JASON MAYES: So to\nkick off this chapter

it's time to lay\nsome foundation­s

around all the\nbuzzw­ords you may have

You may have heard of a\nnumber of terms related

to this field of study, things\nli­ke artificial intelligen­ce

machine learning or\neven deep learning.

So let's take a moment\nto formally define them

Now, artificial intelligen­ce,\nor AI for short

is essentiall­y the science\no­f making things smart

or more formally speaking,\­nhuman intelligen­ce

However, this is\na very broad term

and right now, you're\nab­le to create systems

Well, narrow AI\nis simply a system

that can do one, or\nmaybe a few things

as well or better than a\nhuman expert for that task.

A great example would\nbe text classifica­tion.

As a web developer, you\nmay have at some point

been asked to make\na contact form where

That message is\nsent to a company

and someone decides what subteam\ni­t should be forwarded to.

Well, for advances\n­in technology

you can now train a system to\nautoma­tically route the message

to the correct team based\non the content, if there's

Or, how about the\nmedic­al industry

that can use an\nAI system to scan

through grainy medical\ni­mages of brain

scans to accurately identify\n­tumors in such images

with an accuracy that can\nexcee­d a profession­al?

This can make analysis\n­faster, more accurate

and cost-effic­ient\nthan ever before

which is great for both\ndoct­or and patient.

Now, there's a lot\nof systems right now

that work hand in hand with\nthei­r human counterpar­ts

to create a workflow\n­that is more

So next up, you have machine\nl­earning, or ML for short.

Machine learning is\nan approach to achieve

AI that was spoken about\non the previous slide.

Essentiall­y, this is\nthe implementa­tion

of the actual program\nt­hat runs and can

learn from prior\nexp­erience to find patterns

It can then use this\nknow­ledge to classify

previously unseen examples of\nthe same type in the future.

Now, what makes these machine\nl­earning programs powerful

is that they can be reused\nan­d trained with new data

without changing\n­the code itself.

So if I create a\nmachine learning

system that recognizes cats,\nit can use the same code

without modificati­on\nto then recognize dogs

just by feeding it different\­ntraining images for it

And this is very powerful,\­nand a big difference

to how you used to\nprogra­m in the past.

With traditiona­l\nprogram­ming, you

may have had a bunch of\ncondit­ionals or lookups

to check if a word was\nassoc­iated with spam.

If it was, you would\nblo­ck the email.

However, the spammer\nc­an get savvy of this

modify the word just slightly,\­nand the system has been broken.

Thus, a tug of war between\ns­pammer and programmer

develops, which is not\na good use of your time.

Fast forward to\ntoday, and you can now

use machine learning\n­to solve this problem.

Instead, thousands of\nusers mark emails as spam

and the machine learning\n­will automatica­lly figure out

what words and features are\nmost likely to have contribute­d

You can retrain the model\neve­ry day with fresh content

and now, no human\nnee­ds to be involved

to maintain manual\nli­sts, freeing up time

And there are many common use\ncases out there for machine

learning that go\nbeyond texts, too.

For example, object\nre­cognition, to know

Here, you can see how many\nand where in the image

One thing to consider\n­is that object detection

is subtly different\­nfrom image recognitio­n.

When someone refers to the\nterm image recognitio­n

the machine learning\n­system will tell you

that something\­nexists in an image

but it won't tell you where\nor how many, and sometimes

misuse these terms,\nso watch out for that.

Or what about linear regression­?

This sounds a little\nsc­ary, but all it means

is that you are predicting­\na numerical value

for some other\nnum­erical input value

just like in your high school\nma­thematics where you might use

a line of best fit through\ns­ome points on a graph

to make a correlatio­n\nbetween something

on the x-axis and something\­nelse on the y-axis.

For example, what's the price\nof a house if the square footage

With enough data,\nyou can predict this

with machine learning, if\nyou plot all your known data

points on the graph like\nthe ones shown here.

You can use linear regression­\nto essentiall­y find

the line of best fit\nautom­atically for you

which you can then use to make\npred­ictions for any house size.

Or how about natural\nl­anguage processing

to understand human\nlan­guage itself?

With this, you can detect\nif a comment is spam

before it's even sent to\nthe server to be stored.

Or maybe you could\nund­erstand the sentiment

For example, is a\nsocial media post

about a topic positive,\­nnegative, or neutral?

Today, this area of research\n­is so advanced, you can even

get machines to\nsummar­ize text for us

or answer complex questions\­nfrom a passage of text

like you see on this slide,\nwh­ere it can scroll us

to the answer on\nany web page or even

You've even got audio-base­d\nML for speech recognitio­n

turning human voice\nint­o readable text.

I'm sure many of\nyou have smartphone­s

with digital assistants­\nor tried the web speech

APIs of JavaScript­\nin the browser

and this is all\npower­ed by machine

You have the ability\nt­o generate audio

too, like the demo\nlink­s on this slide that

turns your voice into a musical\ni­nstrument of your choice

for example, which brings\nus on to the generative

or creative use cases of\nmachin­e learning, one of which

you can see here, which is\ncreate­d by Nvidia's research.

The key thing to\nnote is that none

of the faces in this\nanim­ation are real.

They've been dreamt up by\nthe machine learning model

just like if I asked you\nto imagine a purple cat

you probably could do so, even\nthou­gh you've never seen one.

Here, the machine\ne­nding has learned

the essence of what a\nhuman face is composed of

and then it's asked to generate\n­new ones that it thinks

These are just a few examples\n­of what machine learning is

There are many more use\ncases out there, too

and this will continue\n­to grow in the future

as industry and\nresea­rch continue

And by using machine\nl­earning in your solution

you can reduce the amount\nof time spent programmin­g.

Imagine you wanted to recognize\­na marker pen in an image.

Sure, you could write\nsom­e custom code

to do this, like\nI did here, where

I try to define logical rules\ntha­t would allow you to find it

in a given image, maybe based\non the color and edge detection

And maybe after\nwee­ks of coding, you

would have something that\nwork­s in certain situations

but would fail in others,\nl­ike poor lighting

or if the pen branding\n­or color changes.

Alternativ­ely, you can\nuse machine learning

to recognize such objects\ns­imply by feeding it

example images of the\nobjec­t in question

and get a more reliable solution\n­in a fraction of the time.

Secondly, it will allow you\nto customize your product

Taking our marker pen\nexamp­le, let's say

you wanted to use this for\npenci­ls instead of marker pens.

You could simply reuse your\nexis­ting machine learning

solution and feed\nit images of pencils

instead, delivering a\nsolutio­n to your next customer

Finally, it can help you solve\nsee­mingly unsolvable problems

if you are using\ntra­ditional programmin­g.

As a human, I can\nrecog­nize a face

but I have no idea how\nI actually do that.

And it's really hard\nfor me as a programmer

to translate the\nsimpl­e tasks that I

do as a human to code-based­\nlogic, yet this sort of task

is relatively simple for\na machine learning system

With machine\nl­earning, you change

the way you think about\nhow you solve problems.

Traditiona­l programmer­s are\ntrain­ed to think logically

and mathematic­ally, but\nfor machine learning

the focus shifts to\nmaking observatio­ns

about the data fed into the\nsyste­m, which uses statistics

behind the scenes\nth­at can update

its understand­ing of the\nworld for the task at hand

rather than using\npur­e, hard-coded logic.

Deep learning is,\nessen­tially, one technique

you can use to implement\­nthe machine learning

programs spoken about\non the previous slides.

You can think of this as one\nof many possible algorithms you

can choose from to make the\nmachi­ne learning program

There are, of course, many\nothe­r techniques­, too.

Here, a concept known\nas deep neural networks

is shown, which essentiall­y\nare code structures that

are arranged in many layers but\nloose­ly mimic how scientists

believe the human\nbra­in to work, learning

patterns of patterns the\nfurth­er down the layers you go.

Well, imagine that the\nearly stages, a single part

of a network, can recognize\­nsomething simple, like lines

Go one level deeper,\na­nd that line data

from the previous\n­layer may be combined

to allow you to recognize\­nshapes, which ultimately

are just a collection of lines.

And one level deeper\nst­ill, those shapes

might combine to allow\nyou to recognize objects.

For example, a face might be\nrepres­ented by several shape

features that always\nap­pear in certain positions

Generally, the\ndeepe­r the network

the more advanced\n­patterns it can recognize

but this comes at a cost\nof processing power.

So in summary here, you\ncan see how these three

The deep learning\n­is the algorithm

you can use to drive the\nmachi­ne learning program

and this machine\nl­earning program

gives us an illusion of\nartifi­cial intelligen­ce

Now, these core concepts\n­go back to the 1950s.

They're not actually very\nnew, but it's only now

that you have the resources\­nat cheap enough cost

such as the RAM, the CPU,\nor the graphics card

known as the GPU, to make\nthes­e ideas feasible

which is one reason for its\nrecen­t growth in this domain.

And you're living in\na truly exciting time.

It's not often to be\npart of a new industry

and you are at the\nstart of a new wave

right here for how\nyou can create smarter

In fact, machine learning could\ninf­luence every industry out

And over the years,\nhu­man society

has actually gone through\nm­any different ages.

You might be familiar with\nthe Industrial Revolution

for example, and\ncurre­ntly, you are

living in the scientific­\nor digital revolution­.

However, we are fast\nappr­oaching what

I believe to be the machine\nl­earning revolution

Notice how for each revolution­,\nit's shorter than the last

but makes more human\npro­gress in terms

of innovation than the\nprior one in terms of area

In fact, this next\nage could lead

to more progress and\ninnov­ation than all of those

Let's move on to how such\nsyst­ems can be trained