Mind Crunches #25: Meditations on Tech, Decentralization & Complexity Theory
In search of an intellectual punctuated equilibrium
đĄThings I Have Been Thinking
The Convex vs Concave Debate about Decentralization: one of the things I have realized about the world is that most problems/challenges/debates can be viewed as a battle between convex and concave philosophies. If you arenât familiar with these concepts, a concave pov is when someone tends to think that non-extreme (something in the middle) solutions are the best while convex theory promotes a binary way of thinking (this is the only solution). An area that the convex vs concave pov applies very well is that of decentralization. I work in an industry that is obsessed with decentralization and this is one of the reasons I love working in crypto. But nature/history/personal experiences show that an obsession with decentralization can often lead to slow decision making, inaction and chaos. There is a deep biological tendency to work collectively and in a decentralized way (think of cells inside an organism) but there is also a natural urgency to act/decide fast in a very centralized fashion. So how can we build an agile and adaptive model to use Minimum Viable Centralization (MVC) without succumbing to the challenges of centralized concentration of power? How can we learn from nature/biology where organisms often use very centralized decision overruling decentralized systems for the greater good of the whole? So far, the framework of quadratic voting/funding seems to be a well-thought approach to this problem. There is also a deeper philosophical question that arises from this debate. Should we judge the âwholeâ for the actions of one of its âunitsâ? Or does each unit have its own existence/decision making progress/entity? In real terms, should we judge/cancel/criticize a political party/media organization for the bad/unethical actions of one of their candidates/journalists?
On Technological Parasitism: this is something I have been monitoring on a daily basis in terms of how Web2 and Web3 influence one another. It was recently reported that the New York Stock Exchange is considering to open up 24/7 trading as a way to adapt to the always on nature of the crypto economy. Most people, including Uniswapâs CEO, encouraged this and perceived it as a sign of the âtraditional worldâ adopting crypto practices but I am bit more pessimistic. I see this as a form of parasitism (in that case Web2 being the parasite) where the traditional world copies the best ideas and maybe some tech primitives from Crypto but not the core ethos/values of decentralization. If that happens, then we end up in the worst equilibrium where major corporations keep harnessing power/revenues and they can still communicate they âtransformedâ.
On HyperStructures: Since the moment I first read about the concept of hyperstructures, I have been fascinated with them. In its literal sense, a hyperstructure is an algebraic formula but in the context of networks, a hyperstucture is an open, permissionless and expansive protocol that runs on a blockchain and is coordinated by cryptography. The analogies with ecologies and evolutionary biology are numerous. At first, I realized that GenAI will enable us to build entrepreneurship hyperstructures where builders will build rails for other builders and so forth and so on (the inspiration was Shopifyâs announcement of Sidekick). But in a Web2 environment, you still have some closed doors/permissions/moats that donât allow the ecosystem to grow organically. But this isnât the case with blockchains. Lens or Farcaster are open protocols on top of blockchains that can grow exponentially allowing anyone to build their apps on top of them, the protocol teams can build modular ârailsâ to make app creation easier and apps can form new added value when they connect with each other. They are positive sum infinite games and we never had something like that before. The coolest thing about them isnât only their expansive nature but also how they can change the behavior of static players. For example, we now see more and more VCs launching OSS products to kickstart the flywheel of app building on top of hyperstructures. a16z Jolt is a great example.
Tech/Political Ideologies: still debating myself about Thiel's famous analogy AI=communist/Crypto=libertarian. The analogy that seems to work for me at the moment is: AI=mercantilism, Crypto= classical liberalism (democracy+capitalism+meritocracy), ZK=libertarianism. Glen Weylâs post on 21st centuries political ideologies is still a must-read. For people who are interested in what framework is the closest to my political beliefs, I highly recommend reading the Power-Sharing Liberalism manifesto. I look forward to a future with more plural political ideologies that donât fall in the trap of Left/Right but rather create a new worldview through an ongoing and dynamic deliberation of their constituents.
On Procrastination as a Competitive Advantage: Procrastination has a bad name but it can unlock amazing alpha if you accept it and convert it into an intellectual serendipity and combinatorial thinking enabler. I have found the skill of âstaying informed in different fields/disciplinesâ to be one of the most important pillars in my career. Staying informed and developing combinatorial thinking gives you an amazing edge if you are working in Strategy/BD roles. Many people conflate "staying informed" with unproductive procrastination and twitter scrolling but, at least in my experience, it's a skill that requires constant optimization (information diet) and rigorous planning (when to read, where to store info, taking notes etc). It took me many years to understand that "staying informed" is a competitive advantage and I was happy to see Josh Wolfe (Lux Capital) and Haseeb Qureshi (Dragonfly) recently sharing their stories on how procrastination has actually helped them stay on edge and I am now convinced that we have been fooling ourselves for years. Procrastination is good. You just have to embrace it!
Do Your Own Research (DYOR) Limitations: How realistic is to expect everyone to do their own research, especially in highly complex issues? Does DYOR lead to more extremist views and amplification of conspiracy theories or does it lead to a wiser crowd equilibrium? And how can someone rely on his own research when itâs also true that most media/academic journals/government sources are ideologically biased? Blockchains could be (part of) a solution providing a proof of truth.
đ§âđ» Things That Excite Me
The concept of collective intelligence:
I have written in the past that Michael Levin is one of the most fascinating scientists out there. His latest paper argues that intelligence doesnât have a narrow, specific definition (as we tend to believe) but itâs rather a concept of collective decision making between competent sub units. If you donât like reading papers, his Noema article is a great summary of the basic findings.
Sakana AI, a Japanese GenAI startup developing nature-inspired LLMs. Their latest research proposes an evolutionary approach that automatically discovers effective combinations of diverse open-source models, harnessing their collective intelligence without requiring extensive additional training data or compute.
What if we interpret History wrong? What if we could find new patterns that explain differently many historical facts? This research describes how using stochastic processes to analyze historical datasets could reveal surprising patterns that have gone unnoticed in previous models. A stochastic model, which incorporates uncertainty and randomness, would treat historical shifts not as deterministic but instead as probabilistic.
Bacterias are technically motors that can rotate and switch directions.
The Sphere: I recently had the chance to visit The Sphere in Las Vegas. I consider myself a techno-optimist but I also recognize that the older I get, the more cynical I become about tech-driven narratives. Which means that I didnât expect to be amazed by The Sphere. I have to admit that it was a transcending experience. The enormity of the venue definitely creates a wow factor but I believe the reason it touches so many people is because itâs a sensory-centric innovation. You donât experience something new/unknown like you do with AI or robots. In a way, the experience is super basic. You are watching a movie. But then it hits you. Itâs the first time that you donât have to adjust your senses to the size/shape of the screen. You can look in any direction and see something. Exactly as you do IRL with your peripheral vision. The Ultra-HD makes the whole thing hyper-real. Human senses on steroids. The Sphere= Technological Ayahuasca.
đ€ Speaking at Panels/Podcasts
I moderated a panel, organized by 1kx, about the future of decentralized and verifiable AI at ETH Denver.
I was invited at Delphi Ventureâs podcast to discuss about the intersection of crypto and AI with Casey Caruso and Colin Gagich.
Casual conversation with Panos Papadopoulos and Alex Alexakis from Marathon VC about the state of crypto and the transition from Big Tech to Web3.
đ Mind Crunches
If you are into Progress studies (as I am), take a look at the newly formed Abundance Institute and this piece on how we can re-brand Progress.
Plurality, the book from Glen Weyl and Audrey Tang about the future of collaborative digital democracy, is finalized and will be soon printed and published. I was honored to be part of this project from the very beginning and I have come to realize that Plurality is now the framework I am using most often to think of complex, multi-disciplinary challenges. Highly recommended.
Crypto is back baby!
Itâs been 18months since ChatGPT caught the world by storm which means that we can now have a solid understanding of where the money is made, if there are moats and how the stack will evolve. At the moment, most value is accrued on the Semi/GPU level (duh) but the trend shows that Apps will be able to build stronger monetization and rely less on the NVIDIAs of the world. I believe this is a valid hypothesis but I also think we will see more innovation in the Decentralized AI space that will completely transform the pyramid into a different shape with new players like Data DAOs, aggregated compute layers and synthetic data marketplaces.
Hottest trends in GenAI: small, on device, no cloud LLMs and AI agents trained on synthetic data.
I recently had the great opportunity to meet Emad Mostaque in person. He recently stepped down from CEO at Stability AI and he is pursuing a very inspiring vision about decentralized, locally aligned and transparent AI world. One of a kind! More news on what heâs cooking soon.
How to do things if you arenât smart or talented!
Memecoins and DAOs are very GenZ concepts and they have attracted (not always unfairly) a lot of criticism but I find them to be very interesting social experiments on community building, culture alignment and coordination. Higher has the potential to be the first decentralized brand.
Based on the latest Stanford research, Industry has released 80 more AI foundation models than Academia. Not sure how I feel about this.
NTUA is the MIT of Europe. We can do even better.
Lux Capital published its first risk gaming simulation. This (risk games) is something that they have been doing/offering as a service to many organizations but now itâs the first time they publish one.
Regardless of your job, we are all in the business of building worlds. Great read.
What if there isnât a great stagnation? I am personally not convinced by the core argument of the article but the concept of âcumulative cultureâ is fascinating.
Hugging Face is a force of nature. Still havenât figured out how other startups can actually compete with HF. They seem invincible.
Simple steps on how to build a new city in California.
How is Dwarkesh Patel preparing for his podcast interviews? For what itâs worth, I have a very similar philosophy about everything in life. (and mostly in business)
Why civilizations collapse? The tldr; is that itâs very difficult to reform societies that accumulate social and cultural technical debt but make yourself a favor and read the whole thing. Samo Burja is excellent.
Palmer Luckey is hands down the most exciting, refreshing, anti-fragile, fun tech CEO out there.
Take more shots. Whatâs there to lose?
It is now well established that high quality data create better LLMs. But where do you find clean data? Hugging Face released FineWeb, a dataset of more than 15T tokens of cleaned and deduplicated english web data from CommonCrawl.
Recommended podcasts:
Dwarkeshâs podcast with Zuck. Fascinating conversation on the future of open source LLMs and Zuckâs decision framework. Ate-a-Piâs summary is great.
Tylerâs conversation with Jon Haidt. Itâs been almost 8 years that I have been following Cowen and I donât remember a single other episode where he had a more heated debate with someone. I havenât read Haidtâs latest book but it seemed to me that his rationale/argumentation is much more logical than Cowenâs. Also, people start finding holes in Cowenâs polymath namedropping.
Recommended Blogs:
Strange Loop Cannon, Rohit Krishnan
Guzey, Alexey Guzey
Recommended Books:
Quote of the month: âHypocrisy in search of social acceptance erodes self-respectâ
Photo of the month: a glimpse of the hyper-real experience at the Sphere.