00:00:00
i want to show you two video clips one of
them is real and one of them is fake look
00:00:04
at these two clips one of them is real and one
of them is fake can you tell which one it is
00:00:12
learning of the dangers of AI artificial
intelligence AI poses to the social i've
00:00:18
never seen it quite like this this
technology is spreading rapidly it's
00:00:22
really mindblowing deep fakes deep fakes
deep Tom Cruz was a tipping point for deep fakes
00:00:28
we're increasingly in a world where
AI is everywhere
00:00:31
but do we actually know what's really going on
00:00:35
let's dig into what AI really is
with me Crystal Widjaja at Kelas Pakar
00:00:55
yeah AI has actually been around for many years
now um coming from the tech field I started doing
00:01:02
machine learning um with a Python script in
my laptop maybe 20 years ago but today we're
00:01:10
increasingly exposed to AI in the wild in the
wild being things like in our news on Twitter
00:01:17
uh social media we have to ask ourselves
is this post real or did someone deep fake this
00:01:24
and the paradigm is moving so quickly
that it's hard to tell what we can trust anymore
00:01:30
coming from tech I know how AI works
I know when to be suspicious when we see AI
00:01:37
now controlling things like spam or fraud
and risk we start to see AI doing scarier things
00:01:45
and so I'd like to deepen to dive
into what are the actual mechanics behind AI
00:01:58
AI at its front is really just machine learning
applied machine learning has the ability to create
00:02:06
reason and infer and listen to data around it
and what comes out is the ability to create
00:02:13
new journalistic interviews like this or scripts
might be researched by AI or you might see photos
00:02:21
that never actually existed but were created by
a generative AI product and that might be scary
00:02:39
we see companies cutting jobs and saying we you
can be replaced by AI Klarna for example in the US
00:02:46
replaced its customer service callers with uh
AI service agents and that can be kind of brutal
00:02:54
we find that you no longer have that human
touch anymore
00:02:58
I personally don't think that
00:02:59
journalists will be the easiest entity to replace with AI
00:03:03
I think journalism is a media format that
00:03:07
requires trust and frankly if you think about it
most people don't trust other people and so why
00:03:13
would humans trust machines well if you think
about it AI is the first kind of machine where
00:03:19
you've had to doubt what it says usually when
we're given access to a computer we can trust
00:03:26
that it's going to do what it says it will do
because it's been programmed with static data
00:03:31
static information and frameworks so we never
had to ask ourselves could the resulting answer
00:03:37
be false but now with generative AI we do have to
ask that question we have to be a bit skeptical
00:03:43
about the results that we are getting from an AI
program or an AI that's generating information
00:03:50
and so with journalism well at the heart of it
it's a human experience that you have to trust
00:03:55
you to trust that someone has gone through the
facts has talked to other people on the ground
00:04:02
and picked up on data that isn't really on
a spreadsheet it's not really available in
00:04:08
a poll it's something that you have to talk to
humans and ask them how they're feeling about
00:04:13
the situation and so journalism while on its face
you can replace maybe some of the content you will
00:04:19
not get to a place where you have trustworthy AI
and that's something I think will never change
00:04:29
in tech we might see companies replacing workers
who are able to do kind of simple fact retrieval
00:04:36
so in the legal field you may have people who are
able to pull together cases but you aren't going
00:04:42
to replace the person who has to pour through
all of the details and deeply understand it
00:04:47
and in fact the kind of jobs that are most safe
from AI replacing them are jobs where you have
00:04:54
to interact with another human being or need
to have built trust and empathy with the users
00:05:00
in tech you might want you might be able to
replace an engineer in tech with AI because
00:05:05
they've learned to code on the same machines but
how do you replace someone who knows how to sing
00:05:12
or dance or create art that is innovative and new
not something that AI has the ability to do right now
00:05:25
so I'll talk about the there is a maturity
with AI and how we are experiencing it in the
00:05:31
real world today if you think about a product
like Google maps where initially you use it to
00:05:36
get a static result from point A to point B what
turns do I make and how long will it take that
00:05:42
used to be a static snapshot from data based on
you know geographic maps and the static laws of
00:05:50
the streets and then we moved into more predictive
data where you were able to consider you know an
00:05:58
accident might have just happened and so the
map updates itself based on new inputs of data
00:06:04
that would be a reactive data system or a product
that based on a new input is able to correct its
00:06:13
advice and then we move into generative kind of
predictive mapping where in Google Maps it knows
00:06:20
that there isn't any traffic right now um you
know on the way home from work but it does know
00:06:26
based on historical data that within 30 minutes
there will be about a 10-minute slowdown and so
00:06:33
it adds that into the prediction of how long your
route will take to get home um before it actually
00:06:39
happens and that ends up being a example of
predictive AI not all companies have this
00:06:45
uh ability not all companies should be embracing
AI we should like many other things embrace AI
00:06:51
only when it makes sense for our business or
for our company's needs most companies don't
00:06:57
even use the data that they already have today
and so I'll talk to companies who you know they
00:07:03
understand that users are churning or they're
leaving their product but they don't understand
00:07:07
the inputs as to why and instead they want a
very fancy AI machine learning model to tell
00:07:12
them who's going to churn and I say "Great i'll tell
you this customer right here they're going to churn."
00:07:19
And I know that because you know 11%
of customers turn on the first day what are you
00:07:23
going to do about it and they have no answer so AI
when used to make predictions without telling you
00:07:30
the the reasons why it's made that prediction
in the case of most deep learning isn't that
00:07:36
useful it might help you build a financial model
but it doesn't help you act in the real world
00:07:46
so companies there are many companies today
that survive not because they have the most
00:07:51
data or the cleanest data but because they are
a delightful product and delight is rare today
00:07:57
you need people to understand and have empathy to
have one there's the offline services a therapist
00:08:04
um for someone to kind of feel the emotion with
you to hold your hand with you as you are crying
00:08:09
through a bad day that's not something AI is
going to be able to do and if we make a robot
00:08:15
out of it I'm not sure that I'm going to want
that hug um versus you know my parents or a
00:08:20
friend a someone to cry a shoulder to cry on is
not going to be replaced by AI We may make tech
00:08:27
enabled tools that maybe use AI to create
better and faster services or matchmaking
00:08:34
but you're not going to have a product whose
delightfulness it's the ability to scroll and
00:08:40
see animations the ability to create characters
that you love that are mascots of your app
00:08:51
think about the last time you scrolled Instagram
and you thought to yourself "Wow that's a really
00:08:56
cool post or that's a really cool visual." That
probably wasn't AI generated it was probably
00:09:02
someone's taste and taste is something you can't
magically create through data most people don't
00:09:09
have good taste and as a result AI is pretty
bad taste it's quite mid and so if you want to
00:09:16
be someone with who is irreplaceable who cannot be
replaced by AI you got to have really good taste
00:09:23
you'll have to be design first that requires
human empathy an ability to understand how
00:09:30
the world perceives the products and services
that we have to offer and to make something compelling
00:09:42
for those of us who are in engineering
today or maybe you thought you know it's it's a
00:09:46
good time to become an engineer it's still true
it's still true that we need more engineers it's
00:09:51
just that the bar for an engineer has gone a lot
higher it's no longer enough to be able to code or
00:09:58
know a library it's going to be important to be
someone who makes thoughtful decisions that can
00:10:04
understand and perceive how the business is moving
what architectural design decisions have been made
00:10:10
and the types of people at the company whether or
not they might like this type of library versus
00:10:15
another one there's never been a better time to
be an excellent engineer is what I tell people
00:10:22
it's a pretty bad time to be an engineer
who can only write code and instead will
00:10:27
have to care a lot more about the business
engineers who only care about code and tech
00:10:34
and don't look beyond their scope of work
are going to have a really bad time in a
00:10:39
world where AI can replace a lot of the
functional tasks that we used to do as engineers
00:10:50
and instead we'll have
to do what AI cannot do which is
00:10:53
talk to our customers try to design
a compelling experience to consider
00:10:59
uh and make decisions on how people will
experience our product and that can be as
00:11:06
simple as the vibe that you want your landing
page to look like is a decision that most AI
00:11:13
leaves up to chance and if you leave things up
to chance you are no better than an AI system
