The three components of a persistent, credentialed, global education network.
Trustworthy AI Education?
Artificial Intelligence will accelerate and democratize our ability to educate the entire planet. The pursuit of the global education ideal will potentially undercut the higher education business model. However, most concerning is the authenticity of the information AI will proliferate. (1)
We are headed towards a crisis of data accuracy and intellectual property ownership. (2) If AI is to be trained on the knowledge we feel is most valuable to us, we may be incentivized to build an infrastructure to validate it. Teachers must evolve to become authenticators.
To better visualize this, let’s imagine three components of a globally persistent teaching network.
These components are:
1. The AI teacher
2. The Reputation Nodes
3. The Identity Guilds
The AI Teacher
ChatGPT awoke the planet to AI's potential. Developers around the world are experimenting with Large Language AI Models (LLMs). In a short time, AI-powered chatbots will be everywhere, across every industry.
These chatbots may gain traction with documentation and support. Google has road-mapped the rollout of its documentation AI. (3) The blockchain we work with, Solana, is already using AI chatbots to serve up documentation to their developer base. (4) AI will undoubtedly be our internet search companion. (5)
Full-on education, is not far behind.
Kahn Academy is taking a bold step forward using GPT4 with their educational chatbot "Kahnmigo." (6) Duolingo, a language learning platform, has developed its own approach to the model. (7) These aren't just companies providing support documentation for a product. The knowledge these kinds of platforms provide IS the product.
The previous wave of chatbot hysteria turned out to be mostly-canned responses, in an if/then style framework. (8) These new bots, powered by super language models, will promise understanding, adaptability, and provide confidence to the students as they learn.
It's not just curriculum, it's teaching.
AI Teacher: A persistently accessible intelligence that teaches, gamifies curriculum, and credentials an individual with an internet connection.
While language models power the curriculum, generative video can visualize the front end. There is already ample evidence of AI video creation. Synthesia and Movio are two of the early movers in this space. (9) For a monthly price, you can generate a variety of customizable near-human avatars. Admittedly a bit “uncanny valley,” but there is little doubt that the fidelity of the performance will grow to be the face of our AI mentors.
As a teacher, my work consists of me writing lectures, demos, and workshops to perform at a preordained time within a classroom. I get paid by a credentialed institution for this labor. In a few short years, this content will be entirely fed to language models, which will do the teaching instead.
So, what happens to the credentialed institution?
Reputation Network
AI models sometimes lie. They don't mean to. In fact, they don't even "mean". They are simply making an increasingly accurate series of guesses. My email prediction often guesses my responses incorrectly. How accurate will the fully generated New York Times essay be?
In the AI world, the wrong guess, despite reading and feeling accurate, is called a hallucination. (10)
In education, hallucinations are an insanely dangerous thing. We have already seen what falsehoods do to politics. This could be far worse. If AI teachers are to proliferate, there will need to be a way to credential their knowledge. Users will need to know the authenticity, and trustworthiness, of the system.
Universities are credentialed by legally audited centralized organizations. Having these credentials allows universities to become, in the eyes of the law, the rightful issuer of a highly prized subject matter degree. Bachelor, Associate, and Master are all signatory values of accomplishment. They provide evidence to an employer that the student has "mastered" a curriculum of skills. (11)
But soon our curriculum will be lost in the latent space of an AI supermodel. (12) How can a human organization possibly regulate it?
Let's imagine an immutable ledger that tracks the transactions of nodes in a network. Nodes on this network can gain reputation, and thus value, with measured transactions.
Bitcoin has a fiat value. Ideas and artwork have value too. Value can be knowledge, skill, or authority --- Intellectual Property.
What if the top writers, the top artists, the top game players, and the most fabulous fashion models peer-review one another?
Could this be the source of truth?
Reputation Network: A council of peer-reviewed, highly valuable nodes on a blockchain network.
As I watch the development of the web3 ecosystem, I see the emergence of Decentralized Autonomous Organizations. Generally, these organizations are online professional communities that sometimes incubate grant voting systems. (13) Most have not found an operational structure.
But what is increasingly evident is that the success of a DAO may not be the massive number of participants that get pulled into a large-scale fundraiser, but the recruitment of a highly productive few who step up to be the leaders of the network. (14) These leaders provide authority and increase the desire of others to join a valuable organization. The nodes of the thought leaders create the reputation, and thus value, of the network.
If these DAOs do find a path to operation, perhaps they could be used to manage these reputational nodes and provide authenticity to the AI teacher.
The reputational nodes are the most valuable of the network. But valuable to whom? How do we define the network and the subject matter of the data within?
Identity Guild
I spend my time in the communities that reflect my identity.
I am a writer to some. A game developer and animator to others, and still again, a beer-loving drinking buddy to others. I spend time in places IRL or in discord, because I enjoy it, and I enjoy sharing my value with others. My identity is multifaceted, operating slightly differently in each, but also as a unified whole.
During the Renaissance, much of the art, research, writing, and philosophy was empowered within officially recognized guilds. These were collections of like-minded thinkers who had the legal authority to defend an art form’s intellectual property, but also the application and tooling of their livelihood. (15)
If I were a young fresco painter, I might apprentice myself to a journeyman, or move to the city to join a powerful guild. It would be the reputational way for me to make a pledge to an art form, and work my way through (hopefully up) the system. Guilds offer financing and support. They also create networking opportunities, which is really the point of creating a network.
However, one can not research guilds without discovering economic criticism. The largest argument I see is that guilds centralize and plateau knowledge. It's the age-old fallacy of "once in power - resist the change." To the rabid capitalist, guilds delay innovation, suppress outside entrepreneurial work, and form unhealthy reliance on the centralized state. (16)
But let's imagine if a guild network cultivated its own AI teacher, shepherded by the reputation nodes. Everyone who was in the guild had access to the most current ideas through AI. New entrants could be gamed from apprentice to master.
I suspect that guild structures are the natural formations of creative humans. Check the history books, we've done it a lot. (17) It's part of the cycle of centralization and decentralization. We have a king, and then we get sick of it --So, we democratize and vote on stuff.
Companies are started by Steve Jobs and Steve Wozniak, but nations are started with a document that says "We the People." By declaring who they are, a guild may create legal backing to do the things they wish to do as a collective.
They declare their identity.
Networks pull in people who seek identity. By declaring participation in an identity guild, they collaborate to defend it.
Education will change
If AI is to teach us, we need to put systems in place to authenticate and protect its data. Those who are most passionate about their identity will create the strongest guild governance. The reputation of that guild will depend on the nodes that lead it.
If the system can come to balance, potentially, the creation of credentialed AI will be the most trusted in its domain.
Decentralizing control over knowledge may be our only hope for trusted AI. If we rely on centralized, and potentially biased authenticity, the result could be disastrous.
Nye Warburton is an animation technologist and educator. This essay was written in collaboration with the BanklessDAO Writers Guild. Header image generated by the Stable Diffusion Platform Leonardo.ai. Text and essay imagery were written and designed entirely with the human labor of the author. (April 2023)
web2: https://nyewarburton.com
web3: https://mirror.xyz/nyewarburton.eth
Notes and Reference
(1) Most concerning is the authenticity of the information AI will proliferate. | Artificial Intelligence: An Accountability Framework for Federal Agencies, Public Release (June 2021)
*(2) We are headed towards a crisis of data accuracy and intellectual property ownership. * | Key challenges of artificial intelligence. Intellectual Property: Protecting your AI and its creations, Ertugrul Akinci, Kargan Dora, and Allison Bender, Business Going Digital
(3) Google has road-mapped the rollout of its documentation AI. | A new era for AI and Google Workspace, March 14, 2023
(4) The blockchain we work with, Solana, is already using AI chatbots to serve up documentation to their developer base. | Ask Solana U, an AI Chatbot for Solana Development
(5) AI will undoubtedly be our internet search companion. | Reinventing search with a new AI-powered Microsoft Bing and Edge, your copilot for the web, Yusuf Mendi, February 7, 2023
(6) Kahn Academy is taking a bold step forward using GPT4 with their educational chatbot "Kahnmigo." | Introducting… Kahnmigo, Kahn Academy Help Center
(7) Duolingo, a language learning platform, has developed its own approach to the model. | Duolingo is now equipped with GPT-4: Here’s what it can do for you, Sabrian Ortiz, March 15, 2023
(8) The previous wave of chatbot hysteria turned out to be mostly-canned responses, in an if/then style framework. | Example: Chatfuel.com
(9) Synthesia and Movio are two of the early movers in this space. | Example: synthesia.io | movio.la
(10) In the AI world, the wrong guess, despite reading and feeling accurate, is called a hallucination. | “What Makes AI Chatbots Go Wrong?”, Cade Metz (The New York Times On Tech)
(11) They provide evidence to an employer that the student has "mastered" a curriculum of skills. | We don’t value Education. We value the Credential, Brandon Busteed, Forbes
(12) But soon our curriculum will be lost in the latent space of an AI supermodel. | Understanding Latent Space in Machine Learning, Ekin Tiu, Towards Data Science
(13) Most of these organizations are professional communities, perhaps with some grant voting systems. | Example: Nouns.wtf | Bankless DAO
*(14) But what is increasingly evident is that the success of a DAO may not be the massive number of participants that get pulled into a large-scale fundraiser, but the recruitment of a highly productive few who step up to be the leaders of the network. * | A Hyperstructure for Learning, Courtland Leer
(15) These were collections of like-minded thinkers who had the legal authority to defend an art form’s intellectual property, but also the application and tooling of their livelihood. | Medieval Guilds, World History Encyclopedia
(16) To the rabid capitalist, guilds delay innovation, suppress outside entrepreneurial work, and form unhealthy reliance on the centralized state. | The Problem with Guilds, from Silversmiths to Taxi Drivers, by Justin Fox, Harvard Business Review
(17) Check the history books, we've done it a lot. | “The Square and the Tower: Networks, Hierarchies and the Struggle for Global Power, Niall Ferguson, Penguin, 2018