Strange Protocols

Giving birth to a conjecture — a proposition that is suspected to be true, but needs definitive proof — can feel to a mathematician like a moment of divine inspiration. Mathematical conjectures are not merely educated guesses. Formulating them requires a combination of genius, intuition and experience. Even a mathematician can struggle to explain their own discovery process. Yet, counter-intuitively, I think that this is the realm in which machine intelligence will initially be most transformative.

Why mathematics is set to be revolutionized by AI

Pattern: AI-driven mathematical discovery and protocol development.

Weak Signal: Cambrian explosion of mathematical conjectures and protocols.

Counter-Signal: Math and protocol dark matter. Drawing an analogy to dark matter in physics, the proliferation of AI-generated conjectures and protocols may become increasingly difficult for humans to interpret or validate.


The pattern, weak signal, and counter-signal triplet is a form of information compression. This structure allows me to express perspectives and uncertainties that would otherwise require lengthy explanations.

In an AI-mediated world, alongside weak and strong protocols, a third category may be necessary: strange protocols. Drawing on AlphaZero's move 37, which confounded human Go players, and the concept of strange attractors in complexity theory, where intricate patterns emerge from simple rules, strange protocols could represent AI-mediated protocols that emerge from the interactions of vast networks of AI agents operating at unimaginable speeds. These protocols might exhibit unconventional properties that defy existing human understanding, yet could potentially lead to surprising solutions.

I might incorporate strange protocols into my PILL project, The Persistence of AI-Mediated Protocols.

Any thoughts?

1 Like

Starting to sound like quarks names

Strange = within existing protocols but rare because conventional agents don’t think to run the protocol that way?

I’m reminded of legal but weird “not cricket” moves in cricket List of Mankading incidents in cricket - Wikipedia

Robots often learn weird ways to behave that satisfy the constraints and minimize a cost function but look nothing like gaits expected from the body

1 Like

This is a very interesting approach on how we conceive AI. As you say, although AI is not intelligent, its results are something outside human rationality/intelligence and we still don’t have parameters to measure it. Maybe strange protocols will be able to translate?

1 Like

My first thought relates to symmetric escalation (or symmetrical schismogenesis) in the development of AI capabilities and protocols. A loose historical analogy can be drawn from antivirus and antimalware technologies. To address vulnerabilities in operating systems and application software, a counteroffensive developed in the form of increasingly sophisticated cybersecurity measures. This escalation led to the growth of a capable and profitable cybersecurity industry. Similarly, as AI systems become more advanced and potentially disruptive, we will see a corresponding rise in the development of AI counteroffensives and protocols.

A second thought relates to hyperconnectivity and the rapid adoption of AI technologies. An example is ChatGPT, the fastest-growing consumer internet application in history. The public’s rapid adoption of this technology, which essentially offered a user-friendly interface to LLMs, suggests to me that future protocols will not be designed by committee. Instead, they are more likely to evolve organically through a process of AI-mediated public stress testing.

1 Like