Will AI Prompt a New Golden Era? [Podcast]

In this episode of Sanford’s Policy 360 podcast: we’ll explore AI – from deepfakes to the growing importance of social media verification. Our guest Robyn Caplan is an Assistant Professor at Duke’s Sanford School of Public Policy and is currently teaching a class on the transformation of media. Her latest research considers the blue-check verification process that is used on many social platforms. Duke Sanford faculty lead Anna Gassman-Pines hosts.

Conversation Highlights

Responses have been edited for clarity.

what excites and concerns you about AI?

What excites me about AI is that these tools can decrease loads for certain activities, and we can also use these systems to “level up.” In my class, I have now decided to assign harder theory because students have more tools to parse them. So, if we're using these tools in the correct way, we can use them to make our work more robust and more complete. 

But if we over hype our ability to replace expertise, we can risk eroding intellectual infrastructure. There was a study that was put out by Microsoft about all of the jobs that they thought were most susceptible to AI. Number two on that list was historians. Now that is just a complete misunderstanding of what historians do. If you view historians as just synthesizing information that is out there, then sure you may come to the conclusion that AI could replace historians, but historians do a lot more. They produce knowledge, they do interviews, they produce archives. There’s an erroneous and strange belief that all of the information is online already. And so my big concern is that if we over-hype these tools, then we're going to start undermining the infrastructure - journalists, historians, policy analysts, scholars - that not only future societies depend on, but AI depends on. AI cannot produce new knowledge. It requires on the knowledge that we produce. If we are not producing any knowledge, it will have nothing to synthesize in the future.

On ‘DEEP FAKES’

Deep fakes are synthetic media that are manipulated or created using AI to appear real, but they're fabricated and altered. They’re produced by training AI algorithms to generate realistic content. Then you use another AI algorithm to detect whether it is fake. Then they just play off of each other until you get something that is real enough. Now, the big issue with ‘deep fakes’ that has always really intrigued me is that we have been concerned about them for a long time - we have been talking about them as a looming crisis since about 2016. But they have never become the crisis we thought they would be.

REGARDING ONLINE verification and policy considerations

Image

The blue check verification process started as a back-end process. It was a way to say, "Yes, this is the real person or organization.” It changed over the years and we are starting to move towards a possibility that everybody on the internet will need to be verified.

We have had a huge amount of legislation passed at the state level within the U.S. They've been trying to implement age verification laws in Australia, across Europe, in the U.K. That is leading to the possibility that we will all need to verify ourselves when we go online because children do not necessarily have IDs.

Sam Altman is the founder and head of ChatGPT, OpenAI, and he has been putting an enormous amount of funding into this product called World. He's been going into areas, particularly Northern Africa and Argentina and scanning irises and in exchange for cryptocurrency to build this kind of database that he is hoping or at least to try a proof of concept around this world passport system that he thinks is going to be necessary in the future. Policymakers are going to have a really, really big role in how both age verification legislation takes shape and about how these new technologies take shape with regards to age verification. And as policymakers, what we need to be doing is balancing harms. So yes, regardless of what the research says, that is the future that we are going into.

ON THE LONG HISTORY OF FAKE IDENTITIES

What surprises my students is how long we've been having these same conversations about what is real and what is fake when it comes to identity in society. We start the semester with a piece that talks about the idea of the false messenger within the medieval era. Once kingdoms started growing and they wanted to talk to each other, one kingdom would send a messenger from one to the other, but suddenly you had people just claiming that they were the messenger. And so, you had this complicated set of insignia that would be used and then that would then be parodied. So, they had to continue to outsmart each other. 

on the latest in online verification systems

So this is my most favorite thing that I have done at Duke yet, I allowed students to study an existing verification system.  I had one student that did biological athlete passports. How do they keep track of testing for doping at major events? I had another student do Costco memberships. This has totally shifted how I think about Costco and how they're using my data!

I had one student do a project on how fake IDs work in the digital era. Now you have to put in a fake ID online. It's a much more complicated system - it's one that is not done by your older brother's sketchy friend anymore. But my favorite project had to do with Raya, which is an elite dating service. And what the student ended up doing is that she went down a path of investigating the alternative market for Raya referrals that exists on Reddit. All of the people who are trying to get referrals into this elite dating market by going this alternative route and working with other people and buying these referrals

Identity and online verification: What should policymakers consider?

We need to balance the harms, and we need to recognize when is the moment to be doing this and when is not. What is concerning me right now is that we are at a moment where it might be more important than ever to preserve online anonymity, to preserve freedom of expression. And that is coinciding with a moment where we are deciding to erode all of those protections. I don't disagree with verification because there's nothing to really disagree with. We are all verifiable. There is no real privacy online. Everything we do, our IP tracks back to us. Our activities can be tracked to us as well, just in the very basic patterns. But at a social level, I do think that there is something that is important that we need to preserve, that we are in danger of eroding very quickly.

We're at a scary moment where it's going to be very dangerous for young people with marginalized identities to search about their identity online. And that is a very scary thing to imagine. 

We are seeing age verification laws being passed across the US. There are a set of laws that are motivated by privacy in California and New Hampshire where they're trying to differentiate how data is taken and treated between young and old, and that seems very laudable.

There's a set of legislation including legislation in North Carolina that's trying to prevent access of minors to pornographic content. Those tend to be taken up by more conservative states. And then there's a set of laws that are being passed across the US that's trying to limit access to social media for young people. Those were being passed primarily by conservative states, but it is extending to areas like New York and New Jersey as well. And we just have to be really cognizant of the time and place in which these are being implemented that we don't want to limit access to information in an environment where we are seeing information access limited in lots of different areas like libraries.

Robyn Caplan, PhD is an Assistant Professor at the Sanford School of Public Policy. She was formerly a senior researcher at the Data and Society Research Institute. She is a founding member of the Platform Governance Research Network.