Is security already dead? It’s quite possible to fool oneself into believing that you are secure while browsing the internet or, worse, when you connect AI to your systems/accounts. This is especially true in this age of AI. As naive as this may sound, we are not safe, and neither are our nations.
Many ordinary people, myself included, do not worry enough about our privacy being misused or violated. “Let it happen. What do I have to lose, right?” we might think. The problem, however, comes when the government we rely on to keep us safe adopts this new technology and connects it to citizens’ data repositories without proper safeguards.
When I say safeguards, I don’t mean creating policies or laws to restrict the use of such technologies; that is almost impossible as these systems keep advancing. I’m talking about building the right tools to safeguard our nations. But how does one build such tools?
The first step, I think, is to recognise that there is a threat. A government that knows about a threat to its nation ahead of time is in a stronger position to respond. The facts are already out: evolving technologies such as artificial intelligence create new opportunities for sophisticated attacks. Think of biological, cyber, information, and psychological warfare. I am not an expert in these areas, but given the direction we are heading in as a civilisation, it seems clear to me that some, if not all, of these outcomes are true. How can we use this powerful new tool to our advantage without creating a “mess”?
There are limits to what intelligence can predict efficiently. Computational irreducibility is one of them. Some things cannot simply be predicted or simplified; you have to run the system to see what will happen. This is a powerful concept and, I think, a useful guide for humans navigating the computational universe: it helps us understand where even AI faces limits.
What am I proposing here? Irreducibility as a possible line of defense. For some systems, predicting how they will behave requires working through essentially the same computation as the systems themselves. Greater intelligence alone does not remove that limit. Could we use this to make certain attacks harder? My proposal is to explore whether irreducibility can help us build stronger defenses into systems.
Of course, this could work only in certain areas, such as cyber and information warfare, and not in others, such as biological warfare, because biology is weird. Defending against a bio weapon requires other approaches, which might prove harder computationally but it opens up research opportunities in computational biology.
Dilate is built around this principle. Our goal is to build systems designed to resist attacks and remain dependable under pressure, i.e., systems that are “impossible” to hack. To do that, we have programmes such as Loopseed, which I will introduce in my next post, for researching and testing current AI systems.
Other projects, such as Quanta, focus on building and strengthening humanity’s future by investing in children’s computational thinking and preparing them for life in a computational universe. We have pilots in place for these programmes, and we encourage everyone to get involved. I will introduce these projects in future posts, so keep an eye out.
The opportunities extend beyond security. We believe our engineering work will help shape the future of our civilisation through advances in governance, scientific discovery, national security and, most importantly, the development of critical infrastructure.
Finally, Dilate welcomes anyone or any organisation with ideas on how to secure the future of our nations and humanity as a whole. Get in touch if you’d like to work with us.