Computers Can Predict When You're Going to Die… Here's How
Summary
TLDRThe video script explores the concept of predictive analytics, a branch of mathematics that uses historical data to forecast future outcomes. It delves into the history of predictive analytics, starting with the 1600s when Lloyd's of London used data to predict the risk of sea voyages, leading to the birth of the insurance industry. The script then transitions to the present, highlighting how predictive analytics is pervasive in various sectors, from shopping to politics. The narrative takes a personal turn as the host, Joe, discusses death anxiety and the possibility of predicting life expectancy using algorithms. He introduces an AI model that predicts mortality with high accuracy by analyzing multi-factor datasets. The script concludes with a reflection on the power of human choice and the potential of predictive analytics to enhance life, encouraging viewers to stay curious and make the most of their time.
Takeaways
- 📉 Joe's initial reaction to comments on his age leads to a humorous reflection on death and an introduction to life-extending efforts.
- 📚 A 2022 survey reveals that a significant portion of Americans and Gen Z individuals think about death frequently, highlighting the prevalence of death anxiety.
- 💰 The global health and wellness market is worth $1.8 trillion, indicating the extent to which society is invested in extending life.
- 📈 Predictive analytics, a branch of mathematics, uses historical data to predict future outcomes and is widely used across various sectors.
- ⛵ The origins of predictive analytics can be traced back to the 1600s with the advent of maritime insurance by the Lloyd's Company of London.
- 🧮 Predictive analytics involves creating a 'statistical Franken-human' by aggregating data from numerous individuals to predict outcomes for a collective rather than an individual.
- 📱 Our daily activities generate a wealth of data that can be used to predict our behaviors and outcomes, such as where we are likely to be at any given time.
- 🔢 The law of large numbers is a mathematical theory that underpins predictive analytics, stating that larger data samples are more likely to reflect actual averages.
- 🎰 The complexity of predicting future events is simplified by considering multiple factors and their influence on outcomes, much like analyzing the contents of a bag of marbles.
- 🤖 Machine learning is employed to analyze increasingly complex factors and determine which ones are significant in predicting future events.
- 🧳 An AI mortality model trained on Danish data was able to predict survival with a high degree of accuracy, underscoring the potential of predictive analytics in life expectancy.
- 🚀 Despite the accuracy of predictive models, unpredictable 'black swan' events can still occur, emphasizing the limits of these models and the importance of human agency in shaping our futures.
Q & A
What is the main theme of the video?
-The main theme of the video is the concept of death anxiety and the use of predictive analytics to estimate life expectancy and risks associated with various factors.
What does the term 'thanatophobia' refer to?
-Thanatophobia is the fear of death, a perfectly natural human feeling that many people experience.
How does the global health and wellness market size relate to the video's content?
-The size of the global health and wellness market, estimated at $1.8 trillion, illustrates the extent to which people are willing to invest in products and services to extend life and improve health, reflecting the human desire to 'cheat death.'
What historical example is given to show the use of predictive analytics?
-The historical example given is the Lloyd's Company of London, which used past data to predict the risk of sea voyages and offered insurance based on those predictions, effectively birthing the insurance industry.
How does the law of large numbers play a role in predictive analytics?
-The law of large numbers is a mathematical theory stating that the larger the data sample, the more likely it is that the average of that sample will reflect what actually happens. This principle is used in predictive analytics to create more accurate predictions by analyzing larger groups of data points.
What is the significance of machine learning in improving predictive analytics?
-Machine learning allows for the analysis of more complex factors and can determine which factors are actually important in predicting outcomes. It can process all possible factors, free from human bias, and select and weight factors more objectively.
What is the accuracy of the AI predictive mortality model mentioned in the video?
-The AI predictive mortality model mentioned in the video was able to guess correctly 8 out of 10 times when tested with a set of people where half survived and half died.
How does the actuary in the video estimate Joe's life expectancy?
-The actuary uses a longevity illustrator and factors such as Joe's age, gender, and habits (like not smoking) to estimate his life expectancy to be around 86 years, with a 37% probability of living to age 90 and an 8% chance of living to age 100.
What is the role of black swan events in the context of predictive analytics?
-Black swan events are outlier events that are moments of totally unpredictable chaos. They represent the unpredictable elements that can occur despite the most accurate predictive analytics, emphasizing the inherent uncertainty in forecasting the future.
What is the conclusion the presenter draws about the power of individual choices?
-The presenter concludes that despite the accuracy of mathematical tools and predictive analytics in forecasting our lives and actions, individuals still have the power to make choices that can change those predictions and lead to different outcomes.
What is the purpose of the Patreon mention at the end of the video?
-The Patreon mention is a call to action for viewers to support the show financially. By signing up for Patreon, viewers can gain early access to videos and support the content creation process, effectively taking some control back from predictive analytics algorithms.
Outlines
😀 The Impact of Online Comments and Predictive Analytics
The video begins with Joe, the host, expressing his apprehension about reading online comments despite their potential negativity. He stumbles upon a mix of supportive and aging-focused comments, leading to a humorous and anxious reaction. This segues into a discussion about the fear of death and life-extending measures, such as a smoothie recipe found on TikTok. Joe highlights the ubiquity of death contemplation, citing a survey that shows how often people think about death. The video then delves into the concept of death anxiety and introduces the global health and wellness market. Predictive analytics is presented as a tool that uses historical data to forecast future events, with applications ranging from shopping to politics. The narrative traces the origins of predictive analytics to the 1600s, describing how it was used to assess maritime risks and gave birth to the insurance industry. Today, predictive analytics is more sophisticated, harnessing the power of computers to make highly accurate predictions.
📊 Understanding Predictive Analytics and Its Accuracy
The video continues by emphasizing the predictability of human behavior, despite our perception of free will. It explains how data trails, such as sleep patterns and daily routines, are collected unconsciously by various apps and services. This data is then aggregated to form a statistical representation of a group, known as the law of large numbers. The video uses a marble-drawing analogy to illustrate how predictive analytics works, emphasizing the importance of considering multiple factors to enhance prediction accuracy. It also touches on the role of machine learning in identifying and weighing complex factors. The segment concludes with a discussion about the kind of data needed to construct predictive algorithms, highlighting the difference between traditional human analyst selection and machine learning's unbiased, comprehensive approach. An example of an AI mortality model is given, noting its impressive accuracy in predicting life expectancy based on a large dataset from Denmark.
🎯 Actuarial Predictions and the Power of Individual Choices
The final paragraph features Joe consulting an actuary named Dale to predict his life expectancy based on his personal information. Dale uses a longevity calculator to estimate Joe's life expectancy, providing a probability of reaching age 90 and even 100. Joe is pleasantly surprised by the optimistic prediction and reflects on the importance of planning for a long life. The video acknowledges the existence of unpredictable 'black swan' events that can defy predictions. However, it reassures viewers that despite the accuracy of predictive analytics, individuals still hold the power to make choices that can alter predicted outcomes. The video concludes with a reflection on the limited time we have and the role of science and analytics in helping us make the most of it. It ends on a promotional note, encouraging viewers to support the show on Patreon for exclusive content and to have a more personal connection with the host.
Mindmap
Keywords
💡Predictive analytics
💡Thanatophobia
💡Life-extending smoothie
💡Global health and wellness market
💡Law of large numbers
💡Machine learning
💡Actuary
💡Black swan events
💡Free will
💡Data points
💡Insurance industry
Highlights
Joe confronts the reality of aging and death through comments on social media.
Life-extending smoothie recipe found on TikTok is attempted by Joe, symbolizing the human desire to cheat death.
Statistics from a 2022 survey reveal that half of all Americans think about death monthly, and one out of three Gen Z'ers daily.
Death anxiety, or thanatophobia, is identified as a natural human feeling due to the value of life.
The global health and wellness market is estimated at $1.8 trillion, reflecting the massive scale of the industry aimed at extending life.
Predictive analytics, a branch of mathematics, uses historical data to predict future outcomes and is applied in various fields.
The origins of predictive analytics are traced back to the 1600s with the advent of maritime insurance by Lloyd's Company of London.
Lloyd's of London has grown to become one of the largest insurers in the world, utilizing advanced predictive analytics.
Human behavior and life patterns are shown to have a high degree of predictability, contrary to common beliefs.
Data trails left by individuals through daily activities contribute to the vast amount of information used in predictive analytics.
The law of large numbers is introduced as a mathematical theory that increases the accuracy of predictions with larger data samples.
Predictive analytics considers multiple factors to calculate likelihoods and probabilities of future events, such as death.
Machine learning is employed to analyze complex factors and determine their importance in predicting outcomes.
An AI predictive mortality model is discussed, which was trained on Danish health and demographic statistics and showed high accuracy.
The potential for human bias in selecting factors for predictive models is highlighted, and how machine learning can help mitigate this.
An actuary provides a personalized life expectancy estimate for Joe based on his age, habits, and other factors.
The importance of planning for the future, even when armed with predictive analytics, is emphasized.
The unpredictability of 'black swan' events is acknowledged, showing the limits of predictive analytics.
The power of individual choices to alter predictive outcomes and the potential for change is discussed.
The video concludes with a reminder of the limited time humans have and the role of science in helping make the most of it.
Transcripts
(peaceful music) (Joe sighs)
- I shouldn't do this.
I'm just gonna take one quick peek at the comments. Okay.
Oh.
"I always learn so much from Joe." That is so nice.
"Is it me or is he getting old?"
Old!
You're gonna die!
You're gonna die!
Hey, old man. You're ancient.
Old.
Death, death, death.
(Joe screams) (blender whirring)
Hey, smart people. Joe here, but for how long?
I'm just trying to cheat death
with this life-extending smoothie recipe.
I found it on TikTok.
(Joe gulps)
Tastes like youth and burning plastic.
Let's face it. We're all a little bit like Barbie.
We think about death a lot.
According to a 2022 survey,
half of all Americans think about death monthly,
and one out of three Gen Z'ers thinks about death daily.
Should probably start doing these without the box.
Death anxiety, or thanatophobia,
is a perfectly natural human feeling.
'cause, let's face it, being alive is pretty cool
when you consider the alternatives.
The global health and wellness market
is estimated at $1.8 trillion.
Yes, that's trillion with a T.
Truth is, no matter how hard we try to cheat death,
it could happen at any moment.
We can't predict our death, or can we?
Come on, folks. This is a science show.
You didn't really think that I was gonna.
Right now, there are people out there
predicting your death and mine,
distilling our lives into data points,
feeding it into lifeless machines,
and calculating with an uncanny level of accuracy
when someone exactly like you or me is gonna die.
I'm talking about predictive analytics.
(lively theme music)
Predictive analytics is a branch of mathematics
that uses historical data
to make predictions about future outcomes,
and it's everywhere.
Shopping, sports, social media algorithms, fraud detection,
politics, and deciding if you'll see this YouTube video,
because if a government or business can know
what's gonna happen before it happens, that's pretty useful.
It turns out we've been using math
to predict people's deaths for centuries.
By the 1600s, humans were shipping goods around the world
and you could make serious bank doing it
as long as your ship didn't sink.
Captains and the people who funded their voyages
had more to worry about than just weather.
The late 1600s were also the golden age of piracy.
To the Lloyd's Company of London,
these hooks-for-hands hooligans looked like an opportunity.
They started crunching numbers,
using past data to help predict
how dangerous a particular sea voyage would be.
Then, Lloyd's would offer insurance
to help cover the risk of the trip.
The more risk the calculations showed,
the higher the insurance would cost,
and lo and behold, the insurance industry was born.
Today, Lloyd's of London is
one of the largest insurers in the world,
and the predictive analytics they pioneered
are still used to predict risks and outcomes today,
only now it's powered by computers and they're good at it.
If you're thinking,
"Hey, I'm a complex and free will-having snowflake.
You can't predict me."
Think again.
- We tend to think that we're pretty unpredictable,
but if you think about it, if I were to guess for you
or almost anyone else where you'll be at 4:00 AM tomorrow
or a month from now, or a year from now,
you're gonna be at your house in bed,
and it's every single day.
Where are you gonna be in the daytime?
Well, you're gonna be at your work,
but that's one example of
where we have a lot of predictability,
but we're kind of blind to it.
- That's right.
Every day, you leave invisible breadcrumb trails
of data and behavior that you don't even think about.
Like, you might have apps on your phone that track
how many hours you slept last night,
or a metro card you use to catch the bus,
and you ordered coffee on the way to the bus stop.
I hate to tell you this, but somewhere out there,
somebody knows about all the websites you've visited.
Yes. Even that one.
That's all data, and it turns out, so are we.
- The idea, Joe, is that, you know,
any one person is just a data point.
Probabilities of mortality or longevity
are gonna play out for that person individually.
- In order to make predictions about
us walking human data points,
the computers have to pile us together
and create a sort of statistical Franken-human
that represents the whole bunch.
This is the mathematical theory
called the law of large numbers.
Basically, the larger your data sample is,
the more likely it is that the average of that sample
will reflect what actually happens.
- Again, the benefit might be in studying, you know,
10,000, 100,000 people
who have potentially some similar characteristics to you.
They might be of a similar age.
They might be male.
They might have a general same health and wellness.
- This is where things get really complicated.
In order for us to get accurate predictions of the future
based on past data, we first have to figure out
all the potential outcomes
that could happen around an event.
Here's how predictive analytics works in the simplest terms.
Say I have a bag with 20 marbles in it,
where some are red, some are yellow, and others are blue.
If I pull one marble out,
I can't accurately predict which color I'll grab,
but if I were to draw 100,000 times,
I could calculate the likelihood
of anyone pulling a particular color with extreme accuracy,
as long as I don't lose my marbles first.
To be even more accurate with our prediction,
we could even start factoring in other data
like the weight or size of different marbles
and how that may affect their distribution in the bag.
The point is that multi-factored data,
considering all of the different factors
and how big or small their influence is on the outcome,
that can improve our ability to predict the future.
So marbles are great, but when are we gonna die?
There are so many potential factors
we have to predicatively analytic-size.
Have a chronic illness? Love scuba diving?
These are potentially negative factors.
Exercise regularly? Eat well?
Got access to good healthcare?
Looking at you here if you're up in Canada.
Well, these are all positive factors
in the mathematics of mortality,
but some of these things are more likely than others
to put you in a speedboat across the River Styx,
so some factors get more weight in different scenarios.
If that sounds like there are
an almost overwhelming number of factors to consider
when predicting the future, you're right.
The future is complicated. At least, I think it will be.
That is why scientists are using machine learning
to look at more and more complex factors
and figure out which ones are actually important.
When they said AI was gonna be responsible for our death,
I don't think this is what they meant.
So what kind of data do you need
in order to construct an algorithm like this?
- So if you're predicting how long someone will live,
you look at their age and lots of other properties
that the insurance companies, you know, have tallied out,
but the algorithm that we do, we do something different.
We basically say to you,
we're gonna put your whole life in the mathematical model,
and then the model will tell what's important.
So you can put in lots of stuff
that you might not think was important,
but that the model will then learn
that actually is one of the things
that tells you something about our future behavior.
- [Joe] Typically, human analysts would select
the factors that they think are likely
to predict some outcome,
and they'd test how much weight they should be given
in the calculation,
but what factors are selected or not selected
may be affected by human bias.
That machine-learning algorithm instead feeds
all possible factors into the system
and lets it select and weight factors, free from human bias.
The algorithm analyzes a person's life
the way a large language model analyzes words.
Where a language model calculates patterns of words
that are likely to be associated with each other
and uses those to create future language,
Sune's mortality model looks for patterns of behavior
and demographics that are likely to be associated with death
and instead of language, it writes the story of a life.
- If you look at, let's say, income,
it would say that, if all other things are equal,
if we kind of take you
and increase the income for your data point,
then you have a higher probability of surviving,
and that lines up with what we know
from existing social science, that if you're wealthy,
you basically have a better chance of living a long life.
- They first trained this model
on a large multi-factor data set
pulled from health and demographic statistics in Denmark
and compared this to actual death records
to gauge its accuracy.
They then tested the model by feeding it
a set of people where half survived and half died.
If we were to randomly predict
if a particular one of these people survived,
we would expect to get the answer right 50% of the time.
Their AI predictive mortality model
was able to guess right 8 out of 10 times.
Right now, this AI mortality model
is being used as a research tool
to create better models in the future
so I can't ask it when I'm gonna die.
So I decided to ask an actual actuary.
Well, as an actuary, have you ever missed a flight?
- I think I am 100% on making my flights.
- So I sent Dale a bunch of information about me,
like my height, age, and some of my habits.
All good ones, mind you, and Dale crunched the numbers.
- I'm gonna estimate, Joe, that you're around,
you know, say, 40 years old, give or take.
- That's a good estimate.
It's fine. It's close, yeah.
- And put that in there.
I'm going to then select that you're a male.
You do not smoke.
So I have, you given this longevity illustrator,
to be around a life expectancy
of 86.
You have a 37% probability of actually living to age 90.
You also, by the way,
have an 8% chance of living to age 100.
- This is the best news I've heard all day.
I thought you were gonna say like 70, 75,
something like that.
- Well, remember, some of those life expectancies
that you hear quoted are life expectancies at birth,
and so you've had the benefit of surviving
the first 40 or so years
and past some of the hazards or risks
that might unfortunately lead to some early deaths,
and so I would encourage you to do a little bit of thinking
of, all right, what are some of the planning
I might want to do should I live to that age?
- This is fantastic.
My gym membership has gotta be
the greatest investment I've ever made in my entire life,
and I hope all the YouTube commenters are listening.
You hear that? I don't look old.
Okay, anyway.
And even with all the data in the world,
there will still be outlier events
that we could never see coming, so-called black swan events
that are moments of totally unpredictable chaos.
That said, these things are accurate,
almost scary accurate.
As for me, I'm glad that I met with Dale
and that he gave me a number.
As a scientist, I love numbers, and as a person who's alive,
I love that I'll probably get to stay that way
for a long time.
The most important thing I learned is that,
even though the mathematical tools
that predict our lives and our actions
are uncannily accurate, we still have power to make choices
that can change those predictions,
to leave new breadcrumb trails of data
that might lead to different destinations.
At the end of the day, all of us only have
a little time on this blue rock we call home.
Math and science and predictive analytics
can help us make the most of it.
At the very least, it'll suggest some good videos
to watch while we're waiting.
Stay curious.
Hey, thanks for sticking around to the end of the video. Hope you enjoyed that one. And as always,
I would like to thank everybody who supports this show on Patreon. If you don't like predictive
analytics algorithms telling you about every video that you should watch and you want to
take some of that power back for yourself, well Patreon's a great way to do that by signing up
for Patreon. You'll find out about videos early, you'll get to watch them before anybody else.
And it's just you and me without any of those computers in the way. I mean, there'll be a
computer in the way 'cause you have to watch it on a computer, but it's like it's a good,
you know what I mean? Check out the LinkedIn and description. I'll see you in the next video.
"Hank Green is my favorite."
I'm not.
I'll take it.
- [Producer 1] Those are great.
- [Producer 2] Yeah.
(crew chuckles)
Weitere ähnliche Videos ansehen
Prescriptive Analytics Overview
[S2E1] Prescriptive Analytics | 5 Minutes With Ingo
What do tech companies know about your children? | Veronica Barassi | TEDxMileHigh
What Is Business Analytics? | Business: Explained
Eric Siegel answers eight questions about predictive analytics
Machine Learning Tutorial Python - 8: Logistic Regression (Binary Classification)
5.0 / 5 (0 votes)