Liberating the self from calculation: entropy in privacy enhancing technologies

Hello. Is the microphone working? Yes, it's working. Real pleasure to be here and such a pleasure to speak after my good friends Waseem and Paul and their fantastic presentations. And that there is a track at Protocol Berg for philosophy.

I love all of that. So very Nim. Nim is a decentralized noise generating network. And I have a little bit of an academic background also in the field of digital geography and political economy. So that's just a little bit about me.

I'm here with some of my fantastic colleagues. We did some kind of more technical sessions early this morning. But my talk now is going to discuss privacy from a little bit of an unusual angle. I'm going to start off from psychotherapy. A book that was written by Iraqi psychotherapist called Noir Al-Sagir.

And she went to clown school in order to kind of investigate the nature of spontaneous laughter and its role in accessing the true self. So this is going to be the kind of unusual starting point for privacy preserving technologies that I'm going to be discussing today. The other reason why I wanted to kickstart with psychotherapy was also partially because it's kind of fun that it triggers some of the kind of right-wing accelerationists. I don't know who of you kind of saw this retracted ad from Solana last year, but there seems to be a kind of a little bit of a strong resistance from right-wing accelerationists towards the idea that we should at any point in time, you know, look a little bit to our interiors and why we react the way that we do. So a bit of an overview first.

So yeah, as I said, I'll start off with Noir Al-Sagir's work. She is a psychotherapist. So I thought it would be nice to kind of then move on to the work of a contemporary philosopher, Bium Chul Han, who actually looks at also kind of the psychology of the world, but through kind of the lens of digital technologies and the impact of digital technologies on our perception of ourselves. From there, I'm going to move on to investigate the concept of entropy through another contemporary philosopher, Cecile Malaspina, who's done a kind of rereading of the Shannon Weaver mathematical theory of communication. And then to kind of bring it all together, discuss the concept of entropy, you know, the way that entropy actually plays out in privacy-enhancing technologies, in order to then kind of position a kind of bigger-picture framework to suggest that entropy might be a good lens to bring the kind of field of privacy-enhancing technologies into the conversation about artificial intelligence specifically.

So yeah, so starting off with, you know, the title, I'm going to kind of structure this talk very much step-by-step, moving through the actual title itself. So liberating the self. What do I actually mean by that? You know, there's a lot of different kind of ideas and notions around freedom out there, but, you know, here, building on the work of Al-Savir, there is this kind of idea that, like, the relationship that we have for ourselves, it starts from a very young age. So there's this kind of process of socialization that happens from almost a kind of infant age, where we start to kind of notice how we need to behave in order to access things that are important to us.

And those things are connection, love, food, security. And as we start to adjust our behaviors in these micro ways in relationship to our environment, that starts to kind of build up a set of subconscious understandings of the world and how we must be in the world as an identity in order to access the things that are important to us. So the process of psychotherapy is really like, you know, at a certain point in your life, those kind of belief systems or structures of behavior and subconscious ways of behaving no longer serve you. You know, they might kind of actually be hindrances towards the things that you really want to kind of achieve in your life, or they might, in fact, kind of fog your understanding or your relationship to yourself on a deeper level. So the process of psychotherapy is really to kind of understand these deeper patterns of behavior that you've kind of acquired throughout your life.

And in fact, liberate yourself from those patterns and achieve this connection to a much more spontaneous, true self is what is what Nour El-Sedir talks about. So what I found interesting when I was reading this book was that, you know, she's talking about the formation of identity, you know, as we get to know ourselves through the eyes of others that we surround ourselves with. So this kind of socialization that happens, she touches only very briefly on how digital technologies actually have started to affect that. But if we think about it a little bit over the past, you know, a the way that we understand ourselves is increasingly not just through the eyes of others, the immediate others that we relate to our family, our friends kind of on an immediate basis, but also through the eyes of millions of others as they are online or perceived others as they are online or perceived ways of viewing ourselves through the eyes of infrastructure. And, you know, that includes actual people, but it can also include various things like credit rating or whatever, you know, so we start to kind of modify our behavior and our sense of identity through not just other human beings, but also through large scale technical systems.

So someone who's actually worked on this quite a bit, thought quite deeply about this in terms of philosophy is Byung-Chul Han. And Byung-Chul has this concept of the psychogram. So this is an idea that Byung-Chul has developed, kind of taking, you know, going deep on the work of Michel Foucault. So those of you who, you know, are into kind of philosophy, you'll probably have read some of Foucault's work. And Foucault has this idea of biopolitics, right?

And this is very much around the disciplining of the body, and kind of power structures have emerged around the disciplining of the body, where Byung-Chul argues that, you know, biopolitics kind of like only took us so far in terms of history. And actually, we've shifted into a period that he calls psychopolitics. And the reason why he does that is because there's kind of almost like a, let's say, you know, an externalization of cognitive processes that have happened, you know, through the invention of the internet. And that externalization of cognitive processes, you know, as a type of cognitive prosthesis, doesn't just, it's not just a kind of explicitly conscious act, but actually, like incorporate certain of our sub, certain subconscious dynamics as well. And this idea of the psychogram is the, you know, a kind of shared subconscious that is written into digital infrastructure, data infrastructure, data economies, in a way that's almost, you know, it represents a kind of shared subconscious, that Byung-Chul then argues that because it's been externalized in these sorts of ways, becomes kind of ripe for exploitation.

And the type of exploitation that Byung-Chul talks about, is, you know, he talks very much about very much the kind of personal stuff that happens. So, you know, and again, to come back to the concept of freedom, Byung-Chul says, like, freedom itself has almost become a new type of subjugation, where instead of perceiving ourselves as subjugated subjects, as we might do under a refashioning and reinventing ourselves. And there is this sense of freedom that attends from passing from the state of subject to that of project. But at the same time, when we start to perceive our identities as project, this projection amounts to a form of compulsion and constraint, in fact, a more efficient kind of subjugation and subjectivation, because it is us constantly modifying ourselves in relation to the project that has become ourselves as a project. And, of course, you know, there's like very, you know, easy references that we can make here to kind of social media and the effect of social media on people's kind of sense of identity.

But I think there's a bigger kind of deeper piece here that I haven't like fully got the full got the entire sense of yet, but that has to do with, you know, why, you know, identity politics has become such a big thing. In fact, has kind of escalated into a full blown identity war in many ways, at the same time, as we've kind of had these, the social media pervasiveness, as we've had over the past 10 years. So from Jungtul, I want to kind of now get into a little bit like, you know, we have the kind of psychogram, we have the kind of internet, I want to move into this concept of entropy that, you know, somehow can maybe loosen up a little bit of the, you know, some of the kind of disciplinary mechanisms or the subjugation that Jungtul is mentioning. So Cecile Malaspina is another contemporary philosopher, and she's done a rereading of the Shannon Weaver mathematical theory of communication. So what's interesting about this piece of work is like, so if we take the kind of mathematical theory of communication, and the way that kind of more and more understood a little bit as the opposite to kind of signal, right, so you have an increase in entropy in, in the system, then the, you know, you lose kind of signal.

So it's almost like the set in opposition to information. So more entropy, you can't, you know, find the signal, signal versus noise type of understanding. But what Malaspina is actually talking about here is that when you look closer at the Shannon Weaver theory, what they're actually arguing is that entropy, in fact, requires, or is a designation of needing more information in order to fully express the level of unpredictability and uncertainty across the set of probabilities. And so, you know, understanding this a little bit more kind of fine grained, what's happening here is that a piece of information informs us only if it's not redundant, aka, you need a level of entropy in order for information to even qualify as information. So this is something that we're kind of, you know, those that have been around in privacy preserving technologies will be somewhat familiar with.

So saying from the NSA General Baker that you need the haystack in order to find the signal. So that's a little bit the same kind of logic that we're seeing here. But what's what is interesting about the kind of framework of entropy is that, you know, this, this aspect of, can I get some water, it's very hot up here, and I'm losing my trail of thought. Thank you so much. Okay, reset.

So, so we see kind of entropy. We see entropy in the kind of like beginnings of, you know, the NSA projects of being kind of like operationalized in the beginnings of the NSA project of mass surveillance, you need loads and loads of information, in order to set up a kind of have enough in order to kind of start predicting, in order to start making convincing predictions. And, you know, fast forward, sorry, this is I'm, the lights are just very hot. So it's getting a little bit distracting for me. Fast forward to the state of the condition of today, and we've arrived at, you know, I kind of, sorry, guys, I'm just going to take a minute because it's like a little bit too much.

Yeah. Thank you. Hello, I can see you now. Oh, my God. Okay, this is actually so much better.

It's, I felt like I was staring into a kind of blank void, and it was getting a bit much. I'm going to get my notes, and then I'm going to continue from the floor. Thank you, audience, for being so kind and patient, as I get my thoughts back together again. So, right. So you need the needle to, or the need of the haystack in order to find the needle.

And, you know, this, this is, you know, this idea of requiring a lot of information in order to surface kind of predictions that are, start to make plausible predictions is, of course, kind of the, some of the underlying operations also of artificial intelligence. And this created really this, like, hunger for data and information. And, you know, yeah, so we've ended up now with the artificial intelligence agents, or artificial intelligence industry, with the operations to, sorry, guys, this is really, no, it's, it's actually quite frustrating, because I got really quite blinded by the lights. And I'm going to try one more time. Right, so the use of entropy, as it's being used in the artificial intelligence industry, is a use of entropy that is entirely concerned with prediction.

It's trying to surface patterns. It's trying to acquire as much information as possible through ever-expanding, surveillance infrastructures. Thank you. And that hunger for data is insatiable exactly because of, you know, how we understand the kind of Shannon-Weaver mathematical theory of communication, which is that when you have uncertain and unpredictable environments, and you're trying to make predictions in those environments, you need more information. So the hunger is insatiable.

And so we've ended up in this, at this point, also, and this is not, like, of no consequence. I mean, there's, you know, potentially interesting applications of AI, sure, but we're actually in a moment right now where the artificial intelligence industry is directly aligning itself with warfare and genocide. And these things, you know, hugely matter. So the question is really what can be done about it. And this is also why I was so interested and keen in Malaspina's rereading of entropy, because it's an interesting framework that shows, you know, the kind of very famous use of entropy on one hand with the AI industry right now, but there's been actually a much longer term kind of operationalization of entropy in the field of cryptography and privacy-preserving technologies.

And so what I would like to propose is the idea of entropy as a kind of theoretical framework that can hold artificial intelligence on the one hand, but privacy-preserving technologies really as the flip side that really has to do with adding more noise to these patterns, in fact, hiding and disguising the patterns in noise. And we see this, you know, first of all, entropy is used in privacy-preserving technology, obviously, in cryptography, where, you know, you look for entropy in order to have a source to generate secure keys. The higher the entropy, the more secure the key is. But it's also used in, you know, the networks that we're working on, namely MixNets, where it's a little bit like, you know, you zoom out from cryptography and you're looking at actual kind of, you know, look at data packets across an internet. What we do is we try and kind of scramble the patterns of packets as they move across the network.

So this is really a kind of like flip side to artificial intelligence. Rather than discovering patterns, what we're doing here is we're disguising patterns. And, you know, entropy is kind of like what ties this bigger picture together. So, let's see. Next slide.

That is a cryptography noise-generating MixNets. So, I'm going to wrap up here. But, yeah, so the idea really is that, you know, with artificial intelligence, we've got a little bit of a kind of dystopian hellscape going on, the way that the industry is kind of aligning itself right now. And, you know, I would also go so far as to say that the obsession with prediction is a kind of imprisonment of our future. It actually limits, you know, the range of possibilities, the scope that we have.

And the kind of obsession with discovering patterns across, you know, any possible scenario, including things that are actually really quite deeply personal to people, can have quite profound effects. And I do believe that privacy-preserving technologies can be a counterbalance to those dynamics. And that counterbalance is, you know, both by virtue of how we operationalize entropy in a different way, but also exactly because what does privacy really give you? It gives you a chance to breathe. It gives you a chance to be silly.

It gives you a chance to be awkward. It gives you a chance to kind of make mistakes. And in those scenarios, there is the scope for emergence to happen. And I also think that on a deeper psychological level, and the reason why I wanted to kind of also start to bring in the conversation of psychotherapy here is, I think there is something deeper also that happens in terms of the relationship to ourselves. And so, to return back to where I started with Nur El-Sedir's book, which is titled Animal Joy, this is, you know, a little bit what I picture when I picture kind of animal joy.

It's my niece, and this is a kind of, you know, authentic laughter that, you know, comes from a place of, you know, just pure connection to yourself in an environment that is based on, you know, non-judgment and love, rather than based on measurements, prediction. And something that seeks to kind of limit the range of action. Again, thank you so much for the patience. I was relieved that I couldn't see anybody's faces. Thank you very much.

Do you have a question? Thank you. We have a question from our tool. Do you want to answer the question? Yeah, yeah, let's go.

What do you think will be the psychological consequences of AI and use of entropy in 10 to 20 years from now on? Oh, that's a really good question. I mean, you know, I think there is, I think, so I think that the, actually, you know, speaking of like Byung-Chul's psychogram and all that, like, I think that the internet has already had like profound impact on mental health. And I think artificial intelligence is going to also have a very profound impact on mental health. There is something, you know, when I was reading Byung-Chul's work, and this idea of the psychogram and the idea of, you know, the internet kind of potentially capturing, so he wrote the book before this wave of AI boom really happened.

And, you know, I wonder what he's, he's probably writing something on AI right now, probably already has, you know, but what's interesting is like this idea that our collective subconscious has already been, you know, kind of written up into a psychogram. You know, now I'm seeing friends of mine that are using, you know, AI as effectively like their psychotherapists, like they are asking deep, deep personal questions. And if you consider the fact that AI has already been, you know, training on, you know, deeply psychological stuff, the deeply psychological matter that we have kind of collectively been uploading and sharing on the internet, there is no wonder that it's particularly kind of good for that or like seems to kind of resonate with people. But there are, you know, so many like threat vectors, this is really, you know, we're talking about kind of attack vectors that go right, you know, deep, deep into our kind of headspace and our sense of identity and sense of self. So, yeah, I think it's a pretty risky area that needs to be worked on.

Thank you. That's a good question. Give a big round of applause to Tanya.