Latest / Linen Suit & Plastic Tie / What's A Prediction Machine? ft. an AI Professor (Dr. Avi Goldfarb)

Transcript

What's A Prediction Machine? ft. an AI Professor (Dr. Avi Goldfarb) · Linen Suit & Plastic Tie

0:04
Welcome welcome. Welcome to the linen suit and
0:06
plastic type hard cast. This the podcast will be worth
0:09
to unlock the amazing, the suspenders, the Epic D
0:13
influential the Epic again, power of Storytelling and help
0:18
you harness that power in your everyday lives.
0:21
I'm going. And I'm Kevin.
0:25
Kevin did have I told you my Santa Claus and Coca-Cola.
0:28
Fun fact? No, I Don't think it's made into
0:31
the any episode soon. You're cleared.
0:34
Oh, cool. Well, because we're getting
0:37
closer to the holidays, I would love to tell you.
0:42
The story about how Santa Claus got this image of, like, big man
0:48
in red cloaks, and like that hold five.
0:55
And that was actually, the first reported documentation of,
0:58
Santa's is this know? When President in North America
1:02
of this, like big man in a red outfit, with a big beard was
1:06
actually done by Coca-Cola. Coca-Cola did this huge Grand in
1:10
campaign long ago. We're essentially, they showed
1:14
off this Santa Claus is this big round, man?
1:17
In red before, then most of the pictures and images.
1:22
We have a Santa was actually a thin man in a green cloak but
1:26
Coca-Cola changed that because Coca-Cola red is such A central
1:30
part of their brand that for the holiday campaigns, they really
1:34
literally. We branded Santa and now for
1:38
billions mini. No, no, how many people, many,
1:41
many people worldwide. See Santa has this big man in
1:46
red outfit solely because Coca-Cola wanted him to be red,
1:51
because red is so influential and important to the Coca-Cola
1:54
brand. And I just love that.
1:57
I love this idea of Storytelling.
1:58
I love this idea that We talked a little bit about and I'm sure
2:02
we'll talk more about how commercialized holidays are
2:06
especially in America, how there's all these holidays.
2:08
They found things day like Halloween, like, Christmas that
2:13
are huge spending events in America because all these brands
2:17
have created or pumped up these holidays as ways to spend money,
2:22
right? So it's so interesting, how Coco
2:24
jumped on that kind of capitalist mere can train to be
2:28
like, oh, it's Christmas. Think about red.
2:31
Thank you, Lord. Santa Claus.
2:32
Think about Coca-Cola. And there's also Coco's is such
2:35
an old company that there's so many of these fun facts about
2:37
Coca-Cola where they're like, oh, you're kidding.
2:40
And a lot of them has to do with them building this brand.
2:43
That's lasted, hundreds of years and will last longer.
2:47
Yeah, it's really fun, for me at least to look at how a lot of
2:54
holidays, really catch on because of these marketing
2:59
campaign. Like holidays, like, Christmas
3:02
and Halloween are not culturally native to China, for example,
3:07
but this past Halloween. I've definitely seen way more
3:11
people who are in China for more enthusiastic about this Festival
3:16
on social media. Then I was and I been living in
3:21
the United States for six plus years.
3:24
Anyways, that was a little bit of a sidetracked, Sidetrack
3:28
intro today but I can't Who we talking to this week today?
3:32
We are talking with someone who is an expert in a really cool
3:39
field dr. Avi Goldfarb is a professor at
3:46
University of Toronto and he's the chair of artificial
3:51
intelligence in healthcare at the rotman School of Business.
3:55
He is also author of some Sick books related to the field of
4:03
AI, including the prediction machine which grabbed loves and
4:08
also. He is coming out with a new
4:10
book, called power and prediction, the disrupted
4:15
economics of artificial intelligence.
4:18
So this is going to be a very insightful conversation where we
4:24
try to demystify Ai and And AI economics and also how
4:32
storytelling plays a part in to this seemingly, very tactical
4:36
field. So let's get started.
4:42
Today was our glad to be joined by dr.
4:45
David Goldfarb, who has an amazing story and a lot of great
4:50
insights that we cannot wait to get into.
4:54
So Avi to get us started, can you tell us a little bit about
4:58
yourself? What is your story?
5:00
Sure, I grew up in Toronto and then for in grad, school in the
5:07
late 90s, I went to Chicago to Northwestern University in the
5:11
econ Department. Did my PhD in economics And I
5:15
was looking to figure out, you know, what should I do?
5:17
My dissertation on and my first paper was a study of the beer
5:23
industry, particularly the decline of Schlitz beer, okay?
5:26
And in the 1970s, then I was looking for my next paper and I
5:31
was intrigued by a bunch of research on the impact of
5:33
advertising on cigarettes and I had this moment in the middle of
5:37
my third year of my PhD saying I don't want to be the big beer
5:40
and cigarettes guy. Okay so I started Look around
5:43
for something else to study and by then it was 1999.
5:49
And there was this really cool thing happening that no one
5:53
really understood, which was the internet and so I I remember
5:59
deciding. Okay, well, Since no one knows
6:02
anything about that. That's the kind of thing that I
6:04
could do my PhD on and no matter what I find it'll be new.
6:08
So I decided to, you know, to study the internet in the
6:11
economic impact of the internet and my dissertation was about
6:14
search engine competition back when there really was such a
6:18
competition. Google was in my dataset and
6:20
they were the 17th most popular search engine in the country
6:25
from that. I then graduated went to
6:28
University of Toronto moved from economics in I'm in the
6:31
marketing department and my research in the early days
6:36
focused on understandings of how did the internet impacts society
6:41
and impact marketing, that led to a bunch of working online
6:44
advertising and to privacy. And then in 2012, my co-author
6:52
and colleague, Jay Agarwal. Started, this organization
6:54
called creative destruction lab, which is a program for helping
6:57
science based startup scale. In the very first year at a lab,
7:01
we had this company that said they were doing AI for drug
7:04
discovery. And ten years ago, anybody
7:08
talking about a I seem too crazy and the idea that an artificial
7:11
intelligence was going to help anybody figure out which drugs
7:15
were going to save people's lives.
7:18
Seemed like science fiction. And so with that, well, this is
7:22
interesting and let's dig a little into this and the next
7:26
year we had a couple more companies partly because we were
7:28
sitting in Toronto and a lot of these companies were coming out
7:31
of Jeff Anton's lab, a few blocks down, the street, from
7:34
our business school in the in the computer science department.
7:37
And then by 2015, we had a flood of companies coming out of
7:42
University, of Toronto and Waterloo.
7:43
And a few other places that were AI companies for all sorts of
7:47
applications. And at that point, Jay Joshua,
7:51
not my co-authors, and I decided that we really wanted to get our
7:54
heads around this. And so we started to think
7:57
through what what are the economics of a?
8:00
I look like and take the the tools that we learned from the
8:08
first decade and a half of our career, studying the economic
8:11
impact of the internet combined with the insights of the, by
8:15
then dozens of startups. And by now hundreds of startups
8:17
that we've seen trying to commercialize a, i in our lab to
8:22
create our first book production machines and then with five more
8:28
years of data and information, it's led to our new book, power
8:32
and prediction. To kind of Define the context of
8:37
our conversation. What is AI?
8:40
Really what is this AI concept? You are speaking of with you,
8:45
your research because I feel like you know there can be so
8:48
many stories and stigmatization Surround that people don't
8:53
necessarily understand what we're talking about.
8:56
Okay? So the reason we're talking
8:58
about artificial intelligence you know in 2022 and we weren't
9:03
talking about it. Much in 2002.
9:06
It's because a very this very particular branch of AI research
9:11
is very particular branch of, it's a branch of computational.
9:14
Statistics has gotten much better and that's machine
9:18
learning tools. We can think about that as
9:19
prediction, okay? So prediction in any economic
9:24
sense has gotten better faster and cheaper.
9:27
So that's not to say that there's not excite you someday.
9:30
We might not have an artificial general intelligence that can do
9:32
everything humans, do. But maybe better like a machine
9:36
that could really think that might happen someday that's been
9:41
20 to 50 years away, since the 1960s and it continues to be 20
9:46
to 50 years away if you talk to the expert.
9:47
So that's, that's a really cool. Technology is an exciting
9:50
technology, but it's not. The reason were seeing
9:54
commercial deployment of AI today.
9:57
The reason we're seeing all this excitement around AI today is
9:59
because of machine learning technology, we can think about
10:01
his prediction And from there, what is aii?
10:08
Economics, AI economics is first to say that the reason we're
10:14
talking about machine learning today, is because predictions
10:16
gotten better faster and cheaper.
10:17
And then to say, what's prediction for a prediction in
10:21
and of itself has no value, if someone comes to you and they
10:26
say, have an AI and you say, what's it for?
10:28
And they tell you, it'll provide you with the insights, then you
10:31
should run or at least. Back away slowly.
10:35
Okay, that means that person doesn't know what the
10:37
predictions for, what a predictions for is it helps you
10:41
make better decisions. And so if if you ask somebody
10:45
what their AI is for, they should be able to describe the
10:47
decisions that the prediction will support.
10:52
So AI economics is first the recognition that we have better
10:55
faster cheaper production second the going back to econ 101.
11:03
When something is cheap, you use it more.
11:05
So if coffee is cheap, you do more, you buy more coffee.
11:08
If prediction is cheap, we're going to do a lot more
11:10
prediction. And then to say, well if
11:15
predictions cheap, what becomes more valuable, what are the
11:17
compliments? Just like, when coffee gets
11:19
cheap, it's good to be in the cream and sugar business.
11:22
What is it? Good to be?
11:24
What gets more valuable as AI starts deploying its scale and
11:30
we emphasize data, which is no surprise, he came out of a data
11:35
analytics master's program. That's data becomes much more
11:38
valuable cuz that's the core input into the predictions and
11:41
then there's this thing we call. Judgment, which is knowing which
11:45
predictions to make and what to do with those predictions.
11:47
Once you have them and that is an inherently human skill.
11:53
And that is maybe the biggest opportunity in the biggest
11:56
challenge with using machine learning tools because in our
12:00
human brains right now and you make decisions, you're combining
12:03
the prediction and the Judgment together and it's hard for you
12:07
to figure out what is what with the machine learning tool does,
12:10
what AI does is? It provides you with a
12:12
prediction that means it forces you to decouple the prediction
12:17
from the rest of the decision? And that creates opportunity
12:22
because you can make much better decisions once you have the
12:24
prediction in place, but it also creates challenges with respect
12:29
to disruption. Because maybe the people who are
12:34
really good at doing the prediction and judgment together
12:37
are not the people who are going to be great at figuring out what
12:40
to do with the prediction once the machine provides it.
12:43
And so that creates that creates opportunities for entrepreneurs
12:48
to think creatively about what that decoupling of prediction
12:51
for the rest of the season can offer, and it creates threats to
12:54
established organizations established businesses, because
12:57
they might have some of their power, take it away.
13:01
So can I tell you a story around that?
13:03
Yes, storytelling. Okay, loves toys.
13:07
Okay. So the story takes place in
13:11
Flint Michigan and if you remember a few years ago, there
13:15
was a water crisis in Flint Michigan.
13:17
They discovered that a bunch of the pipes in Flint Michigan had
13:20
led and let his deadly neurotoxin and you do not want
13:23
to drink water that has Legend. And so the city among other
13:28
things started to try to dig up and replace all the lead pipes.
13:34
Problem is the only way to figure out if a pipe going into
13:36
a house as lead is to dig it up, okay?
13:39
And so much of the effort in replacing a pipe is already done
13:45
once you discover that the pipe has led so Two, professors at
13:52
University of Michigan. Built machine learning tool to
13:57
predict, which houses likely had lead pipes in the city.
14:02
Use that tool before digging and they were right something like
14:07
80% of the time. Okay.
14:08
So four out of every five houses where they dug to figure out if
14:13
the pipes were led the pipes were LED that's a pretty good
14:18
success rate. Then people started to complain
14:23
because some people were not having their pipes dug up and
14:26
they say how come the people be street over, get their pipes,
14:29
dug up and not mine. That doesn't seem fair.
14:31
I want to know for sure that my pipes aren't LED and so they
14:35
complain to the politicians And the politicians said, well, you
14:39
know what, you're right, it doesn't seem fair, you who vote
14:43
are complaining to me and somehow this algorithm, that's
14:47
we don't really know what's going on.
14:48
Is to is allocating who gets to have their pipe stuck up.
14:53
That's not fair to go systematically throughout the
14:55
city street, by Street house by house and make sure everybody's
15:00
safe and comfortable. So they did that and the success
15:04
rate went from 80% to 20% Which meant that hundreds if not
15:09
thousands of people were now drinking toxic water that they
15:13
would not have been drinking. Had they continued to use the
15:17
machine learning tool the Atlanta called it a feel-good
15:20
story, feel good, AI story gone wrong.
15:24
But the story didn't stop there because because they decouple
15:27
the prediction of the Judgment radio was clear what the right
15:31
decision was. It was, it was clear that it was
15:36
wrong to go with a twenty percent accuracy rate over an
15:39
80% accuracy rate because those people felt it was, you know,
15:42
because it was more systematic. And so, a judge ultimately ruled
15:47
that the city had to follow the professor's prediction tool.
15:53
And it went in that success rate, went back up from 20 to
15:56
about 70 percent. And so, as a consequence, again,
16:03
people got their lead pipes replaced and we're drinking, and
16:07
a lot of people were now drinking clean water who
16:09
otherwise would have been drinking toxic water.
16:13
Now, the story on where does that play out?
16:17
In terms of power is because of that two, coupling once you had
16:20
a prediction it was very clear what the wrong decision was.
16:24
And so the people who always used to make decisions the
16:27
politicians, when it was clear enough of they were making the
16:31
wrong decision, they were overruled.
16:34
And we ended up using the AI combined with the judge's
16:38
decision. In this case that the people
16:42
with the highest likelihood of having lead pipes were the ones
16:45
who should have their pipes replaced, I think that's such a
16:48
clear story but how prediction works and how the importance of
16:52
taking these steps taking these systematic steps in AI to make
16:55
clear predictions and how the drop in cost and petition has
16:59
led to this kind of explosion in the field.
17:02
And I think that's so interesting because it's a way
17:05
that a I was used to directly impact the wellness of someone's
17:08
life. But I can also see because of
17:11
the way stories but II told because of this kind of big
17:15
term, artificial intelligence May I there's fear there because
17:19
it again impacting someone's life.
17:21
Can you tell me a little bit about in your experience about
17:23
how this term AI is seen in the general public and how that's
17:28
affected your work? Sure.
17:31
The starting point for almost every talk I give about AI is to
17:37
point out that this is not the artificial intelligence.
17:40
We see it science fiction. And so, I often will have a
17:46
slide with images from science fiction, whether it's from Star
17:48
Wars, or from the Terminator, or the Matrix or elsewhere saying,
17:53
look, this is the popular image of an intelligent machine.
17:58
It's not impossible. As I said to you before, I
18:01
actually don't think this idea of an artificial.
18:03
General intelligence is impossible.
18:04
I think it's likely But it's not the technology we're talking
18:07
about today. When you hear a I don't think a
18:12
machine that can do just about everything humans do instead,
18:15
think production machine and that's where the advances have
18:18
been over the past 10 years that's where the commercial
18:22
opportunities lie. Then I want to point out the
18:27
prediction is a big deal, right? So it's not, yeah, it's a pretty
18:29
sure machine, but prediction can be transformative because a
18:35
predictions in input into decision, making and decision
18:38
making is everywhere. That's with one important
18:43
caveat. So we wrote production machines,
18:44
we published it in 2018 and we were super excited about a
18:47
revolution that was about to start.
18:49
And so we thought the industry was going to change as people
18:56
realized what they could do with prediction.
18:59
And a lot of the barriers, over the next two years, on the
19:02
technology side starting to be overcome by companies who are
19:05
investing in it like, you know, collect enough data having
19:10
People in master students and others who are good at on
19:12
analytics, that it's not solved. But we've made major progress
19:17
there and yet even by 2021 even today, we don't see that many
19:22
industries that have been transformed by AI.
19:26
And so that's the starting point of our new book which is we
19:30
wrote prediction machines thinking this Revolution was
19:32
about to happen and here we are five years later and it hasn't
19:35
happened. What went wrong.
19:37
There's a number of possibilities.
19:39
One is it's just not that big a deal of the technology to be
19:42
clear. I think that's not what went
19:44
wrong. The other, is that in order for
19:49
a technology to really be transformative, we have to
19:53
reinvent the way we live in the way we work. and that process of
19:57
Co invention of organizational, innovation of complementarities,
20:03
takes time And the the analogy I think is most powerful is
20:08
understanding the story of electricity.
20:11
So if you go back and you read what people were saying in the
20:15
1880s about electricity, it was clear that for those who are
20:20
paying attention, they knew that electricity was going to be a
20:24
big deal. It was clear that the electric
20:27
light in the electric motor and a handful of other Technologies.
20:31
We're going to change the way we work to lift.
20:35
But it wasn't until the 1920s, 40 years later that the median
20:40
household of the median Factory. The US were electrified, Why did
20:44
it take 40 years for a technology that it should be
20:47
clear to everybody was going to be amazing?
20:51
Well, we had to reinvent the way we lived and worked.
20:56
So for example, the first factories that use electricity,
21:01
Took out their old power source and put a new one.
21:04
So you'd have a factory, the 1880s that was organized around.
21:07
Say, the steam engine that was the power source, really mean it
21:10
was organized around this Dimension.
21:11
I mean it was literally organized around the steam
21:13
engine because the farther away the machines were from the power
21:16
source, the more energy got dissipated and so they
21:20
structured the geography of the factory around the power source.
21:25
The most power hungry machines were as close to the power
21:27
source as possible. Then the first factories that
21:31
electrified. They said, oh, you know what,
21:33
for where we are electricity's going to be cheap power, and
21:36
they took out the steam engine put an electric motor in the
21:38
center of the factory but they didn't move any of the machines.
21:41
And it's saved a little bit on energy and this was like what we
21:44
call a point solution, took out one point in dropped, in the
21:48
electricity electric motor in the same place and that saved a
21:52
little energy and didn't really transform the factory.
21:55
It wasn't widely adopted, it was only for the particular Factor.
21:58
He's where the steam engine was an expensive way of operating or
22:03
the water wheel. Then over the next 40 years,
22:06
people realize that electricity wasn't just cheap power, but
22:09
enabled you to decouple the power source for the machines.
22:15
And once you could decouple the power source in the machines,
22:17
you can organize your workflow differently with inputs coming
22:20
in one end and outputs coming out the other over a typically a
22:24
big flat surface where you'd want cheap land because you want
22:27
to be able to spread out the work.
22:28
Flo has a different kind of factory was much more productive
22:33
and it allowed people to allow businesses to build products
22:38
much more cheaply. And with much better quality
22:43
control that would have existed in the old style factories.
22:46
There were steam powered but that took a long time to
22:50
reimagine What the factory would look like to take advantage of
22:54
the decoupling of the machines from the power source.
22:58
That's what we call a system solution in those 40 years.
23:02
Those 40 years are the between times the time between seeing
23:06
the potential of a technology and recognizing what it can
23:09
actually deliver. And so we think with Ki going
23:13
forward, but we haven't figured out what the organization of the
23:19
future is that takes advantage of better faster and cheaper
23:22
prediction. We have a couple examples.
23:25
We have Google and we've a handful of others that are
23:27
really transforming aspects of the way we work and live through
23:33
better prediction but for most Industries and most companies,
23:38
they haven't been able to do that reimagining yet.
23:41
Now, I'd like to hope it's not going to be another 40 years
23:44
till we see the impact of AI on Society at scale.
23:49
But I do think we need to recognize that it's not just a
23:52
point solution. It's not about taking out some
23:54
human process and dropping in a machine that happens.
23:57
Lots of companies will do that, but it's necessarily going to be
23:59
incremental. The big picture is when we can
24:02
redesign a system to take advantage of the technology in
24:05
order to deliver new value to ultimately to our customers.
24:11
That's very interesting because you know, like you said, a lot
24:14
of companies talk about a lot of companies talk about integrating
24:19
Technologies. We see like Facebook talking
24:22
about using AI to find spam or hate speech.
24:28
We here. We're hearing popped a pic and
24:31
Disney talking. A lot about how they want to
24:33
connect Disney, plus to the parks to use petitions to see
24:37
based on what they were doing in the park.
24:39
What should we recommend on Disney plus?
24:41
So we're healing petition being talked about a lot in a I being
24:44
talked about a lot in these big buzz was being thrown out.
24:47
Do you, would you feel Do you think a lot of these big
24:50
companies are trying to kind of implement more Point Solutions
24:53
and kind of the way to go about? This is what working with
24:56
startups to build them in a way that they long-term growth.
25:00
What is that kind of relationship in your mind?
25:03
Okay, so yes, most big companies are thinking about Point
25:06
Solutions because they don't want to mess with their
25:08
workflow, they don't want to change that requires figuring
25:16
out how to often will require a reorg.
25:19
Figuring out who should be responsible for what in the
25:22
presence of better prediction, it will require new ways of
25:26
coordinating and working together because now that one
25:29
part of your organization is responding on a daily or hourly
25:32
basis to these predictions. Everyone else has to come along
25:36
and be able to coordinate. So, like organizations have
25:39
these, we call standard operating procedures.
25:41
They have these rules, and the rules are necessary to, make
25:45
sure everybody is at least Marching In The Same Direction.
25:49
Section. But once you have a prediction
25:51
machine in principle, the idea is instead of just following a
25:55
rule all the time, you're going to see what that prediction is
25:58
and you're gonna make a decision different things based on the
26:00
state of the world. But once you're doing different
26:02
things, based on the state of the world, the rest of the
26:05
company has to know what you're doing and your suppliers need to
26:07
know what you're doing, and your customers need to know what
26:09
you're doing. That requires rethinking often
26:12
of the process of the way we operate and potentially
26:17
delivering a new kind of value that's hard.
26:21
And so instead what we're seeing are people looking at the
26:23
low-hanging fruit saying, hey, here's a process that we
26:27
currently have some manual way to do it.
26:30
Let's use a prediction tool for that.
26:31
So, with the additional Parks example would be, you have some
26:37
prediction about how long the lines are going to be and it
26:39
used to be. There was someone standing in a
26:42
line who would give you like this, this wooden thing that you
26:45
walk to the end and then they'd be able to time it was taking 45
26:48
minutes to get people go through the line.
26:50
And then they're using AI tools to predict that that will help.
26:53
You know, also, tell people what rides are going to like, okay,
26:57
what you just described is potentially a way to deliver
26:59
more value, which is to then use what's happening at the park, to
27:04
provide a new kind of service, a Disney plus, a new kind of now,
27:09
Maybe it's not a system level change, its someone incremental,
27:11
but at least it's delivering a different kind of value than
27:14
just doing the same thing a little bit better, through Ai.
27:17
And that's where a lot of the opportunities lie thinking
27:19
through, what can you be thinking through the mission of
27:23
the company? What do you actually trying to
27:26
achieve? What's your goal and then?
27:29
How can you better deliver on that mission in the presence of
27:32
prediction thinking through the role of mission in delivering
27:37
value is essential? And that's the Very much a
27:39
storytelling aspect of it. So each company in each CEO,
27:44
should be able to describe what the mission of the company is.
27:47
And then once, you know the mission, you can think through
27:51
how better prediction can allow you to deliver that to think
27:53
about an airport that's just at LAX last week.
27:56
It's a huge airport okay? And they they have a lot of
28:01
excellent services at LAX. Okay.
28:03
So there's pretty good restaurants.
28:05
There's all sorts of certainly depending on the terminal The
28:09
terminal fair enough. Okay.
28:10
Say LAX is good enough. Seoul Incheon Airport is
28:13
fantastic. It's often rated the best
28:14
airport of the entire world, okay?
28:16
So if you didn't, it's pretty spectacular.
28:18
So the restaurants and the shopping is a whole nother level
28:21
and there's like theater and Fantastico Talon massage chairs
28:25
everywhere. It's just, it's a great airport
28:27
and it's part of like a multi-billion dollar investment
28:30
on the part of the government over the past few decades, the
28:34
Korean government. Now, what's the mission of Seoul
28:37
Incheon Airport? The mission of Seoul Incheon
28:40
Airport is to ensure smooth air transportation, ensuring smooth
28:44
air. Transportation has almost
28:46
nothing to do with the hotel, the restaurants, the shopping,
28:51
and all that. You need all that stuff because
28:55
you have bad prediction. If you knew how long it would
28:57
take you to get to the airport through security, then you just
29:01
walk right on the plane and the airport itself would be a shed.
29:04
In fact, we have an example of that if you fly private, which
29:09
someday we all may aspire to do, or maybe not.
29:12
I don't know, but I've been told as you fly private the
29:15
airport's, our sheds there's like this is a few chairs and
29:19
some coffee and that's it. Why because if you fly private,
29:23
you arrive at the airport and get on the plane you don't spend
29:25
any time in the airport. You don't have to deal with that
29:28
uncertainty of getting through security and traffic at the
29:30
airport, it's the ultimate airport you know, ensuring
29:34
smooth entry air transportation experience.
29:38
So we can imagine, how could Seoul Incheon Airport or LAX, or
29:44
any other airport? Do a better job, delivering
29:46
value, delivering on the mission, and it's ensuring
29:49
smooth air transportation. Once we have good predictions
29:51
about how to get through to the airport through security, how
29:55
long it's going to take. And part of the answer, might be
29:59
all that architecture. That exists is billions and
30:03
billions of dollars that exist around.
30:05
Keeping you happy when you're bored at the airport are no
30:08
longer necessary. We can reimagine the airport to
30:11
be more like the private terminal in every industry.
30:15
You can think through what's your mission and then once you
30:19
understand the mission, you can think through.
30:22
Are there ways we can deliver on that mission better?
30:25
Once we can do better prediction, do a better job at
30:28
filling in missing information. We very interesting perspective
30:33
and also on that on that story, you told about the airport.
30:37
I thought it was really interesting as well.
30:39
It's kind of, you know, reimagining an airport to be
30:43
like a garage or a parking lot where you don't think about
30:46
having a cup of coffee while you're heading down to your own
30:49
car. So that's something, you know,
30:52
that's really interesting for me, kind of reimagining what
30:56
we're so used to. And we've been talking about
31:00
Out. How AI is sort of the next
31:05
electricity in the fact that so many of the industries has been
31:11
transforming so far. It happened because it's more
31:15
convenient for those Industries. They can more easily deploy
31:18
Point solutions that don't necessarily disrupt their like
31:21
more fundamental workflow. But I know your new book goes
31:27
into uncovering the disruption And he's can you think of an
31:32
example where this kind of more systematic disruption has
31:37
happened? Yeah, okay, so it hasn't
31:40
happened at that much yet, but here's, you're not old enough to
31:44
remember this, but some of your audience might be before the
31:48
internet. If you wanted to go look
31:49
something up. You would first go to the
31:53
encyclopedia that you may have had at home but failing that you
31:59
would go to a place called the library and the library still
32:03
exist, but you probably don't go to the library that often to
32:06
look something up. Okay, but he'll days.
32:08
That's what we had to do. If you want to know the answer
32:10
to something and it wasn't in the the pages, the encyclopedia
32:14
that you might have had at home, you go to the library to look it
32:16
up. And yet, I bet if you went to
32:19
those libraries, you know, 30 years ago, Walked up to the
32:24
head. Librarians of people who've been
32:26
thinking about the library's most and said, what do you think
32:30
the most exciting industries of the future are going to be?
32:34
My sense is not one of them would have said Library science.
32:39
They wouldn't have thought, you know what the future this
32:43
incredible commercial opportunity and lots of money in
32:47
library science. And yet, one of the most
32:51
exciting companies in the world today, Google.
32:55
Is a library science company, that's what they do.
32:58
They organize the world's information.
32:59
That's the the essence of their mission and they do that largely
33:04
through AI. A big part of how they do that
33:09
is by filling in missing information.
33:11
A search is a statement of intent and they respond to that
33:13
statement of intent. But helping you fill in missing
33:15
information through a i tools. So we've seen and it turns out
33:21
that once search is cheap enough, once we can respond to
33:26
people's queries once we can help people look stuff up, it's
33:29
worth billions and billions and billions and billions of
33:31
dollars. So that industry has been
33:38
transformed from being not even an industry at all.
33:41
In some sense in the 1980s, it was something we subsidized to
33:45
today, where it's an incredibly profitable Enterprise.
33:50
We can see a handful of other industries that got disrupted by
33:57
thinking through systems and not Point Solutions, okay?
34:00
I think personal Transportation over the last little while.
34:05
About a decade ago there were a handful of technologies that
34:10
were being developed that had the potential to change
34:14
Transportation Systems. So for example, one of them was
34:18
directions predictions that told you how to get from point A to
34:21
point B at a particular time of day the most efficient way
34:24
another was Digital dispatch. So from your mobile device, you
34:32
could order a car. And a third was prediction tools
34:39
on where there was a lot of pedestrian traffic that might
34:43
want to go somewhere by car and they were companies 10 years ago
34:49
that were commercializing each and all of those sometimes even
34:52
together. So there was a company that was
34:55
taking those predictions on where a lot of pedestrians were
35:00
and using it to advise taxi drivers on where they should
35:03
drive. In order to get the most fairs.
35:07
And that they sold it pretty successfully and taxi drivers,
35:13
used it and it made the less good taxi drivers to be a little
35:20
bit more like the best taxi drivers.
35:23
So it close the gap between the lower income in the higher
35:26
income taxi drivers by about 14%.
35:29
But it's a point solution, right?
35:31
It's helping, taxi drivers into the taxi industry and we had
35:34
similarly, we had companies that were providing digital dispatch
35:36
to taxi companies or directions to taxi drivers.
35:40
There was one company that figured out this at least one
35:44
company. There are a few that these
35:47
together were more than just point solutions that you could
35:51
create a new industry. And so Uber took these and other
35:55
Technologies and built a new system for transportation, this
36:00
wasn't about them just selling back into the taxi industry.
36:04
It was about enabling anybody effectively to be a taxi driver.
36:11
To provide other people with rides, you need a digital
36:14
dispatch. You needed the predictions from
36:17
how to get from point A, to point B.
36:18
Because otherwise, the first time you got into an Uber, and
36:20
the Uber driver didn't know where they were going, would be
36:22
the last time you got into an Uber.
36:24
So actually that, those, those eyes were fundamental to it and
36:28
you needed a way to tell the Uber drivers where to go in
36:31
order to get the most fairs and that was through all sorts of
36:34
creative Innovation on their pricing system.
36:37
So, it was almost the suite of Technologies led to a new system
36:41
that allowed them to really deliver value in a way that they
36:44
couldn't before. Absolutely.
36:47
And I think it's so important. I mean, it's this idea.
36:50
It's kind of how Innovation disruption words in a lot of
36:53
ways with things like the innovators dilemma and things,
36:55
like, and companies like that this idea of Point Solutions
37:00
giving us great lessons and finding ways to connect them in
37:03
finding ways to innovate, and finding ways to question the
37:06
value. Value of y, we innovate and I
37:09
think position machines and AI is such an important part of
37:12
that future. And that's one thing I loved
37:14
about your writing as well and I'm excited about your new book.
37:17
Palin prediction is because it's taking these really complex
37:22
kupets AI, economics position, Machines of complex math and it
37:28
it's it's taking it down to how does it actually work?
37:31
What are the steps? What is the basis of it?
37:33
Because I think this idea of Storytelling is actually, is
37:36
that In pop culture. When we hear a, I, we are
37:39
consistently told stories of what they determine 80 or like,
37:44
the base scary robots, right? And I think it's something that
37:49
is so interesting with your writing, is this kind of taking
37:52
it down to. This is how it works.
37:53
These are the steps it takes. And this is why it's so
37:56
important and why we have to get better at it.
37:59
So, I'm excited about your new book.
38:01
I guess that sometimes thank you.
38:03
I'm excited too. We're it's coming out November
38:06
15th. And we've got a fair bit of
38:10
attention already to it and we're looking forward to really
38:15
like, ultimately the goal is to accelerate the time till these
38:18
Innovations happen. And so the between times for
38:22
electricity and for computers lasted, 40 years, We've learned
38:27
a lot since then and there's reasons to be optimistic the
38:30
between times for AI. So far we're five years in, they
38:34
might only be five or ten years and that's that's our goal is to
38:38
as you think through what those new systems look like, how can
38:42
we build better companies in order to really deliver value to
38:47
change the way we live in the way we work?
38:56
To close out every one of our episodes, we have this segment
39:00
called suspenders and works like this.
39:03
We ask you a fun random question.
39:06
That's not related to anything whatsoever and you can give us
39:09
any answer. You feel like question of the
39:13
day is if you had to wear a hat for the next week.
39:18
What type of hat would you choose?
39:22
So my favorite hat is its baseball cap.
39:27
So I call the hats, I got a baseball cap, so we're limited
39:30
to that. And so I've won baseball cap.
39:37
That's like the National Bureau of economic research was like,
39:40
the economics organization that that I'm a part of, but I feel
39:46
like the next week that's not going to be my one.
39:47
The, the Hat I want is the creative destruction lab.
39:50
A, i Green hat. And that's to say the lab has
39:55
been so important to how I understand technological change.
40:00
Generally, an AI in particular, our book is coming out.
40:04
What do I want to let people know about and show off?
40:08
Is that through our lab? We've seen amazing technology
40:14
and to try to project that you know, as I talk about power and
40:18
prediction fashion Intellectual statement.
40:23
Love it. Welcome back to Top Hat.
40:33
This is the part of the episode where we dissect and analyze
40:36
some of the key. Learnings we got from this
40:38
week's expert Storyteller and this week, we had a pretty cool
40:41
way. I predicted machine.
40:43
Technical expert Storyteller, didn't recap?
40:45
Oh yes, dr. Avi Goldfarb is, obviously a
40:49
very accomplished and insightful expert in the field of
40:55
artificial intelligence. So we got to talk to him about a
40:59
lot of stuff involving. Being a I trying to demystify
41:06
what the concept is a little bit.
41:09
Yeah, for sure and I think one of the reasons we love talking
41:12
about emerging and saying, technology on this show is
41:15
because learning the stories and what technologies have the best
41:21
storytelling through buzzwords? Who articles Foo pop culture?
41:25
Often helps you tell what is going to be the future?
41:28
Because if enough money is excited about it and not Pause
41:31
this toy is happening. That's going to be the research
41:34
that's funded and if the research is funded, that's going
41:36
to push forward the mission and that kind of talks about Kevin.
41:41
This idea of demystifying nai-nai is talked about
41:44
constantly in pop culture and we talked about it with optical
41:47
fiber, little bit. This idea of like determinate,
41:50
ER, or like the evil Warlords or like the future of technology
41:55
and pop culture and science fiction really gets to explore
42:00
these worlds. Because it's Not real because
42:01
it's, it's one world so. But because all these really
42:05
exciting stories have been told about it, it kind of mystifies
42:08
what AI is and there's a lot of healthy skepticism, but that's
42:11
why I what I love about dr. Go pumps work.
42:14
Is that in his book prediction machines, in his new book, they
42:18
they take Ai and they bring it down to its core of position
42:22
machines. They explained that these are
42:24
machines that makes predictions and by we branding it, by
42:28
bringing it down to its core. It becomes a little bit more.
42:31
Audible to talk about and it allows us to focus in a little
42:35
bit more and kind of disconnected from some of this.
42:37
I terminate is that disconnected from some of this broad, deep
42:41
learning or broad known at things and focusing on one part
42:45
of artificial intelligence, that is making impact today and
42:49
allows us to understand the economics behind it.
42:52
Absolutely. And on the flip side of that, we
42:56
also got a bit into the importance of understanding your
43:01
Mission when your business trying to figure out how to
43:06
integrate AI into your day-to-day workflow.
43:11
And as powerful as the prediction machine is, it is
43:15
just a tool and it requires people in the business who
43:20
understand what, what your mission is, what business
43:23
problem, you're trying to solve to be able to figure out.
43:27
Okay, where can I apply a, i Into my workflow, where can I
43:33
plug this in machine in and also, how do I utilize this
43:37
machines to make sure that it'll make predictions that out?
43:41
Allow me to make more efficient decisions and mission, of
43:44
course, is a very important, very storytelling involved
43:49
process. So good, understanding of your
43:52
mission and your story is also crucial to knowing how you can
43:57
leverage the prediction machine. Yeah, it's seen the force behind
44:02
the trees, right? It's this idea that if you
44:05
remember what your mission is, is stay true to it.
44:07
It helps you avoid the airport scenario where instead of making
44:12
it a more efficient, you're making an airport, more fun, and
44:15
making it more nice and right. Oh, I airports great.
44:18
While you're waiting, you can do these ten things.
44:21
Most people would like, well, instead of waiting, I rather
44:24
just do those things at home and then be able to get on my flight
44:26
fast. Right?
44:27
It seemed the the for Fiona Trees and remembering that even
44:31
with exciting technology, even these buzzwords even these
44:33
short-term profit gains. One of the most key things is to
44:36
remember your mission and see how you can build that
44:39
technology in your story. So, remembering the story of the
44:42
technology. You might mean the story of your
44:44
company and saying, hey, is this just exciting?
44:47
Is this is something something talking about is this persuasive
44:49
story time? Or is this actually going to
44:51
help our core mission on for story?
44:53
But anyways, it was really interesting to talk to dr.
44:58
Goldfarb this Week to talk about technology.
45:01
How storytelling affects emerging technology?
45:03
Just talk about how storytelling affects businesses and how AI is
45:07
a disruptive technology that can really change the world?
45:12
And that's going to do it. This has been another great
45:15
episode of the linen suit and plastic tie podcast.
45:19
If you like our content, make sure to subscribe and follow us
45:23
wherever you listen. Also, feel free to leave us a
45:26
comment or review to let us know what you're thinking.
45:29
Follow us on Instagram at LSP, teapot LinkedIn lettuce and
45:34
plastic tie and we'll see you next time.