Cloud Native Computing Foundation KubeCon + CloudNativeCon China 2019 (Shanghai), 5 Jul 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: How I Teach my Second Grade Kid what AI is and How this Relates to BI that Impact... Cupid Chan

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

How I Teach my Second Grade Kid what AI is and How this Relates to BI that Impact... Cupid Chan

One day, my kid asked me: “Dad, what are you doing at work?” That struck me because being an AI and BI practitioner for about 20 years, I have helped clients on a daily basis, but my how do I explain it to my kids? Well, I didn’t tell them because I thought they would not understand even if I tell them. But that is MY problem. As a consultant, it’s my job to explain to my “clients” so that they understand my proposal, right?! So, I decided to “teach” my kids AI, in the language that they can understand. It turns out this is a very good exercise to help me explain this concept to other people including non-technical executives, as well as technical geeks. It also provides a good framework for me to tie this back to a long-standing BI technology and help my client identifying cases to apply AI/BI in their business.

https://sched.co/NrsC

A

Disclaimer this presentation session was prepared or accomplished by Keuka chen in his personal capacity. The opinions expressed in his talk are the author's own and do not reflect the view of any United States government agency or Department good afternoon.

B

Okay, give me some feedback good afternoon. Okay, thank you very much. You know what just this section I actually have two big challenges, number one. This is the last presentation for the whole summit, so I need to keep you here. So this is.

B

Chinese English.

B

B

I'm, the CTO of index analytics one of the small business in United State focus on the.

B

B

This is actually a pack-off dies. Okay, so then, okay, of course, I also get this I also get this for me, I'm, not sure. Where days it looks like me, but they're just me.

B

Now this is so tricky one, and this is the hardest one.

B

This is of car right, so, for example, if I flip it, you tell me to stop so okay. So how many to keep in keep it right here when you pick it to the yourself: okay, you're, on okay, one, this time, kill it yourself and stop okay. So there are three new cards: no, what is the percentage.

B

B

B

Each other, because that's basically the foundation of why I are so powerful, the pop the ballistic, the problem, of course, when you compare that, together with matrix, manipulation, calculus differentiation, integration, very simple, something like this: this is the choo-choo for a machine, and one of them can you guess this is very complicated.

B

Conference and to a speaker then I realize you know this probability and relying heavily on probability and statistics. I come from us Washington DC.

B

In Pennsylvania is called Hershey Park and then for those who this is just another theme park like Disney. They have characteristics like this I think China also have this Hershey chocolate right and then there are wise. We look whole stop for people to join, and then there was a line for the Rye and then suddenly my son, a second grader kids, ask me.

B

Well, very good question you know at work. I will first of all data all the data using ETL process and then trying to apply some line your face and try to pop the data. If the data is not really that satisfying, then I will apply Network and then hopefully there is I will be better. Do you really think that's how I explained to my son? What is a I, of course, because he is not my audience to understand and.

B

B

Perception when they I I further explain. Do you still remember? How did you pinch your younger sister, a pan versus an apple? So basically we will bring a pan and then you say, hey a pen and then we bring a monopoly. You say this is an apple and hopefully, if I keep to do this younger sister, what is the pen and what is an Apple right? So why? Because she started doing amazing the wet one thing for that is, and the more pointy thing it's a pencil pen.

B

This is supervised learning because.

B

Here the pointer wrong word, no visible feature in machine learning and Apple and pen. They are the label for machine learning and then my son said: oh I think this guy is actually talking about.

B

Can help my song a second greater, to remember or understand what is supervised, learning I would just let it be.

B

Supervised learning has a lot of application. This is one of them. So, instead of a baby or a kid trying to remember what is a pen and an Apple is actually a computer, we place indicate the pen and the Apple is we're pleased by this kind of or MRI scan. Who is the parents horse to pick up whether it's the doctor, the doctor, you tell the computer?

B

What is this cell is that the cancer cell, or not so through this repeat process, the computer will start recognize what is a cancer cells, so one of the application for machine learning trying to be the expert to explain to you what is the layer of entropy or second or thermodynamic.

B

Case it is actually saying that unless you put extra ology to keep everything in the state, things were just going to be Messier and Messier. So that's basically it a very high-level explanation of law of entropy.

B

Now, if we apply to a kids playground, what's that mean is unless we really spend time to clean up the playground after you let play with that the playground for buy time from time to time it will become something like this very messy and then in the house is actually my mother-in-law taking care of this kind of responsibility with my kids. So one incident, my mother-in-law went back to Hong Kong to take her break, so guess what happened? The whole playground is becoming so nasty to a point that my wife sat: I, don't care.

B

Mom comes back from Hong Kong next week.

B

Instruction but just ask them to clean up the paper and as well in the next two hours, the kids will try to be like a bundle or prune all the third wheels boxy thing and they called it class Oh the flavio phase. They group them together. They call it stuffed animals and then all the blocks that can stack together. They put it inside one box and they called it level.

B

Instruction they can uncover the pattern after messy paper and then in machine learning. This is called unsupervised. Learning is to uncover some heating pattern inside the data sets and then in the worldwide unsupervised. The running can be used for preparing the customer together. So, for example, how many love? What's the income for a certain people.

B

Married so after everything into a computer and the computer, those people together and then hopefully in the marketing campaign, you can target certain group of people based on the characteristics that the machine can find for you. So this this one of the Haitian food unsupervised, the running and for teaching the kids. You know that it's not always true that you keep showing them the pan or pencil is the most effective way. It's not just trying to tell them this, and then they can uncover certain pattern. Sometimes.

B

If it isn't be bad right, it does reinforce the behavior of a kid and then in machine learning, same concept also happens and we accorded reinforcement, learning debtors when the system performs something, let's say a robot. It's performing something good to the system, but if the system does not perform as you expect, we reduce the point from the system.

B

So, let's imagine the robot, you ask to be able to work in a straight line: okay and then, because the relevant has tomato and the sensor around the robot. So all the data is collected and then into the system so that when the motor light bars that refer down so next time, you will vote it. But when the successful step up one step two steps, then all this will also be record and then being tried to fit it into the model and help the system through one, as the time goes so fast.

B

The reinforcement warning for their machine learning and then we cover so far. We cover supervised the unsupervised and reinforcement learning, and these are the manger Blanche for AI. But that's not the end, because when we think about how to teach a kid where this I, you know them at the very popular question, a kick we ever ask is why I think for those who have kids.

B

You experience that the precise asking the question why the other two very popular questions that I will ask is I'm talking about when they are very young, when they first learn how to talk, and then they first learn how to interact with people. They ask hey: how are you or a newborn girl?

B

So they were also artists us this kind of question.

B

This is also the fundamental for an algorithm that that needs to understand is a number and then in machine learning is called regression because we are trying to probate some certain actual number.

B

That's the that's the class of a class application. The machine learning algorithm has to understand so.

B

It is an apple or it is a pen last tickler, the best as a developer. Let me go back and forth to assure you. Do you see the difference? Well, stammers, a difference that the difference is not allowed.

B

Using the ruse, the if then else, statement, Elvis orders.

B

Of course, we hope the result is good and if it is screw, you keep using that. If somehow there is a case that bricks, then we start incorporating an.

B

B

As a result, so yes well, this we then become self-sufficient. This is why I I is much powerful than the traditional way of programming, because all the developer do is to formulate the AI irrelevant and then the algorithm as we fit more and more data to the then we learn better and better. Just like this model, we're improving, you may have heard about if a man a fish, you feed him for a day.

B

But if you teach a man to fish, you feed him for a lifetime right, so now lovely builders and then they'll convert us to keep it profit. Let us give a computer told you are actually helping to solve only one situation. However, if you teach a computer cold, you are helping this computer for solving a collection of situation. This is the real reason of why AI is so powerful and when we talk about AI, we need to think about what is memorizing.

B

The verses generalizing and the goal here is not to understand not to just memorize what is two plus one, because when you teach your kids, where there's two plus one they would say three. Are you sure you know where that's two plus one they say? Yes, then you ask what is one plus two? They say you haven't told me. I didn't know this bogus too, but if you really understand what 1z is just like know how to do it exactly, but generalizing is really understand. What's the point.

B

B

Memorize the answer- it's not something preferable is also more preferred because it's just randomly throw something in the dark, and then they may get get an answer randomly. So this is also the concept for machine learning and defeating this, my oil overfitting.

B

Now, if we apply this for feeding actually this spot as a human, we understand this is actually bad right. This is the exception that we want to avoid.

B

As normal, so that's you want to avoid, especially for fault detection. You know that this is a you want to catch. It.

B

Interesting is.

B

B

And if I can guess me right, I win. If you can guess me why first you win and I observe my kids to play this game and the first. What they are doing is themselves at the Conte, have and scan through one by one you having the one fishing. Are you having the one playing soccer? Are you playing the one playing like eat all, but few days later they actually forget better smart array. They will ask question like I, won't have to run wearing anything, maybe a glass, maybe something but wearing anything.

B

Or do you have to come with purple in that, so what they're doing they trying to narrow it down and do their decision tree? This is also one of the very popular algorithm in AI. Not only the kid, but AI machine learning also have this, but the whole process of developing this. It's actually as easy and s fund a fund.

B

You know, as the kids playing game, of course, for human gasps, who may be the game, the mastermind, which require a lot more much more sophisticated algorithm.

B

Decision tree may not be enough so depending on which game and which situation we are targeting, we will have different kinds of algorithm, and this is only a rough categories for all the AI algorithm ever exist so far, and you can imagine, as we learn more as we derive more many other Elohim can be derived it.

B

This is Tesla and this is uber. Uber actually is in China. Okay. So can you tell me what is the difference between these two company and wasn't the same for this to come in? First of all, let me ask you what is the same similarity for these two companies anyway, I actually heard here some fans already. What is that car car industry?

B

B

Also, the car, the rest of defense Tesla, we invent the whole engine for the car, the technology of the car.

B

Invent anything new they use map, they use app. The use GPS, which is like a really exist, ten all technology in the business model. Why do I bring this up? Remember this question my kids ask me so: hey I can help me to do homework and the teacher we know if I know that I did not do that right.

B

So I know from the bottom of my heart. I will be so proud of my second grader kids. If he can really.

B

Picture that it's the machine, doing the not him but I was proud of him if he can just utilize the arrow from someone else to Ada, and then he applied to do the homework for himself. So why is it matter?

B

Because there is no right or wrong answer when you are dealing with AI is only right or left depending on what kind of path that you're picking. Sometimes you may want to invent the the technology, and sometimes you may want to utilize the technology, and this is also the main left. The second part of this conversation this session when an enterprise, when a company saying that I want to invest in AI, the company has to understand which part of AI you want in that invest into.

B

You may want to invest in some like technology, so that will improve the AI irrelevant, but simple as utilized that the that the other people has invented for you either way is good, is just a matter of which way you are going to use or employ so now. I think that another key point I want you to take. We got into this session is AI is that we are the traditional traditional BI of the reporting and the dashboarding.

B

Is that, especially for a second day that kid is understand, kind of the AI concept is that unless you incorporate a and other approaches in your bi strategy, that's the remaining five minutes. I know that I have five minutes left, so we don't need to raise that the card. For me, it's serious to conclude my talk in order to make sure that your bi is actually running. Okay, a render a is so prominent right now, the long process, framework, architecture and principle, let's go through them number one process today love this proofs, the M.

B

This is the cross industry standard process for data mining. Look at this.

B

The whole process started from the business understanding to the paid understanding and try to prepare. Your data model is actually done by manually and final deployment, so that's a status. The process so remember that the BI strategy incorporate a I into that in the modeling, and next is the architecture.

B

Architecture is the one here: I shall technology and gnostics you can now find, although you can not find cloud or you cannot find sequel server in this architecture diagram because I intentionally make it technology agnostics.

B

This is the second point that when you build up your beyond strategy, keep in mind, you should not take a look at the technology first, but then we still need to follow some principles, and these are the principle, for example, to have a single version of truth for the data, and it should be open, open ecosystem and apply API to that and provide the self-service analytics tool to add users. Those are the key principles for applying the AI to bi and finally framework.

B

We need a framework to have it so, first of all the solution, of course we have the solution, but most we have the solution. We need to Ali Ridge say that the other people will also understand what you are talking about. The decision.

B

So that whatever you have the skill sets also, they can also be capable for doing this. That's the workforce, development and an able law that is, we apply different things to enable all this together and this we call it as so framework.

B

This is the QR code for my youtube channel. I have just started two weeks ago so and the very very first it's talking about how I did I talk, teach my second million kids about AI. Okay, some people say very good story. Cubed, that's for presentation, I like it.

B

You teach your kids about AI as the way.

C

And uh so computer there something wrong.

B

I can tell you, this is a true story.

B

Thank you very much.

B

And so like a twisting of terrain for Question and Answer.

B

Okay, okay, so like maybe we may not have time officially, but if you can stay behind, we can talk. We can still talk with every okay.