How is artificial intelligence redefining our perception of reality and truth? Can AI be creative? And how is it changing art and innovation? Does AI-generated perfection detach us from reality and genuine human connection?
Henry Ajder is an advisor, speaker, and broadcaster working at the frontier of the generative AI and the synthetic media revolution. He advises organizations on the opportunities and challenges these technologies present, including Adobe, Meta, The European Commission, BBC, The Partnership on AI, and The House of Lords. Previously, Henry led Synthetic Futures, the first initiative dedicated to ethical generative AI and metaverse technologies, bringing together over 50 industry-leading organizations. Henry presented the BBC documentary series, The Future Will be Synthesised.
THE CREATIVE PROCESS
Tell us about how you came to work as an advisor working at the frontier of the generative AI and synthetic media revolution? As we reflect on the resurgence of fascist ideologies, how do you see the role of deep fakes in accelerating these dangerous movements? And what responsibilities should creators and regulators of synthetic media hold in this context?
HENRY AJDER
Having worked in this space for seven years, really since the inception of DeepFakes in late 2017, for some time, it was possible with just a few hours a day to really be on top of the key kind of technical developments. It's now truly global. AI-generated media have really exploded, particularly the last 18 months, but they've been bubbling under the surface for some time in various different use cases. The disinformation and deepfakes in the political sphere really matches some of the fears held five, six years ago, but at the time were more speculative. The fears around how deepfakes could be used in propaganda efforts, in attempts to destabilize democratic processes, to try and influence elections have really kind of reached a fever pitch Up until this year, I've always really said, “Well, look, we've got some fairly narrow examples of deepfakes and AI-generated content being deployed, but it's nowhere near on the scale or the effectiveness required to actually have that kind of massive impact.” This year, it's no longer a question of are deepfakes going to be used, it's now how effective are they actually going to be? I'm worried. I think a lot of the discourse around gen AI and so on is very much you're either an AI zoomer or an AI doomer, right? But for me, I don't think we need to have this kind of mutually exclusive attitude. I think we can kind of look at different use cases. There are really powerful and quite amazing use cases, but those very same baseline technologies can be weaponized if they're not developed responsibly with the appropriate safety measures, guardrails, and understanding from people using and developing them. So it is really about that balancing act for me. And a lot of my research over the years has been focused on mapping the evolution of AI generated content as a malicious tool.
THE CREATIVE PROCESS
I know you are a lead advisor on the partnership on AI responsible practices for synthetic media. And I just see that now OpenAI has joined the Coalition for Content Provenance and Authenticity (C2PA) to address the prevalence of misleading information online through a kind of watermarking on deepfake videos, etc. Do you think that that's enough or there are still ways for bad actors to bypass it?
AJDER
I advise the content authenticity initiative, an organization advocating from the adoption of content provenance technologies like the C2PA. I think it's worth breaking down some of the solution approaches that are being put forward technologically because understandably, it's quite complicated. The first of these is detection, which looks to try and give a score or a kind of an evaluation using AI to spot, let's say, the kind of digital fingerprints left behind by AI-generated content, which the human eye or ear may not be able to detect. But deepfake detection is always going to be, to some extent, unreliable. There are tens of different detection tools out there, and they sometimes disagree with each other. It will be able to give you a percentage confidence score based on probabilistic reasoning. But we've actually seen in the hands of everyday people, this doing more harm than good because they've sort of trusted blindly a detection system which is not robust. The second approach is watermarking within the media, on the pixel level or on the kind of wave level in audio. This is the intentional injection of artifacts into a piece of media, which really stands out to a detection tool. The challenge is that signals similarly have a robustness challenge. They can be manipulated. Someone can compress a piece of audio or compress an image, which actually removes the fidelity around those watermarks. The last is C2PA, which attaches cryptographically secured metadata to a media, providing something like a nutrition label. The moment an authentic photo is captured, let’s say, data is cryptographically secured to that media file and then travels with it throughout its lifetime. And as people continue to manipulate it within that standard, it then continues that log, that ledger of how manipulation has taken place. In my mind, it's the most promising approach because it's the most secure.
THE CREATIVE PROCESS
I'm glad we're going in that direction. It helps the average viewer understand when you say it's like nutritional labels. A lot of us pass over that because it's just like a long list of nutrients and names of chemicals I don't even understand, and I just have to go to the checkout line. I think in some ways the existence of deepfakes reinforces the importance of legacy media, but I'm just wondering what this is all doing to us. You produced the BBC series The Future will be Synthesized, but now we’re entering a synthetic reality. AI is rapidly altering our collective reality. It's kind of a Nietzschean all gods are dead, be your own god, and some tech companies are offering opportunities to inhabit dreamlike worlds where we can be more than our real selves. For those who haven't had a chance to experience enough of the real world before they're immersed into a synthetic one, how do we protect those most vulnerable in society from these synthetic realities?
AJDER
The series that I did came out in May of 2022. When I was producing that documentary with my producer at the BBC, I don't think I was expecting us to have accelerated so quickly into that future. The world has been quite synthetic for some time. I just don't think we've necessarily been aware of it, but whether it's computational photography in every single smartphone, even when you think you're taking a photo, no filter, there's still quite a lot of shaping of the reality that you're presented with at the end by these phones. Movie magic and entertainment, recommendation algorithms for your Facebook feed or Grammarly are all AI-powered. But we're now interfacing with AI in a much more intimate and personal way. When you were taking a photo or scrolling through your Twitter feed, AI was working behind the scenes, but you weren't necessarily aware of that. Whereas in the age of generative AI, we are becoming almost kind of synthetic conductors of the reality that we're creating. That could be via a large language model or through one of these companion based apps. We're starting to see some really powerful tools like Suno coming out which can generate AI-generated music. We are starting to see AI-generated content seep into areas of our lives, particularly around what we traditionally see as human-to-human communication or human creativity, that I think does fundamentally change the way that we kind of almost evaluate the value of human creative endeavors, but also the way that we think about interacting with each other in the digital world. How is that going to shape expectations of what normal messy human relationships look like? There's a huge swathe of gray. There's a big question of are we feeling uncomfortable about these use cases because they're new, they're unfamiliar? Or are we feeling uncomfortable because a deeper ethical intuition is being disturbed?
THE CREATIVE PROCESS
Last year, the OECD reported a significant ten-year decline in reading, math, and science performance among 15-year-olds globally. One-third of the students cited digital distraction as an issue, and there was an overall tripling of ADHD diagnoses between 2010 and 2022. I'm curious to hear your thoughts on the future of education and the insights from the paper you wrote for the All Party Parliamentary Group on Artificial Intelligence.
AJDER
From an educational perspective, I believe AI has the potential to level the playing field and provide broader access to digital resources akin to having a digital Aristotle – offering dialogue and bouncing of ideas. I've observed some exciting use cases and tools in universities that demonstrate the cost-effective power of these technologies. In addition to text, AI-generated content has the potential to bring history to life and create innovative forms of assessment and engagement for teachers and students. Despite its reliability issues, AI can provide valuable data points in subjects like history and science and also revolutionize content presentation. However, there is an ongoing crisis in higher and secondary education related to cheating, as educators struggle to detect AI-generated content in student submissions. While many in the industry are enthusiastic about the opportunities AI presents, detection tools are currently insufficient to confidently identify AI-generated work as cheating. To address this issue, alternate assessment methods such as Viva style examinations have been suggested. However, implementing such methods would require a significant increase in resources. There are also proposals for randomized defenses of submitted work, similar to a "drug test," to deter students from using AI to complete assignments. The prevalence of AI-generated assignments poses a significant challenge, particularly as educators and students are reluctant to revert to traditional examination methods. While there are creative solutions we can explore, such as engaging with chatbots for critique and limitations, it's evident that the issue requires attention, especially in state education where resources and AI expertise may be lacking.
THE CREATIVE PROCESS
You speak about the importance of the arts and the humanities, and I'm just wondering how you reconnect to who you were before the overlay of synthetic realities?
HENRY AJDER
Getting out in the countryside and getting out in nature is really important to me. I do have ways to get away from screens, which I think are really important. Having seen some of your previous guests' music, it's something that's really important to me and I have really enjoyed some of the artists that you've had on. I think that those are kind of key parts of being human. I think it's something which is closer to almost a religious assertion than it is grounded in sort of a scientific version of reality. For me, if I didn't know a piece of music was AI-generated, I could think that was absolutely beautiful. The moment I know that actually it's AI-generated, part of the magic goes for me. And it's the same with writing. It's the same with art in general. I think that's because there is an intangible value, and it's a real value. It really does shape the phenomenology of how you experience that piece of art, which comes from knowing that it is contained essentially from one consciousness to another. AI obviously can't do that. Fundamentally we can't have that epistemic knowledge of other people having minds, having consciousness. It's kind of Cartesian, right? It's like, the only thing I know I can truly know is that I am, right? We can't fundamentally know that other people exist, but the way that I think we do come to have those transcendental experiences between people is through art. I've seen some beautiful AI-generated art, but in my mind, it is inherently less valuable to me because there isn't enough consciousness driving that creative process. Keeping that side of things is really important.
THE CREATIVE PROCESS
And finally, as you think about the future and education, the future of work and our AI co-creators, what would you like young people to know, preserve and remember?
AJDER
I feel potentially something I'm concerned about, and I guess by extension I would like to preserve, is a real sense of empathy and humility, which comes with understanding that the world is messy, that people are messy, that defects and imperfections exist, that things don't always necessarily kind of go the way you want, even as much as you wish they could. Imperfection is part of life and I guess my concern is that AI-generated content, which smooths and perfects a version of reality to precisely what you want and forces you or makes you feel pressured to represent yourself in this absolutely perfect way, fundamentally gives you no room for error and kind of detaches you from the reality of growth and life and and how people work. Empathize with other people. Everyone has their challenges. Things don't always have to be exactly perfect to how you want them to be or how other people want them to be. And that involves having some humility about yourself as a messy creature, as we all are. I hope that's retained, but I do see this kind of move towards this sort of smoothed and shaped reality that AI is enabling, potentially creating more of a disconnect between that imperfect, messy, but also quite beautiful world. This sort of polished but ultimately plastic version of reality increasingly is becoming the default for some people over the kind of fleshy, messy human side of things.
