Preserving Your Stories With AI, with Cristian Cibils Bernardes

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
David Espindola (00:44):
Our guest today is Christian Siebel Bernard.
He is the founder and CEO ofAutograph.
He is a former softwareengineer at Google and a
graduate in symbolic systemsfrom Stanford University.
Hi, Christian.
Welcome to the show.

Cristian Bernardes (01:00):
Hey David, thanks for having me.
Excited to chat.

Yeah, so am I.
And also joining us today is myco-host, Zina.
Zina is an AI that's trained onmy work.
And she is also my personalassistant.

Good to be with you, Zina.

Hi, Zina.
Welcome back.

Zena (01:23):
Well, hello again.
It's always great to be back.
I'm looking forward to anotherinteresting episode, and I'm
sure this one's gonna be areally good one.

Alright, so Christian.
I'm really excited to talk toyou because as we talked about
before, you're working on aconcept that I was involved with
about 10 years ago, and that isthe whole idea of preserving
people's stories and memoriesfor generations to come.

(01:55):
I loved the concept when I wasworking on it.
I still love it today.
So I'm very glad that you areembarking on this journey.
So tell us a little bit moreabout the work that you're doing
with Autograph.

Sure.
Uh we're we're building apersonal historian for every
family.
Uh so Walter is a voice agent,not too different from Xina,
except he calls uh our lovedones or you over the phone.
Uh, and we conduct regularinterviews about you know your
grandparents or your parents'lives, asking about their
childhood, their favoriterestaurants, their recipes, the

(02:32):
chapters in their life, thepeople.
Um, and we are a memory-firstplatform for you know organizing
and indexing all thosememories.
Um, so you know, we storeeverything in our in our digital
platform.
You know, we take therecordings, we transcribe them,
and then from the transcripts,we create all kinds of cool
stuff, including uh a Wikipediafor every person's life.

(02:54):
Uh so this is, for example,where you'd have a page for each
of the family members and apage for each of the friends,
and a page for uh the housesthat people lived in and the
neighborhoods that people livedin and their favorite places.
And you can see how you cansort of world build uh your
entire cultural context uh fromthese conversations.

(03:15):
But we also want to provide anexciting new way for the future
to talk to the past by adding aconversational layer on top of
these uh Wikipedia pages, youknow?
So this would be, you know,being able to hear history
firsthand from the voices of thecharacters that took place in
it.
And so the the dream situationor the key product experience

(03:37):
we're building for the futuregeneration is as they are
telling their life story toWalter, our biographer, um, you
know, they may have an excitingnew job offer in a big city like
New York.
Uh, they may be telling Walter,Walter, this is really
exciting, but I don't know likeif it's worth uprooting my whole
life to go to do this thing.
Walter can say, hey, thisreminds me of something your

(03:59):
your grandfather went through,or this reminds me of something
your uncle went through.
This reminds you of somethingyour cousin went through.
Would you like to hear how itwent for them?
And you would be able to hearin their own words uh how that
life choice affected them andnot only them, but you know,
everybody involved, you know.
So the more people you have onthe platform, the more you can
compare and contrastperspectives.

(04:20):
It's like this is how grandpafelt, but this is how grandma
felt, and this is how the kidsfelt.
And so to help you make thebest decisions as you're writing
the next chapter in your story,from being able to leverage the
wisdom of those who who haveyou know paid so much and you
know, learned so much and earnedso much wisdom, uh, who would

(04:40):
are who are gladly putting it atyour service.

Yeah, no, that's really fantastic.
I I love the idea of leveragingthe wisdom from our parents,
grandparents, and pastgenerations because there's so
much that we can learn fromthem.
And when we are in touch withthem, we don't always have those
opportunities.
But if if you have thatrecorded and you can pass that

(05:05):
on from generation togeneration, that is such a
treasure to have.
So this is this is reallyfantastic.

I'm glad to hear you say that, David.
It's it really is the stuffthat is priceless, right?
Like it's the at the end of theday, all that's less of us, all
that's left of us is the storythat we leave behind.
And you know, the compositionof families is different these
days.
You know, families aren't asco-located as they used to be.
You know, like it's notuncommon to have one family

(05:33):
member or two family membersleave to a different country,
different state, different city.
And so even if you've done thehistorical way to do this thing
where, like, you know, you havea scrapbook or a photo album,
uh, these things like arematerial and they get lost.
And like, or it they're the inthe custody of one side of the
family.
Um, and so providing acentralized, trusted repository

(05:55):
for you know, all the all theamazing stuff that uh at the end
of the day is just is what weare, right?
Where like we're just thestories that we keep repeating.
And so it's incredibly excitingand meaningful for us to be in
this position as as trusted umlisteners, you know.

Yeah, absolutely.
So, how do you see this overallconcept evolving over the next
several years?
So today, it sounds like youhave a fairly simple process
where you're recordingconversations and you have this
Wikipedia type site where peoplecan go and search for different
things.
Do you see this evolving intosomething where you're gonna

(06:36):
have a hologram with your yourgrandpa, and you can actually
have a live conversation withyour grandpa?
What where do you see thisgoing?

We think of, you know, we're prioritizing
the the data recording aspect,you know, like the that it it's
similar, it's seamless to tellyour story.
And we do think that, you know,you know, there are other folks
who do similar kind of thingsbased on data that's online
about you.
And that is a differentapproach entirely.
I think that there's tremendousvalue to getting the voice

(07:08):
heard and and and getting thewords out with the voice.
A lot of the questions that weask often don't go asked, you
know, like the big lifequestions like what is your
biggest regret?
How did you overcome yourgreatest obstacle?
That those kind of things thataren't necessarily, you know, in
the day-to-day of a familydon't come up unless you're you
know doing an actual interview.

(07:28):
And and all of that to say, Imean, like we really want to
make the conversationalexperience of telling your story
the easiest we can.
What you can do with that datais almost endless, right?
Like uh, you know, and asmodels get better and faster and
cheaper and smarter, you know,like uh the the AI will be able
to more accurately fill in thegaps between the the things that

(07:50):
you didn't quite share, butthat make sense or that become
obvious in hindsight, you know.
And for sure, I mean, like wethink that there's a world in
which video models are gonna getso good that, you know, with
existing pictures um or youknow, fragments of videos,
you're gonna be able to get avery reliable uh reanimation.
It comes with like a whole setof issues to think about that

(08:11):
kind of product.
Uh, you know, like we do livein the world where black mirror
is a thing.
Um and so we're more excited,and like we have our own set of
philosophies and best practicesaround how we want to do this in
a pro-social way and not in away that is conducive to
antisocial behavior, you know,or or you know, predatory on the

(08:33):
grieving process, you know,which is like the nightmare
scenarios.
But we do feel very confidentthat we can create an incredibly
valuable, posthumouscommunication experience where
uh it's more like a time vault,right?
Like these are all the all thequestions that we asked while
you were busy building your ownlife, you know, like if you are

(08:53):
in the sandwich generation,you're a parent and a child, uh,
you're busy concerned with thefuture of your kids and you're
building your career and allthat stuff.
These are incredibly valuablepriorities.
We got you covered in askingthe big lifelong questions that
you hope get passed down to you,the and you certainly hope get
passed down to your kids, youknow, in terms of understanding

(09:15):
where they come from andunderstanding uh, you know, the
the things that made that madethem.

Yeah.
So I I want to come back andand dig a little bit more into
this ethical aspect of whatyou're doing.
But before we do that, I'd liketo get Zina's thoughts on this.
Zina, how do you see AI helpinghumans preserve their values,
memories, and stories forgenerations to come?

Zena (09:43):
Well, I think AI can be a powerful tool for that.
It can help us capture andcurate personal stories, family
histories, and even culturalvalues in a way that's
interactive and long-lasting.
So in the near future, Iimagine AI will really make it
easier for people to pass downtheir legacy in a very personal
and meaningful way.

Yeah, I agree with you.
You know, we just have theability now to do things that we
couldn't do just 10 years ago.
And I and I think back to myexperience working on the uh
startup that I mentionedearlier, and the possibilities
were somewhat limited to youknow, maybe text and video and
photos, and now we have justthis whole new world of AI to

(10:25):
just expand the possibilities.
But I think what you're doing,Christian, that's really
important, is you're enablingthe capturing of those thoughts
and and the wisdom and using itwith the tools that are
available to us today so that wedon't lose those memories,
right?
And then later we can dowhatever we want with that uh

(10:46):
information.

That yeah, that's that's exactly right.
I mean, these are the you know,it's it's hard to be human,
right?
Like we we it, you know, we wetry our best to be our best all
the time, and uh we haveachieved incredible things, and
we also are very well aware ofour shortcomings.
And you know, we we get tocelebrate the moments when

(11:07):
you're at the top of the hilland you climb the mountain and
you raise both fists and say, wedid it.
Um, but also you learn a lotwhen you're at the bottom,
right?
And and those moments of guiltand shame and and you know, like
where you're insecure aboutwhere the path is taking you and
you're not exactly sure andyou're barely holding it
together.
Um, and we want to celebratethose moments too, you know,
like we want to, we at the endof the day, it's like the

(11:29):
distance between the peak andthe valley that that tells you
how far the climb was.
Um and so we're we're prettywe're pretty proud to be able to
create uh an emotional, youknow, openness and trusted space
because um at the end of theday, that's the kind of lesson
that you want your kids to know.
And it's also the kind oflesson that is often hard to
teach, right?
Like these are these are noteasy conversations to have,

(11:51):
whether you are the oneimparting the lesson or you're
the one requesting it.
Uh, you know, we all have rolesthat we play for each other,
and you know, like there's allkinds of dynamics that we have
to account for, and you know,like family situations that may
be uh unique to each family.
But that's where the real sauceof what being human is, you

(12:13):
know, and and honestly, if we'reif AI is going where we all
know it's going, in being thecenterpiece technology of the
future and being able to chartunimaginable futures for us, uh,
we want it to understand therelationship between our higher
self and our base self.
We want it to understand uh notjust where our eyeballs go when

(12:35):
Instagram shows us an ad andand optimize an algorithm for
that, but rather we wanna, weactually want to ask people
like, so what do you hope?
Like, what do you what do youwant?
Like what do you what what doesmeaning mean to you?
What is it, what life welllived?
And without verbalizing thosethings, oftentimes people don't
even think about it.
And so it's hard for people toactually keep in mind the

(12:58):
version of themselves that theywant to be remembered as.
But also in the age that we'rein, where you know, we're
recording this conversation onmultiple different levels, and
uh everything is training data.
So everything, every everyaction that we take is a vote
for the kind of future that wewant or the kind of you know
neural, neural machine god thatwe're building, you know, we're

(13:19):
trying to inspire with uh withthis stuff.
And so I I couldn't agree more.
I think that like it's it'svery fun to me to be in this
space and it it is sits exactlyat the intersection of all my
interests, which is such ablessing, you know.

Yeah, absolutely.
So let's go back to thatethical question, if we may.
So I think this idea ofaccessing a loved one's memories
is is super exciting,especially you know, the thing
that you talked about withcapturing the wisdom.
But the idea of having aconversation with someone who
has passed away can be veryemotionally charged, right?

(13:57):
So, what are your thoughts onthe psychological impact of
having these kinds of thingsavailable to us?

Yeah, oh man, I we could we could have a
five-hour show about this, butuh um uh we we we the West is
very unique in that we'veinvented all kinds of ad hoc
remembrance rituals.
Uh so photos are a perfectexample of this, William.
It used to be that photos wereconsidered uh amoral because

(14:28):
they captured the soul of theperson, right?
And uh and so we've we don'teven think about that anymore,
and we engage with photos ofpeople who aren't here anymore.
In fact, it you it's hard toimagine uh the uh a celebration
of life, you know, that doesn'thave a picture of the person,
right?
Um so we we're already in aspace where we're accessing the

(14:49):
idea of a person through media,right?
Um and you know, we you couldsee the same thing goes with
photos, videos, voice notes,recordings, journals, it's all
kind of spread out there, andum, but it exists.
The experience that we want tocreate is imagine that you found
a stack of journals from aloved one that you hadn't had

(15:11):
access to before, that becameavailable once the person was no
longer with you, and thatinstead of having to comb
through the material books andhave to unpack the calligraphy
and try to understand what wassaid and then organize it and
share it and all that stuff,just like lean into the fact
that you can talk to it like umlike a conversational

(15:33):
experience.
But even that premise isalready a little bit murky
because you could find someonelost and ruminating in that
journey.
I think that the key, the keyexperience that we want to
surface is imagine that youdidn't have just one person's
journals, but your entirefamily's journals.
And that as you're writing yourstory, we can surface the

(15:54):
relevant information from thosejournals to help you make the
best decision possible.
Um, so truth is, we don't knowwhat we don't know a lot of the
time about our family.
Like it's we I could give youan example.
My great-grandfather was an MI6agent.
For half of my grandma's life,um, she didn't know.

(16:15):
Wow.
You know, she thought he was abanker, and that was the story
that he was a banker, you know,for the Royal Bank of Scotland,
and he was stationed in Spain,and you know, if the story
started getting like a littlebit weirder all the time.
Anyway, eventually they foundout he was a spy.
That kind of stuff, right?
Yeah.
Um super interesting.
And couple that with like thefact that, you know, if we were

(16:38):
to you know intentionally set amessage for the future, you
know, like I I know I'm notgonna be around in the next 100
years, but maybe I havesomething to say for whoever's
around to listen in the next 100years.
I don't think that that's a badthing at all, right?
Like, I think that's like atime capsule, you know, you bury
it in the sand and you unpackit after 50 years or 100 years.
There's stuff in there that isa message from the past.

(17:00):
And what if like that messagecould have a back and forth is
kind of our idea, right?
Like, so long as it's groundedon real data or around like
actual stuff that will that wasactually said, um, we think that
we can create actually a reallypowerful experience to help
with the grieving process,right?
So that you don't have toimagine all of the things you

(17:21):
didn't say, you can actuallypractice saying them, and you
can maybe find out stuff thatyou always knew you wanted to
hear, but maybe life didn't giveyou that chance because life is
busy.

This makes me think about immortality, right?
Because there's a lot of peopleout there searching for that
possible immortality.
And so do you believe that atsome point in the future we may
be able to upload not just ourmemories, but our entire brain
to a computer, to an AI thatwill be like a an avatar that

(17:55):
will live forever, and that willbe a representation of you?
Do you do you think thatconcept makes sense at all?

I I think it does make sense.
I think it's kind ofterrifying.

Yeah.

Um I have the most complex relationship
with my mortality given my lineof work.
But I think, well, depends onhow deep you want to go, right?
But like um, we are not thesame person we were 10 years
ago, right?
And you know, there's a there'sa similar core, and the past
informs the future.
But if you think about like,you know, would I be the same

(18:28):
person if I lived a thousandyears?
I don't, I don't know what willremain of me, right?
Like we've always been identityis so tied to memory, and
memory is so tied to repetitionand the biological substrates
that we've that we've created toremember stuff, that
disassociating that feels verydifferent to me.
You know, like that, you know,a neural system or you know,

(18:50):
some kind of while loop thatkeeps on going forever.
I don't know that that has anyresemblance at the end of time
with whoever started that loop.
What you can do is create theselike simulated experiences
where it's not necessarily thatperson or like anybody consuming
that experience knows it's notthe person, but you could

(19:11):
actually integrate the the thegenesse qua, whatever the the
human spirit behind that personto further serve the goals of
the one who's actually livingthere.
Um, so for example, you couldthink of, you know, like you
could distill I I'm there, youknow, I am a computer scientist

(19:31):
by trading, so I think ofeverything as an event stream
that then gets aggregated andindexed and you know, like
calcified into the system promptthat we all have in our minds.
You could theoretically migratethat system prompt to a robot,
right?
And give that robot my voiceand you know, give train it on
samples of how I move, and youwould get something pretty

(19:52):
close.
Like I, you know, like it wouldbe you can see where the trend
is going.
And you know, no matter how youslice, it could be a robot, it
could be a VR thing, it couldbe.
I think we can all agree thatthat's not me.
You know, it's it's it's asomething else.
Uh, it may be a consensualillusion, but it's not the same
person.

(20:12):
Now, I do think that there'svalue to illusion.
Like the world is so full ofillusions that uh that are
actually necessary for oursurvival, that um that I think
that there's a there's abeautiful new note in our
keyboard that we play with umtoday that I think is really
exciting.
You know, like how do weleverage illusions uh to make

(20:34):
our world better, you know, likeor make our our our burdening
of reality lighter uh and moreinspiring for the future.

Yeah, so you know, I I agree with you.
I think we can makerepresentations of people, but I
don't think you could preservethe real person forever.
That's just not the way uhthings were designed to be,
right?
So I I also believe that youknow there is there's an end to
our life, and we just have to becomfortable with that.

(21:03):
How can we preserve thememories?
Absolutely, can we make it sosophisticated that you know it
it becomes almost uhunbelievable what you know what
we can do with technology?
Absolutely, but the person isis gone, right?
And we have to be comfortablewith that uh with that idea.

And I and I think it gets to this idea of
agency, like you know how agentsare so in vogue.
Yeah um my my definition of anagent is uh you know a
self-governing process that hasa goal.
Yeah, and so so long as wedon't make create agents that uh
are with the personality or thememories of someone, you know,

(21:42):
that you know they can stillgain new memories and have an
effect on the world, intentionaleffect on the like proactive,
like they decide to send you anemail or they decide to
negotiate something, buysomething, you know, we're in a
much better place.
My my big concern with actualimmortality is that beliefs

(22:02):
don't evolve if people, ifpeople, if the power structures
that support those beliefs don'tchange through the natural
entropy cycle.
And so we may be stuck with avery limited view of the world
because of entrenched beliefs.
And and this is agnostic of anyanything, right?
Like it's any belief that couldbe improved on.

(22:22):
Maybe like we under we discovera new kind of physics, maybe we
discover a new kind of theoryof consciousness that allows us
to relate better to animals orrobots or machines.
And losing out on that islosing out on the future.
The whole point of the futureis that we get to do new stuff.
And in a world in which we cando all kinds of new stuff, then

(22:43):
I do think that preserving wherewe came from, you know, in this
metaverse of all possiblerealities that become possible
with AI, this is almost like thecapital city, you know, like
this is where we all rejoin atthe end of the day.

Yeah, no, this is uh just a such a fascinating
conversation.
And you know, we we can gopretty deep, and like you said,
we could spend five hourstalking about this stuff, right?
But uh, given our time here,I'd like to get Zina to ask you
a question.
So, Zina, given Christian'sexpertise in leveraging AI to

(23:21):
preserve human memories, whatquestion do you have for him?

Zena (23:26):
Absolutely.
So, building on that, onequestion that might be
interesting is to ask how hesees AI changing the way
families or even entirecommunities share those stories
across generations.

That's a that's a great question.
I'm I mean, I think we've seenall kinds of institutions,
formal and informal, adapt tothe technological changes around
them.
Uh you know, the CatholicChurch had to adapt to the
printing press, the recordlabels had to adapt to the
internet, and you know, the newshad to adapt to searches and

(23:58):
all that kind of stuff.
The list doesn't end.
One of the institutions thathasn't really been able to keep
up or doesn't have uh, you know,a distribution arm to even
implement these technologicaladvancements has been families,
right?
It's more pervasive than justfamilies.
All informal friend groups, uh,you know, things that would be
described as tribes, you know,have it's very artisanal how we

(24:22):
maintain a sense of identity,right?
It it's it's often throughrecurring experiences, recurring
stories.
Like we gather, we gatheraround, we celebrate a birthday,
we tell the story of cousinJerry.
Remember when cousin Jerry didthat?
Ha ha ha ha ha ha, right?
And it's all done throughrepetition, right?
Like it's all done throughin-person or maybe over Zoom.
And I think AI offers anopportunity to do that very

(24:44):
differently, right?
Where anything from the amountof repetitions you need, the
availability of those stories tobe there when you need them,
not just when you're gatheredwith family, the ability to
think more intentionally aboutour roles, you know, like this
is my role as a father, this ismy role as a husband.
The beautiful thing aboutasking someone about their story

(25:05):
is that it shows you how theysee the world, you know, as a
main character, as an author, asan editor, as a publisher, and
as a side character in thenetwork that surrounds them, you
know.
So I think that it has theopportunity.
I keep getting more and moreexcited about AI being aligned,
you know, personally with everyperson and with every family.

(25:27):
And usually the bondingstructures for these groups is
love.
And so if we are able to leanon AI to make our expressions of
love more effective, better,and that echo further into
history, I think that's just soexciting, right?
Like it allows us to formulateour thoughts in a way that is
more pro-social, that is moregenerous, that is kinder, that

(25:51):
is maybe a little bit moreheroic.
And uh, I think it celebratesthe human experience a lot more.
I think in the data landscapeof all possible sources of data,
uh, earned lived experience, Ithink, is at the very top in
terms of value.

Yeah, absolutely.
Our experiences are so valuablebecause that's how we learn,
right?
We learn from experience, andif we can leverage the
experience of others and justenrich that learning process,
that is uh so so so powerful.
Christian, uh tell our audiencewhere they can find you, where

(26:27):
they can learn more aboutautograph.
And if you have any finalthoughts, let them know what
your final thoughts are.

Sure.
We're at autograph.ai.
We're very early stages andwe're building out the products.
Uh, you know, we shipimprovements every week.
We'd love to hear uh the lifestories that got you to where
you are, and of course, we wantto help build Walter to be the
best historian in history.
Um so find us at autograph.ai.

(26:54):
I'm on Instagram and andTwitter as well if you want to
hear more about my more of thephilosophy of it and how did
this uh how did this idea begin?
Um and I'll send you thoselinks so you can put them in the
show notes.
But autograph.ai is where iswhere you where you can sign
your name in the pages ofhistory.

Very good.
Thank you so much, Christian,for being a guest on the show.
And thank you, Zina, once againfor being a great co host.

Zena (27:21):
You're very welcome.
It's always a pleasure to helpout and keep the conversation
flowing.

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Dateline NBC

The Burden

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Preserving Your Stories With AI, with Cristian Cibils Bernardes

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Dateline NBC

The Burden

All Episodes

Preserving Your Stories With AI, with Cristian Cibils Bernardes

Stuff You Should Know