Values & Philosophy Are Everything
How We Build Smart Agents Matters Deeply
Values Are Reflected In Code
A friend once told me that projects don’t compete on technology, they complete on philosophy. I have found that to be very true.
For example nothing prevents the Bitcoin developers from copy & pasting over the Ethereum codebase tomorrow. The reason this doesn’t happen is because the Bitcoin developers have fundamentally different set of beliefs from the Ethereum developers and this is reflected in the code.
- Proof of Work vs. Proof of Stake.
- 21 Million Bitcoin limit vs flexible issuance & burning of ETH.
- A narrow set of scripts vs. Ethereum Virtual Machine/ Smart Contracts.
- UTXO (Unspent Transaction Outputs) vs. Account / Balance Model.
These are all technical choices, rooted in beliefs. Now certainly both projects also share many values. Especially their peer to peer networks, censorship resistance, immutability, & open source foundations. However they have picked different ways to pursue these goals. As a result they have attracted very different user bases & are stronger in different use cases.
Most Bitcoiners will discuss Bitcoin’s ability to store value as digital gold.
Whereas Ethereans talk about Dapps, DAOs, DeFi & Smart Contracts.
The Filecoin community is all about IPFS, storing data & Web3 & so on.
There were intense debates about all these values, choices and trade offs in 2013 / 2014. Famously the Bitcoin developers turned away Vitalik’s proposals for scripting on Bitcoin and so projects such as Ethereum were launched to explore and build out areas of philosophy where Bitcoin’s developers refused to go.
I believe we are at that type of pivotal moment again in 2023.
Value Choices For Smart Agents
So lets examine the landscape of AI Agents.
OpenAI, Google, Bing all these are large centrally hosted LLMs are:
Monitored, Censored, Centralized,
Meta is the only of the large tech companies to mostly open source their weights for LlaMa and now LlaMa2. Though they aren’t providing a UI or interfaces just the raw generative chat model.
There is a massive open space for a personal AI. One that runs locally, keeps the user’s information private and runs all on open source so there aren’t any license fees or monthly subscription costs.
Where Smart Agents stand:
Open Source, Private, Decentralized, Permissionless, & Aligned To Individual Humans is the objective.
This is where the Smart Agents are positioned.
What Linux is to Servers.
What Android is to Mobile Phones.
Smart Agents are to AI.