Latest / Chief in Tech Deep Dive Shortcast / Chief in Tech S1E16 - Daniela Barbosa : Authenticity in the Age of AI: A Guide to Decentralized Identity, Verification, and Trust

Transcript

Chief in Tech S1E16 - Daniela Barbosa : Authenticity in the Age of AI: A Guide to Decentralized Identity, Verification, and Trust · Chief in Tech Deep Dive Shortcast

• 0:00
  All right. Let's unpack this today. You know
• 0:03
  that that weird feeling when you get a text from
• 0:05
  a number you don't recognize? Oh, yeah, the generic
• 0:07
  ones. Right. Exactly. It says something super
• 0:09
  generic like, hey, it's me. I lost my phone.
• 0:13
  And for a split second, your brain does this
• 0:16
  this little calculation like, is this my mom?
• 0:19
  Is this, you know, a buddy from college or is
• 0:22
  this a scammer sitting in a basement halfway
• 0:24
  across the world? It's that is this real calculation.
• 0:27
  And I mean, we are doing it dozens, maybe hundreds
• 0:32
  of times a day now. Exactly. And that split second
• 0:34
  of hesitation, that is the sound of trust breaking.
• 0:37
  Because we are living in a moment where the line
• 0:40
  between human and machine isn't just blurring,
• 0:42
  it's actively dissolving. We're seeing CEOs in
• 0:46
  videos saying things they never said. We're hearing
• 0:49
  voices of family members that are actually AI
• 0:51
  clones. It really just feels overwhelming. It
• 0:53
  feels overwhelming because the tools we use to
• 0:56
  navigate the world are eyes, our ears, our gut
• 0:59
  instincts. they are suddenly totally unreliable.
• 1:02
  We are operating in this digital environment
• 1:04
  that has zero native way to verify who or what
• 1:08
  is on the other end of the connection. We're
• 1:09
  just flying blind. So today's deep dive is really
• 1:12
  about fixing that. We're unpacking a presentation
• 1:15
  by Daniela Barbosa from the Linux Foundation.
• 1:18
  And usually, when we talk about AI and security,
• 1:20
  it's all doom and gloom. Yeah, the robots are
• 1:22
  coming. Run for the hills. Run for the hills,
• 1:24
  exactly. But this source is different. This is
• 1:27
  about the architectures of trust, like the actual
• 1:30
  blueprints being drawn up right now to rebuild
• 1:32
  the internet's foundation. And it's a massive
• 1:35
  topic. But I love how Barbosa grounds it immediately
• 1:38
  with an analogy that I think we need to start
• 1:41
  with this because it reframes the whole problem.
• 1:43
  She talks about driving on a highway. Oh, the
• 1:45
  car analogy. Let's visualize this for everyone
• 1:47
  listening. Say I'm doing 70 miles an hour on
• 1:50
  the freeway. There is a two -ton metal box three
• 1:54
  feet to my left. I don't know the driver at all.
• 1:57
  No idea who they are. I don't know if they're
• 1:59
  angry or tired, drunk, texting. But I don't swerve
• 2:02
  off the road in a panic every time a car passes
• 2:04
  me. Why is that? because of infrastructure. You
• 2:07
  aren't trusting the person inside that car. You're
• 2:10
  trusting the system that put them there. You
• 2:13
  trust that they had to pass a test to get a license.
• 2:15
  You trust the car had to be registered, had to
• 2:17
  pass safety inspections. You trust the painted
• 2:20
  lines on the road and the threat of traffic enforcement
• 2:24
  if they cross those lines. There is this layered
• 2:27
  architecture of rules and credentials that makes
• 2:31
  the chaos manageable. And her entire point is
• 2:34
  that the internet just doesn't have that. Not
• 2:36
  at all. I mean, the internet was built on the
• 2:39
  principle of move fast and break things. It was
• 2:42
  designed for connectivity, not security. It's
• 2:44
  like a highway with no lanes, no licenses, no
• 2:47
  speed limits, and everyone was wearing a mask
• 2:49
  so you can't even see who they are. And for a
• 2:52
  long time, we just patched over that with passwords
• 2:54
  and databases. Which brings us to the first really
• 2:57
  major insight from the source material. We got
• 2:59
  to talk about why the old way, the password way,
• 3:01
  the database way, why it's failing so spectacularly.
• 3:06
  Barbosa dropped a stat that honestly made my
• 3:08
  jaw hit the floor. Oh, you're talking about the
• 3:10
  2024 health data breach statistic. Yeah. 275
• 3:14
  million people. It is staggering. In just one
• 3:17
  single year approximately 275 million individuals
• 3:22
  had their protected health information compromised.
• 3:24
  Wow. That averages out to about 750 ,000 records
• 3:28
  stolen every single day. That's effectively the
• 3:31
  entire adult population of the United States,
• 3:33
  give or take. But let's go a bit deeper than
• 3:35
  the number. Why is this happening now? Is it
• 3:38
  just that hackers are getting smarter? Hackers
• 3:40
  are getting smarter, yes. But the architecture
• 3:42
  itself is practically hitting them. We currently
• 3:45
  store data in centralized databases, these massive
• 3:48
  servers held by hospitals, banks, social media
• 3:52
  companies. In security terms, we call these honeypots.
• 3:55
  Honeypots. I actually really like that image.
• 3:57
  It's perfect because it explains the whole If
• 4:00
  you're a hacker, why would you try to rob 1 ,000
• 4:03
  individual houses to get a little bit of cash
• 4:05
  when you can just rob one bank vault and get
• 4:07
  everyone's gold at once? Right. It's a single
• 4:09
  point of failure. Exactly. Centralized databases
• 4:12
  are sitting ducks. And we, meaning the users,
• 4:15
  we aren't exactly helping, are we? I felt personally
• 4:17
  attacked by that password stat she cited. The
• 4:20
  Forbes report. Yes. She said 78 % of people use
• 4:23
  the same password across multiple accounts. And
• 4:26
  look, I am completely guilty. I have like 100
• 4:29
  accounts. I can't remember 100 complex codes.
• 4:32
  So I use variations of the same thing. And that
• 4:34
  is the human error fiction point. But here is
• 4:37
  the new variable that changes everything, AI.
• 4:40
  Barbosa describes AI as a threat multiplier.
• 4:43
  In the past, hacking a password took time. You
• 4:46
  had to type it in, guess it, brute force it.
• 4:48
  Now, AI can guess passwords at lightning speed.
• 4:52
  But even worse, it can bypass the human verification
• 4:55
  layer entirely. This is the deep fake issue.
• 4:57
  Correct. If I call your bank and I sound exactly
• 5:00
  like you because I've cloned your voice from
• 5:01
  the three -second video you posted online, the
• 5:04
  bank teller's trust radar is totally useless.
• 5:06
  Because they think it's me. Right. The old system
• 5:08
  relied on something you know, like a password,
• 5:11
  or something you are, like your voice. AI can
• 5:15
  now fake both of those things effortlessly. If
• 5:18
  we can't trust our eyes and we can't trust the
• 5:21
  centralized databases, we clearly need a completely
• 5:23
  new model. And this is where the deep dive gets
• 5:26
  really sci -fi, but in a highly practical way.
• 5:29
  Barbosa uses the example of autonomous cars,
• 5:32
  specifically those Waymo taxis in San Francisco,
• 5:35
  to explain how trust actually should work. This
• 5:37
  is such a brilliant visualization of machine
• 5:40
  to machine trust. Picture an intersection in
• 5:42
  San Francisco. It's foggy. There are no stop
• 5:45
  signs, no traffic lights. Just two robot cars
• 5:48
  approaching each other at speed. It's a digital
• 5:50
  game of chicken. In a way, yeah. But think about
• 5:53
  the constraints here. They can't hesitate. They
• 5:56
  can't roll down the window and wave the other
• 5:58
  guy through. They can't make eye contact. Right.
• 6:00
  No human cues at all. None. In milliseconds,
• 6:03
  car A has to tell car B, I am a certified Wemo
• 6:06
  vehicle. My software is up to date. I am traveling
• 6:09
  at 30 miles per hour. I intend to turn left.
• 6:11
  And car B has to believe it instantly. Instantly.
• 6:14
  Because if car A is lying, If it's actually a
• 6:17
  hacked car trying to cause a crash or it's a
• 6:20
  malfunctioned peaceful, people die. So they don't
• 6:23
  rely on looking at each other. They exchange
• 6:25
  cryptographic proofs. They verify credentials
• 6:28
  mathematically. And this all happens faster than
• 6:31
  a human could ever blink. Much faster. And this
• 6:34
  right here is the North Star for the new internet.
• 6:36
  We need to treat every interaction, logging into
• 6:39
  email, buying shoes, reading the news, with that
• 6:41
  exact same level of cryptographic certainty.
• 6:43
  So moving away from just taking someone's word
• 6:45
  for it. Exactly. We need to move from, trust
• 6:48
  me, I'm a doctor, to... Here is a mathematically
• 6:51
  unforgeable token proving I am a doctor. OK,
• 6:54
  so let's get into the toolkit here. How do we
• 6:56
  actually build this? The presentation threw around
• 6:58
  some acronyms that we really need to decode for
• 7:00
  everyone listening. DDs and VCs. Let's start
• 7:03
  with a DDI. OK, so DD stands for decentralized
• 7:06
  identifier. Think of it as a permanent digital
• 7:09
  name tag that you completely control. How is
• 7:11
  that different from my email address, like hostname
• 7:13
  at gmail .com? Your email address is rented.
• 7:17
  It belongs to Google. If Google decides they
• 7:19
  don't like you tomorrow, or if they shut down
• 7:21
  the service, your identity just vanishes. You
• 7:25
  are essentially a digital serf living on their
• 7:27
  land. The DID is entirely different. It's anchored
• 7:32
  on a decentralized network, like a blockchain
• 7:34
  or a distributed ledger. No company controls
• 7:36
  it. It's yours. So it's like owning your own
• 7:39
  house instead of renting an apartment. The landlord
• 7:41
  can't just kick you out. Yeah, exactly. You hold
• 7:43
  the keys. Okay, so I have the name tag, I have
• 7:46
  the house, but a name tag doesn't really prove
• 7:49
  anything. I can write astronaut on a nametag,
• 7:52
  but that doesn't let me fly the space shuttle.
• 7:54
  I need actual proof. Right. And that is where
• 7:56
  the VCs come in. Verifiable Prudentials. These
• 7:59
  are the digital equivalents of the plastic cards
• 8:02
  in your physical wallet. Your driver's license,
• 8:04
  your university diploma, your Costco membership,
• 8:07
  your gym pass. But digitized. Digitized. But
• 8:10
  strictly better than digitized. See, right now,
• 8:13
  if you take a photo of your driver's license
• 8:14
  and email it to someone to prove who you are,
• 8:16
  that's just a dumb image. Right. It can be faked.
• 8:19
  It can be photoshopped. It can be intercepted.
• 8:22
  A verifiable credential is a piece of code signed
• 8:25
  by the issuer. So the DMV cryptographically signs
• 8:30
  your digital license and you store that in a
• 8:33
  digital wallet on your phone. And this leads
• 8:35
  to my absolute favorite part of the presentation
• 8:37
  because it solves a problem I think we've all
• 8:39
  had, the student discount problem. This really
• 8:42
  clicked for me when she explained it. It's the
• 8:44
  perfect example of what the industry calls selective
• 8:46
  disclosure. Walk us through it. Okay, so In the
• 8:49
  old world, the way we do it right now, say you
• 8:51
  want to discount on some software. The site asks
• 8:54
  for proof you're a student. You upload a picture
• 8:56
  of your student ID card. What did you just give
• 8:58
  them? Well, I gave them everything on the card,
• 9:00
  my full name, my student ID number, my campus
• 9:03
  address, the expiration date, and probably my
• 9:06
  date of birth. Now ask yourself, does the software
• 9:08
  company actually need to know your birthday to
• 9:11
  give you 10 % off? Do they need your campus address?
• 9:14
  No. They just need to know, is this person a
• 9:16
  student? Precisely. In the new world, using verifiable
• 9:20
  credentials, your digital wallet talks to their
• 9:23
  site. The site asks, are you a student? Your
• 9:26
  wallet checks your credential, sees the university's
• 9:29
  signature, and sends back what we call a zero
• 9:31
  -knowledge proof. It essentially says yes. Just
• 9:34
  yes, nothing else. Just yes. It proves the statement
• 9:37
  is true without revealing the underlying data.
• 9:40
  The vendor knows you're a student, but they don't
• 9:43
  know your name, your age, or your address. You
• 9:46
  get the discount. They get the assurance, and
• 9:48
  nobody gets to hoard your private data. That
• 9:50
  is a massive shift. We go from data hoarding,
• 9:53
  where companies collect everything just in case,
• 9:55
  to actual data minimization. And think about
• 9:58
  the security implication here. If that software
• 10:00
  company gets hacked next week, what do the hackers
• 10:03
  steal about you? Nothing. Just a log saying this
• 10:06
  user was verified. Exactly. The honeypot is empty.
• 10:09
  There was nothing to steal. Now, I can hear the
• 10:11
  skeptics listening right now thinking, you know,
• 10:13
  this sounds great in theory. But is anyone actually
• 10:16
  doing this? It sounds like some futuristic concept.
• 10:19
  And that is where the Bhutan case study comes
• 10:21
  in. This isn't just a pilot program in a lab
• 10:23
  somewhere, is it? No, this is actual national
• 10:26
  policy. In October 2023, the Kingdom of Bhutan
• 10:30
  launched the National Digital Identity Wallet,
• 10:33
  or NDI. Why Bhutan? Smaller nations can often
• 10:36
  pivot a lot faster on infrastructure. They looked
• 10:38
  at the landscape and decided to basically skip
• 10:41
  the centralized database phase entirely and go
• 10:44
  straight to self -sovereign identity. So how
• 10:46
  does it actually work for the average citizen
• 10:48
  in Bhutan right now? They download the wallet
• 10:50
  app, they secure it with biometric, so a face
• 10:52
  scan or a fingerprint, and once they're in, the
• 10:56
  government issues their credentials directly
• 10:57
  to that phone, their tax ID, their national status,
• 11:02
  banking details. So the data physically lives
• 11:04
  on the phone? That's the key. It's at the edge
• 11:06
  of the network, not in the center. If a citizen
• 11:09
  needs to prove to a bank that they are tax compliant,
• 11:12
  they beam that specific credential from their
• 11:14
  phone to the bank. The bank verifies the government's
• 11:17
  digital signature and the transaction is done.
• 11:21
  No central database was pinged and no physical
• 11:23
  papers were handed over. It's flipping the whole
• 11:26
  model upside down. Instead of the government
• 11:28
  holding my data and me asking permission to see
• 11:31
  it, I hold my data and I grant permission for
• 11:34
  them to see it. It is a fundamental shift in
• 11:36
  power dynamics. And frankly, it's the only way
• 11:39
  to secure data in an AI world. You can't breach.
• 11:43
  275 million records if they are stored on 275
• 11:47
  million separate devices protected by biometrics
• 11:50
  because you'd have to you'd have to hack 275
• 11:52
  million individual phones right economically
• 11:54
  impossible for hackers okay so we've solved or
• 11:56
  at least we have a really good blueprint for
• 11:59
  identity proving who I am but we still have the
• 12:02
  other side of the AI coin deepfakes Content.
• 12:05
  Because if I see a photo of a war zone or a video
• 12:08
  of a politician accepting a bribe, how do I know
• 12:10
  that is real? A digital wallet doesn't really
• 12:12
  help me there. Right. And this is the second
• 12:14
  pillar of the architecture, content provenance.
• 12:16
  We are moving from identifying people to identifying
• 12:19
  the history of media itself. Barbosa highlighted
• 12:22
  the C2PA standard. The Coalition for Content
• 12:25
  Provenance and Authenticity. Which is a bit of
• 12:29
  a mouthful. It is, but the tech is completely
• 12:31
  revolutionary. The best example she gave was
• 12:34
  the Leica camera, specifically the Leica SL3S.
• 12:37
  She said it has authenticity built in. What does
• 12:40
  that actually mean? Is it just like a digital
• 12:42
  watermark? No, watermarks are easily cropped
• 12:44
  out or faked. This happens at the hardware level.
• 12:47
  Inside the camera, there's a secure cryptographic
• 12:50
  chip. As soon as light hits the sensor and the
• 12:53
  image is captured, the camera cryptographically
• 12:55
  signs the file. Almost like a digital wax seal.
• 12:59
  Exactly like that. It stamps the file with metadata.
• 13:01
  Who took it, when, where, using GPS, and with
• 13:04
  what specific camera model. And if you take that
• 13:07
  file and open it in Photoshop to, say, add a
• 13:09
  tank or remove a person, the seal breaks. Well,
• 13:11
  it doesn't just break though, right? It updates.
• 13:13
  Right. That's a really important distinction.
• 13:15
  It creates a manifest. If you edit the photo,
• 13:17
  the file history will say, original taken by
• 13:19
  Leica at 2 p .m., edited in Adobe Photoshop at
• 13:23
  4 p .m. So as a viewer, I could theoretically
• 13:25
  click a little info icon on the image and see
• 13:28
  that entire journey. That's the vision. Imagine
• 13:31
  scrolling through your news feed. You see a shocking
• 13:34
  photo from a conflict zone. You hover over it.
• 13:37
  If it has the C2PA credentials, you can see the
• 13:40
  chain of custody from the photojournalist's camera
• 13:42
  to your screen. And if it doesn't have it, you
• 13:45
  should be extremely skeptical. It's like an ingredients
• 13:47
  label for digital content. We know what's in
• 13:50
  our food. We absolutely should know what's in
• 13:52
  our news. And it extends to AI too. She briefly
• 13:55
  mentioned the Hedera AI Studio. If an AI generates
• 13:58
  an image or even makes a decision on a loan application,
• 14:02
  we need an audit trail. We need a public ledger
• 14:05
  that permanently timestamps it so we know, this
• 14:07
  was made by a machine. Transparency really seems
• 14:10
  to be the recurring theme here, but there is
• 14:12
  a snag. And Barbosa was very honest about this
• 14:14
  in her presentation. The world is a massive place.
• 14:17
  There are hundreds of countries, thousands of
• 14:19
  companies. If Bhutan has one wallet system and
• 14:22
  the EU has another, and Google builds a third,
• 14:24
  do they actually talk to each other? The interoperability
• 14:27
  challenge, yes. This is by far the biggest hurdle.
• 14:30
  If my digital driver's license only works in
• 14:33
  my home state, it's useless for travel. We cannot
• 14:36
  afford to have these walled gardens of trust
• 14:39
  where your identity gets stuck inside one corporate
• 14:42
  or government ecosystem. So what's the fix? She
• 14:44
  used an analogy I use every day but never really
• 14:46
  think about? Email? The SMTP analogy. Think about
• 14:49
  it. You can be on Gmail. I can be on Outlook.
• 14:51
  Someone else can be on a corporate Yahoo server.
• 14:53
  We can all email each other perfectly seamlessly.
• 14:56
  We don't need to be on the same platform to communicate.
• 14:58
  Because there is a shared underlying language.
• 15:00
  A protocol. SMTP. Or Simple Mail Transfer Protocol.
• 15:04
  Barbosa argues we desperately need an SMTP for
• 15:07
  trust. She calls it the Trust Spanning Protocol,
• 15:10
  or TSP. She gave a really heavy emotional example
• 15:13
  of why this matters, the refugee scenario. I
• 15:15
  think this is so important because it takes this
• 15:17
  out of the realm of just cool tech and puts it
• 15:20
  firmly into human rights. This really brings
• 15:22
  it home. Imagine a refugee fleeing a conflict
• 15:26
  zone. They have to leave their home quickly,
• 15:28
  their physical papers are burned or lost. But
• 15:31
  they have a digital wallet issued by a UN agency
• 15:34
  right on their phone. So they have a VC, a verifiable
• 15:38
  credential that says, I'm a refugee. I have these
• 15:40
  skills. This is my medical history. Right. Now
• 15:43
  they cross a border into a European country.
• 15:46
  The border control agent is using a completely
• 15:48
  different software system. Maybe it's a proprietary
• 15:50
  government system. If those two systems can't
• 15:53
  talk to each other, that refugee is basically
• 15:55
  back to square one. They are undocumented. They
• 15:58
  can't prove who they are. But with a trust spanning
• 16:00
  protocol? With TSP, the EU system can ping the
• 16:04
  UN credential, verify the cryptographic signature
• 16:07
  across different platforms, and instantly accept
• 16:09
  it. It acts as a universal translator for trust.
• 16:12
  That's not just convenience. That's dignity.
• 16:15
  That is the literal difference between being
• 16:17
  a ghost in the system and being a recognized
• 16:19
  human being who can access services. It allows
• 16:22
  trust to travel. And in a global economy, trust
• 16:25
  has to travel. If your credentials die at the
• 16:27
  border, you aren't truly free. So let's recap
• 16:30
  the stack here for everyone listening. We have
• 16:32
  the wallets, the DIDs, to hold our data. We have
• 16:34
  the credentials, the VCs, to actually prove who
• 16:36
  we are. We have the provenance, the C2PA standards
• 16:40
  to prove what content is real. And we have the
• 16:43
  protocol, the TSP, to make sure everyone speaks
• 16:45
  the same language. The blueprints are absolutely
• 16:47
  there. The foundation is being poured right now.
• 16:50
  But Barbosa's final point is that this isn't
• 16:53
  a spectator sport. Yeah, she listed a lot of
• 16:56
  organizations. Hyperledger, the Open Wallet Foundation.
• 16:59
  She's basically looking at all the developers
• 17:01
  and business leaders in the room and telling
• 17:02
  them, you have to build this. Because if we don't
• 17:05
  build it, or if we let one single big corporation
• 17:08
  build it just for themselves, we end up right
• 17:11
  back where we started. Centralized, vulnerable,
• 17:15
  and controlled by a very few people. This infrastructure
• 17:18
  needs to be open source. It needs to be a public
• 17:21
  utility. It needs to be exactly like the physical
• 17:24
  roads we drive on, available to everyone, owned
• 17:26
  by no one. There was one last concept she left
• 17:29
  us with, and I think it's the perfect place to
• 17:31
  land this deep dive. It's an idea that sounds
• 17:33
  so simple, but is actually totally mind -bending.
• 17:36
  First -person credentials. Oh, this is the ultimate
• 17:39
  paradox of the AI age. Explain what it is for
• 17:42
  the listener. It is simply the ability to cryptographically
• 17:45
  prove that you are a human being. Not who you
• 17:47
  are. Not, I am John Smith. Just what you are.
• 17:51
  Exactly. Just, I am not a bot. Think about it.
• 17:54
  In a future where 90, maybe 95 percent of all
• 17:58
  internet traffic is AI agents talking to other
• 18:01
  AI agents, spamming each other, generating endless
• 18:04
  content. human attention is going to be the absolute
• 18:08
  scarcest resource on the planet. So being able
• 18:10
  to raise your digital hand and say, hey, I actually
• 18:12
  have a pulse is going to be incredibly valuable.
• 18:15
  It's going to be the ultimate premium status.
• 18:17
  Being able to verify I am a biological human
• 18:19
  allows you to access spaces, communities or services
• 18:22
  that are designated humans only. It prevents
• 18:24
  swarms of bots from influencing elections or
• 18:27
  crashing stock markets or just overwhelming social
• 18:29
  networks with noise. It's wild to think that
• 18:32
  proving your humanity is going to be a technical
• 18:34
  feature. We used to worry so much about the Turing
• 18:36
  test. You know, can a machine pass as a human?
• 18:39
  Now we have the reverse Turing test. Can a human
• 18:41
  prove they aren't a machine? We need more technology
• 18:44
  to preserve our humanity. That is the great irony
• 18:47
  of all this. But it's also the hope. These tools,
• 18:51
  these architectures of trust, they aren't about
• 18:53
  controlling people. They are about empowering
• 18:56
  people to own their digital lives in a world
• 18:59
  that is becoming increasingly artificial. So
• 19:01
  for you listening right now, next time you unlock
• 19:04
  your phone with your face or you use a digital
• 19:06
  boarding pass at the airport or even just drive
• 19:09
  down the highway, I want you to think about the
• 19:11
  infrastructure behind it. Think about the invisible
• 19:13
  lines of trust that keep everything moving safely.
• 19:16
  and realize that the Wild West era of the internet
• 19:19
  is finally ending. The roads are being paved.
• 19:22
  The real question now is, who is deciding where
• 19:25
  those roads go? And if you are in a position
• 19:28
  to influence that, if you write code, if you
• 19:31
  lead a team, if you make policy, you should really
• 19:34
  look into the Open Wallet Foundation and Hyperledger.
• 19:37
  Look into C2PA. Because as Daniela Barbosa made
• 19:40
  so clear, trust is the only currency that actually
• 19:43
  matters in the age of AI. It's the one thing
• 19:45
  we just can't automate. It really is. That's
• 19:48
  a great point to end on. That's it for this deep
• 19:50
  dive. Stay curious, stay skeptical, and most
• 19:53
  importantly, stay human. We'll see you in the
• 19:55
  next one. Thanks for listening.