Why Free Learning Needs Free Artificial Intelligence

 

'Free' learning means two things to me: first, the idea that learning ought to be zero cost, and second, the idea that learning ought to be open and inclusive. Together these paint a picture of access to learning as part of the basic infrastructure of a society. 

We're not good, yet, at providing basic infrastructure. Most societies have a free and open road network along with access to basic security, including fire and police service. In Canada we also have free and open public education service for children up to Grade 12, as well as access to a range of essential health care services. Still, basic necessities like water, sewage, food, housing, electricity and information services are all things we pay for, with the result that too many go without. This is even more true on a global scale.

My focus is on education because it's where I ended up and what I know well. My understanding of 'education' and 'learning' in a society is far broader than schools, colleges and universities. Yes, these are essential institutions, but learning is a much broader concept, including workplace learning and personal learning. It is also a concept that ranges from access to information services, including news and data, through to access to assessment and entry into the marketplace. 

It takes a lot to provide these services. Buildings and infrastructure must be constructed and maintained, people need to create and provide access to learning resources and opportunities, people need time and space in their own lives to engage in learning activities, and the whole system must be coordinated and paid for. We spend 6.6 percent of our national GDP on education, or maybe 130 billion dollars, give or take, plus whatever we spend out of our own pockets for data and information services such as books, internet, and software.

It's a lot of money, and I don't consider the need for it to be self-evident. Learning isn't an end in itself for most people; it's a means to other ends, and it's these other ends that I am really working to support. I've tried to capture that in my own vision statement, expressing the idea that each of us ought to be able to live, grow and develop in our own way, to the maximum of our capacities, if we wish. It's an extension of John Stuart Mill's maxim that "the only freedom which deserves the name, is that of pursuing our own good in our own way", along with the Kantian idea that we should "treat humanity... at all times also as an end, and not only as a means."

Learning is a necessary, though not sufficient, condition for the realization of this vision. That said, a society in which learning is widely accessible and widely attained is one that is far more likely to achieve this vision, as people realize for themselves first, their own inherent good, and second, the inherent good of others. By 'inherent good' I mean here goodness in itself, and not measured against some external standard such as 'value' or 'worth' or other economically laden terms. 

Learning is not only resource-intensive, it is also difficult for the learner. We cannot, and probably never will, simply absorb knowledge through some sort of direct transfer. Learning is, as some people like to say, embodied, which means that there is no distinction between the informational content of our brain and body and their physical structure. I have described this embodiment by saying it is connectivist, that is, that knowledge is the connective neural structure in the brain, and that learning is the development of that structure. This means that knowledge is grown, not acquired, through a process of practice, experience and reflection.

In this way the development of a learning infrastructure is analogous to the development of an infrastructure for agriculture and human growth. We can't just focus on the provider 'giving' food to a person; an entire system needs to be contemplated, where transfer of food from producer to consumer is only one small step in a social system that grows the food and a personal system that uses the food to grow, a system that (interestingly) represents about 7 percent of Canada's GDP.

It is interesting that in the statistical accounts learning is represented as a cost while agriculture is represented as part of our national income, as though they were somehow distinct in nature. But neither is uniquely a form of income or expense; they may be either for any individual, and in a society, they are both, and globally, their status as income or expense balances out: the cost of learning is exactly what people earn from providing learning, while the cost of food is exactly what people earn providing food. 

So when I talk about 'free learning' there are two imperatives that much be considered: first, the real cost of providing access to learning to everybody, and second, the real income from learning that a large number of people depend on for their own livelihood. Supporting 'free learning' isn't simply a matter of reducing the cost of learning to zero, nor even of lowering the overall cost as a percentage of social cost and/or income.

It is a question of re-balancing; of providing access to all no matter what their income (or other demographic status) instead of providing more access to some and none at all to the rest. And it is, similarly, a question of providing access to (potential) earnings or revenue from learning to all, instead of providing a large revenue to some, and none at all to the rest. Both of these need to be in some sort of balance.

To put the matter more concretely: suppose we are concerned about access to education in Eswatini. Though the country spends 5 percent of its GDP on learning, its GDP is so low the result is that education is chronically underfunded, and many do without. Simply 'giving knowledge for free' is insufficient here. It's important that Eswatini develop its own learning economy, so that its capacity to generate income increases as well as its capacity to provide learning opportunities. To do otherwise is to further exaggerate the disparity between rich nations and poor nations.

This discussion is often conducted using the lexicon of 'colonialism'. In my previous post on this subject I defined colonialism loosely as follows: "Colonialism is (at least in part) the appropriation by one society of some other society or culture's productivity or wealth for its own gain. And it is (at least in part) the imposing of laws, values and cultural elements by one society onto another." 

This definition makes clear the idea that the imperative of 'free learning' isn't simply one of income and expenses. There are other more ineffable currencies that form a part of any national economy: its language(s), its system of laws, its culture and traditions, its social knowledge, its values and beliefs, to name a few. The practice of 'giving knowledge for free' is as harmful to these parts of a national economy as it is to the financial side of the economy.

When a nation can no longer develop its own learning economy, its language and culture (etc) are depressed. The learning provider extracts these from the economy without providing any return, and uses it for its own purposes, while at the same time earning an income developing learning resources that are then distributed (for 'free') to the recipient nation. The money - in the form of international aid, donations from grants and foundations, and even volunteer labour developing (say) software and content - is spent entirely in the 'donor' nation.

The development and use of artificial intelligence has made this process clear for all to see, which is why its use has resulted in a range of very reasonable objections, which again I discussed in my previous post. Its use results in a very apparent form of colonialism, in which a society's language(s),  system of laws, culture and traditions, social knowledge, values and beliefs, etc., are extracted, repackaged by the 'producer', and then sold back to us, in a form inevitably altered by the values and beliefs of those who produced it.

The only real difference between what we'll call 'AI colonialism' and 'Good Old Fashioned Colonialism' is in who is being colonized and who is doing the colonizing. In the case of GOFC, it was one nation colonizing another. In the case of AIC, it is one sector of the economy colonizing the rest. Though if we pause and consider for a bit we'll find it's not so different after all: in most societies, developed and otherwise, there is a structural colonialism, where one wealthier sector of society extracts value from the other, and then sells (or in the case of charity, 'gives') it back as a value-laden alternative.

I am so sympathetic with those who are opposing AI on these grounds, though my charity is extended only grudgingly to those who have only recently made the switch from colonizer to colonized. And my real loyalties are with those who have always been colonized - not only those in Eswatini (who have to their credit have resisted colonization better than many) but also those in my own society and those like mine, who contribute with their language(s), system of laws, culture and traditions, social knowledge, values and beliefs, etc., and find an educational system - and knowledge economy generally - sold back to them, inevitably changed by the values and beliefs of those who performed the appropriation.

This is an unsustainable model. Over time, it not only reduces the wealth of the subjected population, it also reduces the capacity of the provider (or 'donor') community generate wealth without these inputs (one imagines that a company like Disney would flounder without the privilege to incorporate and repurpose Arab or Indigenous culture and folklore). 

This is why, when I wrote for the OECD on sustainable open educational resources (OER), I wrote in favour of a community-based model. "The distinction between producers and consumers need to be collapsed. The use of a learning resource, through adaptation and repurposing, becomes the production of another resource." It's a model of a community producing learning resources for itself, rather than importing them at great cost in money and self-identity from a 'provider'.

As I said off the top of this article, we're not very good at providing this infrastructure. What I meant by that is that even before the arrival of artificial intelligence - long before, in fact - knowledge and learning resources were produced by extracting value from a community, repurposing it, and selling it back to that same community. What was removed was not only the community's own wealth, but even the very possibility of that community providing learning resources for itself.

It's a model created in two steps: first, defining a knowledge and information 'commons', and then applying some logic from Locke to convert it into property: "It being by him removed from the common state nature hath placed it in, it hath by this labour something annexed to it, that excludes the common right of other men." In other words: he who picks the apple gets to sell the apple. And thus the fruits of our society and culture, one by one, with inexhaustible patience, have been appropriated by private industry; and having amassed the lion's share of the world's wealth, they circle around, seeking to appropriate the last.

I am in favour, as are many others, of taxing this industry, to achieve a redistribution of wealth, restoring some balance to our society, which cannot continue in its current fashion. But a mere transfer of wealth will not address the underlying cause. The model of extraction and resale is itself what perpetuates the inability of a people to fend for itself - to feed itself, or to educate itself.

Working toward this end has been a long and arduous road, through pre-Open Source licenses in software like George Reese's Nightmare, through a 'freenet' in western Manitoba (called 'Westman Community Network' because someone trademarked the widely used term 'freenet'), through open educational resources (OER) and massive open online courses (MOOC), through local news cooperatives, to where we are today.

The purpose and intent was never to create a 'commons' that could be aggregated and exploited by commercial interests - that was the source of my longstanding debate with David Wiley. It was to create and enable mechanisms that helped a community build and support its own learning resource network that would not need to depend on proprietary journals, textbooks, or educational resources - and for that matter, to provide an alternative to proprietary and costly educational institutions. I always felt - and still feel - that with his emphasis on 'quality' and commercialization, Wiley's approach to OER was directed toward making the rich richer, even while it was cloaked in a facade of 'giving' learning resources to those less fortunate.

Indeed - and I've made this observation before - much of the existing structure of educational institutions is dedicated to preserving the fortunes of those already fortunate. It is indeed ironic that those who work for such institutions, and those who supported the academic publishing industry as editors and authors, are now finding their own work aggregated and exploited. What we are seeing with the rise of artificial intelligence isn't a new pattern. It is just the continuation of an existing pattern. 

And this is why it does no good to argue that 'AI must be regulated' or that 'AI must be blocked'. Whether or not AI succeeds as a technology is moot; neither blocking AI nor regulating the industry will alter the model of aggregation and exploitation that it exemplifies. The knowledge, learning and information industries will continue to exist, and with or without AI will continue to harvest community language(s), system of laws, culture and traditions, social knowledge, values and beliefs, etc., and in some fashion reshape them according to their own values and sell them back to the community.

Perhaps the elite status of the academic community would be saved, at least for now. But this does nothing to help the rest of us living in different communities around the world.

And this brings us back to what, to my mind, is the real purpose of open educational resources. They represent a means, mostly (though not exclusively) through digital technology, for a community to communicate with itself, to gather and share knowledge, to pass along its values and mores, its ideas and beliefs, and to be able to do this without reliance on external knowledge, information and learning providers.

We saw this in social media, before it was privatized and converted into a toxic mess. We saw that a community could create its own form of public journalism, reporting on and sometimes forming mass movements, breaking news, and (closer to my own sphere) educating and informing itself. Open Educational Resources - and then, later, Open Practices, became a form of learning empowerment. 

It - to me, at least - was never about giving people an education (or giving them rights, or freedoms or anything else). It was about people being able to create these things for itself.

Now a lot of my theoretical work over these years have been focused on the question of how this happens. My answer - and you don't have to agree with me on this to see the point - is that knowledge is produced through connection and interaction. A community creates its own wealth - not only financial but also social, cultural and intellectual - through the process of aggregating, remixing, repurposing and feeding forward. It's a system that works so long as every part of society implicated in the process benefits from the process.

This is, of course, a network theory of wealth - but, importantly, not a commons based theory of wealth. It is based, not on collaboration toward a single goal or entity or nation state, but on cooperation, each with the other. It's an ethos based not on extraction, but on sharing. And it is also a network theory of knowledge, culture, language - whatever you want, whatever is produced by a society, whatever is produced by a network of individual, autonomous and self-governing individuals working together.

To my mind, AI is just the latest version of such an entity. AI is based on the same principles that enable a society to create all these goods, and for an individual human mind to create them as well. This production is in all cases the product of learning, and learning is in all cases the product of practice, experience and reflection. 

Indeed, it is not a question of whether AI will 'continue to exist' or not. At its core, AI is nothing more than a set of mathematical principles that, when embedded in a physical system, can produce actionable knowledge. These mathematical principles aren't even all that complicated - it's not complexity that creates AI, it's scale (and, to a lesser degree, efficiency, so we can reduce the scale). These principles will continue to exist independently of a few companies, business models and economies. They will exist in individual humans - who will continue to be the means of production - and they will exist in societies. And probably, they will continue to exist in computers.

So now we turn - at last - to the subject of free artificial intelligence. For it is my belief that we are facing a question not of whether there will be artificial intelligence, but rather, who will own it. Will it become an instrument of colonization, or of liberation?

Over the last few months the Open Source Initiative (OSI), which created the 'open definition', has taken it upon itself to define Open Source AI (and the existence of this definition is one reason why I opted to use the title 'free AI' and not 'open AI'). It resembles in most respects a definition of free software, but with one major exception: limiting itself to free and open "information about the data used to train the system", but not the data itself.

The very content that is harvested and aggregated from society at large, whether via a common crawl or some other mechanism, will be the part of the 'Open AI' system that is proprietary. This preserves a model in which only one part of the social ecosystem owns the product of that ecosystem. The danger here isn't that some company or another will be able to produce the 'best' AI using proprietary data. The danger is that, by ensuring that the data are proprietary, they will be able to prevent anyone else from creating AI.

After all, the mechanism described by OSI, that "a skilled person can recreate a substantially equivalent system using the same or similar data," is a theoretical mechanism, that depends on having 'the same or similar data', which cannot be obtained in practice, if the AI company owns it all. And the more opponents of 'open AI' use copyright restrictions to argue against the aggregation of content by AI, the more they entrench the lockdown of AI enabled by the OSI definition.

The worst case scenario - and yet the one made most likely by most objections to AI extant today - is that only a small number of companies (which we can all name) are able to create and use useful AI models, and that our economy become dependent on paying (ever increasing) royalties and other fees to the use of what was, originally, our own knowledge. And if that happens, learning will never will be free, not by any meaning or definition of the term.

The only thing that protects open knowledge, open source and open learning is open data. 

Sure - we could just say we don't get to have open learning and the rest because the data used to train AI is de facto not open - it is owned, and the copyright held by the many people who have authored papers, articles, and Twitter posts makes it so. It would be unfortunate, because we could never then have a free society of the sort I envisioned at the top of this paper, but such a society just is incompatible with private property.

But I don't need to argue for the extermination of private property, and indeed, would find it difficult to imagine people enjoying life to the greatest extent possible without it. And the art or artifact genuinely produced through the fruits of one's own labour is, quite rightfully, theirs. 

But this does not entail stealing from the commons. There are two kinds of art and artifice in the world: that produced by individuals, and that produced by the community as a whole. And true, there is a whole lot of overlap between them. But I think we can make this distinction quite easily using language as an example.

There is, as Wittgenstein famously said, no such thing as a private language. This is not because it is impossible for a single individual to form their own words or grammar, but because the concept of a language precludes it being anything other than a tool for communication between at least two people. Language is produced by the community as a whole; nobody owns it, nobody owns the grammar, and nobody (with some exceptions) owns the words.

In broader society, language is much more than words and sentences; it is the full range of shapes, colours, sounds, memes, fads and fashions, gestures, whatever, that we use to communicate with each other, and within that communication, to express a body of knowledge about ourselves and the world.

This (in part) is what an artificial intelligence is harvesting when it aggregates data. It is not aggregating your 'content', it is aggregating the part you borrowed from the rest of society in order to get your idea across. The actual 'content' is very likely to be left behind, unless it stands out as a meme or widely used expression. Your claim to own that part of your content makes it legitimate for an AI company to say it owns that part of the content (as expressed in the (now) proprietary data it has collected. And this legitimizes the transfer of that part from the community to the company.

And we can't recreate it because it's the only community we have, and the knowledge that we could once share with each other in order to actually form a society is lost forever.

If we go back to it: the whole point of open educational resources should never have been to declare ownership, it should have been to enable sharing, through the recognition that this content, at least, is not for sale. This content, at least, belongs to the community, and its sole purpose and intent is to be used by the community, any community, to advance itself, to grow itself, and to express itself.