Who Gets to Interpret You? | The Machine-Readable Self

WEBVTT
00:00:19.920 --> 00:00:26.879
What happens when your identity stops being something that you declare and starts becoming something that's inferred?
00:00:26.879 --> 00:00:29.679
Not by other people, but by machines.
00:00:29.679 --> 00:00:41.039
That's what we're here to investigate in the machine readable self, a compilation episode of archive conversations from the last few years of The Future of You, the podcast.
00:00:41.039 --> 00:00:44.880
Here's Holly Herndon.
00:00:45.280 --> 00:00:59.679
As soon as something is captured in media, or as soon as you take a photograph or a video of anything, then it becomes machine legible, and then it can become part of a training canon, you know, to train a model that you may never even interact with.
00:01:00.079 --> 00:01:02.240
Holly's point is deceptively simple.
00:01:02.240 --> 00:01:05.359
Once something is captured, it can be read by machines.
00:01:05.359 --> 00:01:09.280
A photograph, a voice, a gesture, a fragment of media.
00:01:09.280 --> 00:01:12.000
All of it becomes potential training data.
00:01:12.000 --> 00:01:18.079
So identity is no longer just what we say about ourselves, it's what the system can extract from us.
00:01:18.079 --> 00:01:27.359
But all this kind of digital phenotyping, treating us almost like patients, capturing some of the atmospheric or environmental data around us, the way we move.
00:01:27.359 --> 00:01:39.120
I was talking to somebody about, you know, motion prints in VR, for example, the way in which our gesture with our head and our hands, well, you know, it's so individual that it's like a fingerprint and they call it motion print.
00:01:39.439 --> 00:01:45.439
I think it all depends on kind of where we end up in terms of who's using this data and for what purpose.
00:01:45.439 --> 00:01:55.280
I like to think of AI as us in aggregate, as a collective human accomplishment rather than a kind of like alien other that's here to just replace us.
00:01:55.760 --> 00:01:57.200
That is the first shift.
00:01:57.200 --> 00:01:59.120
The self becomes data.
00:01:59.120 --> 00:02:03.359
But once the self is readable, it can also become searchable.
00:02:03.359 --> 00:02:07.120
And that is where the consequences become much harder to contain.
00:02:07.120 --> 00:02:09.039
Here's Kashmir Hill.
00:02:09.360 --> 00:02:14.159
When I first discovered Clearview AI, what it was doing was a secret.
00:02:14.159 --> 00:02:16.080
It was not known by the general public.
00:02:16.080 --> 00:02:22.159
People found it really shocking that their photos had been collected by this random company.
00:02:22.159 --> 00:02:24.960
And honestly, people were upset.
00:02:24.960 --> 00:02:28.800
What Clearview is doing is, you know, they built this database.
00:02:28.800 --> 00:02:30.879
They wanted to make money off of it.
00:02:30.879 --> 00:02:38.240
They would end up selling it to police departments, law enforcement officials, thousands of agencies.
00:02:38.240 --> 00:02:43.919
And it was tried by police departments around the world, including in the UK.
00:02:43.919 --> 00:02:48.080
There is an events venue here called Madison Square Garden.
00:02:48.080 --> 00:02:51.039
It's a really iconic venue for concert shows.
00:02:51.039 --> 00:02:58.400
And they started using facial recognition technology, not Clearview AI, a different company a few years back to address security threats.
00:02:58.400 --> 00:03:10.000
But in the last year, the billionaire, James Dolan, who owns that venue, decided to ban lawyers that worked at firms that have sued his company.
00:03:10.000 --> 00:03:12.319
I went with a lawyer, I watched it happen.
00:03:12.319 --> 00:03:13.759
She got pulled aside.
00:03:13.759 --> 00:03:15.599
They said, You're not allowed in here.
00:03:15.599 --> 00:03:17.280
You work at a firm that's banned.
00:03:17.280 --> 00:03:18.879
And she said, I'm not on that case.
00:03:18.879 --> 00:03:20.800
You know, I'm not working on anything related to this.
00:03:20.800 --> 00:03:22.159
They said, it doesn't matter.
00:03:22.159 --> 00:03:26.080
You're banned until your firm drops its lawsuit against us.
00:03:26.080 --> 00:03:38.400
And so that is just, oh, you know, a very striking way that this technology can be used to discriminate against people in new ways, even based on the work that you do.
00:03:38.400 --> 00:03:46.240
The other risk is just that you get slippage, that it starts out with, we don't want you coming in here because you're a threat.
00:03:46.240 --> 00:03:49.840
But then you realize there's other ways you can wield the superpower.
00:03:49.840 --> 00:03:56.639
But once you get this infrastructure in place, it might be used in a way that wasn't originally envisioned.
00:03:57.120 --> 00:04:04.240
Kashmir's example is so powerful because it shows how quickly a security technology becomes an identity technology.
00:04:04.240 --> 00:04:18.800
The face is no longer simply something that people recognize, it becomes an access point, a filter, a permission system, a way of deciding who gets in and who's kept out, and on what grounds.
00:04:18.800 --> 00:04:20.560
Here's Andrew Bud.
00:04:21.680 --> 00:04:24.720
To me, a digital identity is a set of facts about me.
00:04:24.720 --> 00:04:28.879
But a person is not the same as a set of facts in a folder.
00:04:28.879 --> 00:04:35.839
So there comes a moment in the life of every digital identity when that folder has to be bound to a real life physical human being.
00:04:35.839 --> 00:04:37.040
And that's our job.
00:04:37.040 --> 00:04:45.839
So our job is to bind that buff folder full of facts about you to a genuinely present human being who is the owner of that folder.
00:04:45.839 --> 00:04:50.480
I'm sometimes asked, uh, you know, what's so special about a f about a face?
00:04:50.480 --> 00:04:51.759
You know, why not the iris?
00:04:51.759 --> 00:04:53.040
And the answer or the voice.
00:04:53.040 --> 00:04:59.600
And the answer is, well, frankly, when the government creates credentials of my identity, they don't put my iris into it.
00:04:59.600 --> 00:05:09.839
The only thing that binds this passport that I'm invisibly waving at you at this moment, Tracy, uh, that shows that it's mine is my face on page two.
00:05:09.839 --> 00:05:21.680
The face will continue to become a B, and it's the coarse root of trust in the digital economy, and assuring the genuine presence of that face will become fundamental to the disability of society.
00:05:22.720 --> 00:05:24.800
Andrew gives us the other side of the problem.
00:05:24.800 --> 00:05:31.680
In an age of deep fakes, robots, and synthetic media, we do need ways to know that a real human being is present.
00:05:31.680 --> 00:05:35.279
So, the face becomes the root of trust.
00:05:35.279 --> 00:05:37.600
But the paradox is obvious.
00:05:37.600 --> 00:05:43.199
To prove we are human, we increasingly have to submit ourselves to machine reading.
00:05:43.199 --> 00:05:45.199
Here's Byron Tao.
00:05:45.759 --> 00:05:54.000
The first era is the traditional data broker, and these are companies who are kind of sort of household names if you pay attention to tech and to advertising.
00:05:54.000 --> 00:06:01.600
They're companies like Thomson Reuters, Axiom, credit bureaus like TransUnion eventually got into this kind of market.
00:06:01.600 --> 00:06:24.240
And what they have historically done is gone out and taken either credit data from banks, commercial data from things like magazine subscriptions, public data from things like courthouses, marriage records, hunting licenses, driver's licenses, and created essentially lists of where people live in the world, what assets they own, and some basic demographic information, right?
00:06:24.240 --> 00:06:27.199
The second generation is social data brokers.
00:06:27.199 --> 00:06:39.839
So when social media became a big part of the way we communicate and the way we express ourselves in the early 2000s, there started to be these data brokers that began monitoring the social conversation.
00:06:39.839 --> 00:06:50.079
Governments also increasingly became interested in acquiring social media data and analyzing it and looking at people's connections with each other and who they talk to and what they say.
00:06:50.079 --> 00:06:52.000
Then there's location brokers.
00:06:52.000 --> 00:06:56.800
So these are brokers that started collecting information off of mobile phones.
00:06:56.800 --> 00:07:09.439
And then the weirdest kind of data I write about is what I call in the book gray data, you called it esoteric data, but you know, it is data that we do not normally associate with, we don't even really think about.
00:07:09.439 --> 00:07:11.920
It's this kind of weird digital exhaust.
00:07:11.920 --> 00:07:15.439
So I'm talking to you through a pair of Bluetooth headphones.
00:07:15.439 --> 00:07:25.600
Well, Bluetooth headphones are constantly pinging out a signal, and if you know how to collect that signal, uh you can say Byron's Bluetooth headphones were at this particular address.
00:07:25.600 --> 00:07:34.879
There's just all manner of data sets that you really need to be a technologist to understand, and governments are moving to collect that kind of data as well.
00:07:35.279 --> 00:07:40.240
The machine readable self is not a future scenario, it is already here.
00:07:40.240 --> 00:07:54.399
Our faces, our voices, our movements, transactions, preferences, and behaviours are becoming part of a larger identity layer, one interpreted less by people and more by systems.
00:07:54.399 --> 00:08:02.480
This is the shift I've been exploring through my work on everything in the future of you through to the database self and the database brand.
00:08:02.480 --> 00:08:11.040
How people and organizations are increasingly understood through data, signals, relationships, and inference.
00:08:11.040 --> 00:08:14.639
The question now is not simply who are you?
00:08:14.639 --> 00:08:18.720
It is who or what gets to interpret you.
00:08:18.720 --> 00:08:27.839
Thanks for listening, and please do join us next time for another archive essay taken from the conversations we've had on the Future Review.
00:08:27.839 --> 00:08:30.639
And please do subscribe and like and share.
00:08:30.639 --> 00:08:32.159
Thanks for listening.