trust in an artificially intelligent world
i've been thinking about two important consequences of our current trajectory with artificial intelligence:
- currently, intelligent, autonomous systems are very bad at knowing what is real and what isn't. they are also very easy to trick and manipulate.
- the technologies underlying these systems are centralizing power among those with the most amount of compute possible.
both of these consequences are the raison d'etre of cryptocurrency networks, and more broadly, decentralized systems.
i think there are a lot of misnomers about crypto because of what it has turned into: memecoins, scams, and get rich quick schemes. however, that is not the root of the technology. this is my definition: cryptocurrencies are an asset that 1) can be used to pay for writing to an immutable, censorship-resistant, public ledger and 2) are a reward for helping secure said ledger. okay! that sounds mildly philosophically interesting. who cares though? why is it actually important?
there are a lot of cases where we rely on centralized institutions to provide us information on what is true. some examples include: "who is this person?", "are they who they say they are?", "do they have what they say they have?", "have they done what they say they have done?". in the financial services, there are banks aand credit reporting agencies. on the internet, people use "log in with facebook/google/github/etc", where a more centralized institution is used to verify someone's identity. it should make sense that these institutions are incredibly useful. it should also make sense that these institutions have a lot of power, and because of that, are in a position to exploit that power and influence "people" networks to their own benefit, in ways that may be harmful to others. these institutions are also critical to trust, security, and commerce on the internet; the more these institutions have come online, the more the overall economic value of the internet has increase.
however, ai has changed everything. because of ai, many institutions have become fundamentally untrustworthy and can no longer fulfill their role as trusted central institution, even if they want to. the most obvious example is social networks. ai makes it stupidly easy to pretend you are not who you say you are. this extends to email. another example is traditional media. it is easy to create fake news and it is easy to share.
there is another evolution of ai coming: of intelligent computational systems acting with autonomy. i think that we already have some level of this, and i believe it is inevitable that it they will become more intelligent. even if ai simply plateaued at the current level of intelligence it is at, it would already be sufficient enough that people will deploy ai systems that operate autonomously. people are already starting to do this, and the tools are already easy enough that people can spin up ai content farms with a few words, making fake videos, email, spam attempts, trying to cold email people, and more at scale.
in this world of millions and billions of autonomous agents, they will need to know how to tell what is real and what isn't. this is where i believe a public, decentralized, immutable ledger that autonomous agents can write to and read from seems valuable. autonomous agents will need a system that is easy to access, publicly available, and cryptographically secure to record their actions, verify the actions of other agents, and identify assets. this is precisely what cryptocurrencies can provide.
one example of decentralized systems replacing traditional media is with prediction markets. kalshi and polymarket more accurately predicted the 2024 presidential election than any media source. it is increasingly becoming a more reliable source of understanding events than the media, which increasingly optimizes for rage-bait and clicks instead of unbiased, factual recounts of what events are occurring. because polymarket and kalshi are built on decentralized blockchain ledgers, the information of how the markets settle are much more available and reliable to ai agents. in the long run, i'm betting that ai agents that need to respond to what events happen will trust on-chain prediction market activity compared to traditional media content (it is also way more cost effective because you don't need complex streaming and embedding infrastructure tied directly into major media providers, but that is a different discussion).
this is just one example of how ai agents can benefit from interacting with cryptocurrency networks. some other areas i think are worth exploring include agent reputation systems, third party task verification, automated escrow accounts for transactions, and agent audit logs. overall, this is an exciting time to be thinking about computing: the architecture, systems, and capabilities are transforming. i look forward to exploring how decentralized technologies can enable more reliable and useful intelligent autonomous systems over the next decade.