And so what I normally do with Intuition tutorials is I go away, and, so I pick a topic and it doesn’t matter which course it is, whether it’s Artificial Intelligence A-Z, Deep Learning A-Z, Blockchain, Machine Learning.

For those of you who’ve taken our courses, you probably know by now that I record the intuition side of things, intuition tutorials, and Hadelin mostly works on the practical tutorials, with the coding. And I personally really enjoy intuition tutorials. I love digging into the, understanding a topic very well and then being able to explain it.

And so what I normally do is I’d take a topic, whichever I pick, whether it’s Neural Networks or Self Organizing Maps or whatever it is. And I work on it for about a week. So I collect all of the materials I can find on it. I look for research papers, I look for videos, blog posts, comments, ideas, suggestions, practical applications and lots and lots of materials. I go through them, I read them, takes me days and days and days. I always think it’s going to be faster. I always think is going to be a two day process or maybe I’ll get it done in a day. But actually on average takes about a whole week.

And so from that, then I process all that with my mind. And as a result, I’m able to record between 15 minutes of content to maybe 90 minutes of content. So as you can imagine, a week of work goes into about one to one and a half hours of video content, maybe less. And, so basically it’s a condensed and processed version of all the things that I went through, all the things that I researched and I pick out all the important things and I put them in the right sequence so that it’s easier to understand. So that’s the process I go through with intuition tutorials. And specifically here, what I found was that I thought it was quite a short topic, because I already knew how recurring neural network, and I’ve already explained that.

So now I only needed to add this one small component, the mixture density networks. But what I found was that instead of taking me the one or two days that I actually thought it would, I got to a point where I hit a roadblock and I was trying to push through it. I was trying to really get, get through it, get past that, whatever I was not understanding as I was… I set myself a deadline. I was like, okay, it has to be done today. And I spent, I spent a whole day doing that. I couldn’t push through the roadblock. There was this one component, one part of the MDN RNN that I was just not understanding. So then I went away and a few days later, I did that again for a whole day. And again, no results. And a third day again no results.

So I just kept trying to push through it and I spent, I just threw three extra days at it and it wasn’t giving me a result. And so what happened on the third day was that instead of pushing for the whole day, I decided to myself, okay, I got to this stage instead of pushing and pushing, pushing and then be disappointed and then being put off this topic for an extra a week or maybe a couple of days, I’m not going to look at it because I have other things to do. What I’m going to do now is I’m going to stop, I’m going to go do my other things that I’m interested in and then I will sleep on it and I’ll see what ideas I come up with in the morning. And that actually really helped. And from there I was able to the next morning wake up and I could see it from a different perspective.

Somehow, I don’t know how, I got these ideas in my sleep and I could understand the topic well now. And from there I recorded the tutorials and everything went super well. I don’t remember the exact details on whether I stopped towards the evening or at mid day. This was about a month ago. But the point is, I made the conscious decision to stop and sleep on it and then, attack it again from the next day with a fresh perspective, with a different mindset instead of putting myself into the same dead end every single time. And so that’s the power of sleeping on it. And actually it’s a good idea to think about this topic right before going to bed, to actually finish your day with that problem that you are struggling with so that your mind can work on it at night.

And in preparation for this podcast, I actually looked at some examples from around the world of when this had happened to people before, and what kind of inventions came out of this. And actually it’s quite surprising. There’s dozens of really powerful inventions that have come from people sleeping. So here’s an example. You know, the sewing machine. Have you ever paid attention to the sewing needle on the sewing machine? Well, the needle has the eye of the needle, the little hole in the needle in a sewing machine is towards the tip on the needle. Whereas in a normal needle, the one that you use for sewing with your hand, the tip of the needle or the hole in the needle is towards the back of the needle, towards the other end. And so guess what? That invention was made in 1846 by Elias Howe in the US. And, yeah, that revolutionized the whole industry and that’s how sewing machines came to be. Just by putting that hole of the needle on the other side, that’s what was required.

That’s what gave this breakthrough to sewing machines. So that’s one example. Another example is, 1869, we probably all know this from school. The periodic table of elements was invented. Well, what not everybody knows is that it was invented by a Russian scientist, Dmitri Mendeleev, and he invented it in his sleep. Again, he was working for three days nonstop. And then he went for a nap, slept for 17 hours. And in that, in those 17 hours, he had this dream of the periodic system of elements, table of elements.

The movie Terminator was actually a nightmare that James Cameron, the director had while he was suffering a fever. And so that turned into Terminator and that turned into a whole sequel of movies. So as you can see in many different areas of our work of where humans work, sleep has played a role.

In fact, Thomas Edison, the famous inventor person who gave us electricity was… actually had this habit or method where he would take naps. But while he was falling asleep, he would hold two steel balls, a one steel ball in his hands. And as you fall, he would fall asleep, the steal ball would fall, and there was a big metal surface that it would fall on. And so it would wake him up. So he’s doing some work. Then he decides to take a nap, takes the steel ball. And as he’s falling asleep, he drops the steel ball. And instantly in that moment he starts analyzing what ideas he had because he was tapping into his subconscious. So that moment when you’re falling asleep, that’s usually where all the creativity, all the ideas apparently happen. And so that’s when he was tapping into his idea. That’s how many of his inventions or ideas come to life through that process of actually tapping into subconscious, that moment when he was falling asleep.

So just puts it into perspective that indeed a lot of different inventions and a lot of creativity has actually happened in people’s sleep. And so in data science, we need to be creative. There’s oftentimes when we have problems, in fact, our work is all about solving business problems and coming up with creative models or solutions and insights from analytics. And it all requires creativity.

And, so whenever you hit a roadblock in data science about how to do certain thing, or maybe you need some fresh ideas, some fresh perspectives, try not to, or maybe stay away from the habit of just pushing really hard to get a result then and there. If you find that you’re hitting a roadblock, maybe it’s a better idea to step aside. Take a step back, pause and let your mind relax, work on something else, and then before you go to bed, actually, think about that same problem for another half hour or so that you were trying to solve during the day and then sleep on it. And see what comes up in the morning. So there you go. That’s a quick maybe hack and hopefully, it’ll help you one day solve a big data science problem or a roadblock that you might hit maybe on this weekend. Think about where in your career you have hit a roadblock or what kind of problem that you have you tried to solve and you couldn’t, and try to apply this method. See what you come up with.

On that note, thank you so much for being here. I look forward to see you back here next time and until then, happy analyzing.