This is a brain dump. I am thinking out loud here, so apologies in advance if this is unfocused and lengthy. If you want something more polished, wait a bit and check back some time later. I may continue evolving this page toward something more concrete and coherent.
I am writing this because a toot of mine got some attention, which made me think it'd be a good idea to expand on my thinking in case there proves to be something of value in it for someone. At the very least, it seems like an important first step toward finding consensus around solving, or at least mitigating, a/some problem(s) I'm not alone in experiencing.
First, the problem of actual human relations on the Internet. We all share a material world, and that material world is primary; communication networks like the Internet are useful inasmuch as they provide a way to create a common and shared experience in line with our actual experiences and preferences as rooted in that material world, because we're all using said networks to work how best to engage with said world. However many layers of abstraction we build up, however esoteric our memes, the physical world is and always will be that which underlies it all, and to which all our thoughts and ideas flow as actions.
Bots and sock puppets are influence multipliers which can be, and which are, used to create false consensus and exhaust and drive out people trying to engage in good-faith discussion. The advent of LLMs and generative AI severely exacerbates this problem, as it makes producing that sort of bad-faith engagement a lot cheaper. Having a sense of how likely it is that the person on the other end of the line is a real human seems valuable to me for this reason. It is also, of course, valuable because authenticity matters; if you are engaging authentically with someone who is engaging inauthentically, it precludes the formation of actual communal bonds. Reciprocity is essential to relationship, and reciprocity with LLMs is impossible.
(As I review these two paragraphs, I realize they sound a lot like the bullshit Facebook was spouting as justification for requiring "real names" on profiles; please don't think I am advocating for using real names online. I am actually very opposed to this for safety reasons, which of course in turn has implications for authenticity. A consistent pseudonym with a human, and only a human, behind it will be more authentic and honest than a name which anyone, of any trustworthiness or intent, could possibly track back to your physical person.)
The foundational fix is to revive physical community. I catch glimpses of this idea in the writing of Wendell Berry, whose writing often argues against the mindless, the homogenized, against "art oriented exclusively to production, like coal mining." (The last point of which resonates especially strongly as more artists come to realize the hazards of this mentality in the context of generative AI.) In Writer and Region, Berry writes, while quoting Emerson:
And action, of course, implies place and community. There can be disembodied thought, but not disembodied action. Action -- embodied thought -- requires local and communal reference. To act, in short, is to live. Living 'is a total act. Thinking is a partial act.' And one does not live alone. Living is a communal act, whether or not its communality is acknowledged.
You live in a physical location, among others in the same physical location, and your problems and physical fates will be tied together by this physicality. You know these people are unique humans; astroturfing and sock puppetry are much harder in the physical world, among the people you've been living with and encountering day in and day out, than they are in an abstract world of strangers and partial embodiment such as the Internet.
If there are to be actual human relations on the Internet, they are best rooted in actual human relations in the physical world. By transitivity of trust, we may possibly knit these together into a landscape more hostile to anti-human forces than what we have now. It will be a smaller Internet, inhabited by ghosts outside of our trust networks whose existence may not be assumed, names best unmentioned, at least within those networks, because of the dubiousness of their existence.
(To measure and categorize trust is a hard problem which smarter people than I have written many papers on, none of which I have read in recent years. I don't propose anything here beyond a bit which says "this person exists and I know of no other names under which they are operating in this space." I don't imagine this is sufficient, but this is at least the beginning of an idea.)
The second problem I perceive is the lack of enforceable boundaries. This is an even harder problem, I think, than the first one is, and cuts across many layers of abstraction. Fundamentally, I think it's a problem of sovereignty; of not owning and controlling the means of production oneself, and likewise not owning and controlling the fruits of one's own labor.
Tante's talk on what open source can learn from Luddism touches on this second problem a bit. The Luddites had been weavers who largely worked out of their own homes, who were paid for filling orders and otherwise had complete autonomy. All the efficiency gains they were able to realize through their own cleverness accrued to their own benefit, to use as they saw fit. With the advent of the autoloom, they found themselves forced into factories, where they had to work according to someone else's schedule and under someone else's direction. They were made to produce fabric of an inferior quality in order to produce a greater volume of it for the benefit of the factory owner. All efficiency gains, accordingly, accrued to the factory owner, no longer to themselves. They lost control of the means of production, and they lost control of the fruits of their labor.
Similarly, we're seeing our own work being used to fuel machines which produce poor approximations of said work, against our wishes and, indeed, against our legal rights. Copyrights and licenses are being ignored just as surely as our robots.txt files. Inasmuch as our governments fail to hold all to the same standards of justice, our rights are unenforceable against those with greater privilege; ergo, as individuals, licenses and copyrights amount to requests rather than any sort of actual protection.
Banding together to enforce license and copyright compliance is the accepted solution to this issue, of course, and we're seeing that in action right now. There are, according to my brief search just now, 20 or so lawsuits pending against OpenAI. It will be interesting to see how they turn out. I will be pleasantly surprised if the courts choose to kill OpenAI over its failure to follow the same rules the rest of us are bound to. I'm not holding my breath.
A more concrete, though also more costly and difficult, form of protection is simple exclusion. "Information wants to be free," of course; that is, you cannot enclose information indefinitely. It will leak and propagate. But the boundaries one puts up around it can change the rate at which it propagates, which is practically just as good as keeping it completely secret in many cases. Likewise, having concrete boundaries in place may make enforcing personal liability, as a deterrent, an option. A lawsuit between two organizations, one representing capital and one representing workers, is unlikely to end in much more than a fine against the organization representing capital. A lawsuit between an organization and an individual would have some teeth to it.
Having a set of "known good actors," or even simply "known unique humans," could go a long way toward addressing both of these problems. I do not think we can, or even should attempt to, create a universal and objective set. Smaller, subjective sets could, in theory, compose with others and contribute to subjective "confidence" scores regarding trustworthiness and unique humanity.
(Actually, this brings to mind Circles UBI, which had a problem posed in similar "set of unique humans" terms and attempted to solve it by making "Sybil attacks" effectively pointless. I remember not being completely convinced for some reason. I'll have to check in on that experiment again and see how their design has worked out for them and whether there's anything that might be borrowed here. I wonder if they worked out a way for people without smartphones to participate.)
The second problem points strongly toward the need for a political solution, and therefore a material solution. We are, after all, in this political situation because of the material realities of capitalism. (This is unfortunately where it gets murkiest for me.) I realize that having to provide services to the capitalists who gain from our further disempowerment, our further alienation from our labor, fuels exactly those problems. It would seem to me to be ideal if we could provide for our own material needs without having to resort to collaboration with those who have designs on pressing us into ever-greater servitude. This poses a bit of a chicken-and-egg problem, I realize. Thinkers greater than me have spent a lot of time and ink on this question, and I have not read most of them. But I presently am inclined toward a sort of cooperativism, a federation aimed at extending the democratic socialist economy down toward the soil and then up the "tech tree" from there. Some sort of enclosed commons to provide for its members in order to free them from the capitalist hegemony and enable the free and open development, exclusive of capitalists and their coercion, of our science, technology, and art.
The root of the problem with regard to LLMs and generative AI, again, is that we do not in fact control what is done with our work. Such control cannot possibly be guaranteed if we do not also control the means by which we do our work, by which we sustain ourselves. In order to address this problem at the root, we need to address capitalism. We need to develop for ourselves an alternative, parallel economy in which capitalist rentiership is disallowed, and in which production is not coerced (and thereby warped) through artificial scarcity.
This is a brain dump, and maybe I've said too much. But I see all these pieces as inextricably interconnect. Pull on one and you begin pulling on the others. They move with each other, and so they must all be moved together, I think. There is no clean separation of concerns in reality.
I'll continue add to and refining this page as my thoughts develop, as I do with every other page in this "garden." My actual work will probably only touch on the barest edge of everything I've identified here. I'm primarily a technologist, after all, and technology is hardly the primary problem here.
screwlisp suggests Veilid and Kitten could be components of a solution. I've been meaning to check out Kitten's technical details, with a view toward possibly writing an alternative implementation for fun. I've also had my eye on Spritely Goblins as another potential building block.
I am adding to this brain dump because I have been reading an article in Compact about Urbit while also partcipating in an exchange on Mastodon about ideas tangential to the ones on this page. I am not trying to work those things into a more unified whole with the rest of this page because, at the moment, I am taking time away from visiting family in order to jot down enough of my thoughts to recover, expand, and incorporate them later.
Capital always warps what it comes into contact with. Large, universal solutions and entities like the ones we've been trained to expect by Silicon Valley, are only possible at present with the support of capital. Even large entities, absent capital in the traditional sense, engender problematic power differentials. Fundamentally what people seem to want does not seem to be achievable without centralization and the corresponding power differentials. Yes, if computing power and computer science advances, we could one day make a decentralized Twitter that's easy enough to self-host and that doesn't have the synchronization problems of ActivityPub, but by then it won't be Twitter that folks want. Centralization will always have distinct advantages in terms of ease-of-use and efficiency. Efficiency is centralization's whole bag. The problem is that what is optimized for is generally not in the best interests of the users, but in theory that "setting" could be tweaked enough in the face of a sufficiently threatening challenger to keep folks not turned off by the fundamental structure on the farm.
The unsatisfying answer may be that the very concept of a whiz-bang universal solution may itself be incompatible with any notions of a more humanized future. It may be that real human networks, with real human beings whose material and emotional/psychological interests are necessarily small and heterogenous, and that they take time to grow, and that they don't generally connect across great distances but through sequences of people who happen to straddle multiple networks. The whole notion of a "living room for the world" may be poisoned at its root. What we are doing now, with building tildes and getting off corporate social media and reviving older forms of connection, which will necessarily be, on average, more authentic because they are less convenient and harder-won, may be what the more humanized future looks like. It may be what proper decentralization looks like. Maybe we need to stop trying to pretend we aren't actually a million different networks with a million different needs. Maybe the difficulty itself, in making connection and in joining these things, is part of what keeps them viable, keeps them honest. When things are free and wide open, they become dominated by the well-resourced and by those who have something to sell. Only things with some kind of cost, whose membership requires some kind of surety, (which may well be measured only in time spent with others rather than anything else) can hope to avoid that kind of domination, I think.
I think Wendell Berry is right about this aspect of things, about the need to see the landscape one occupies, and to appreciate its history and its particularities, and to endeavor to become a part of it, rather than to foolishly try to separate oneself from it and dominate it from afar. (I mean to invoke his image, here, of someone "high up in the air-conditioned cab of a tractor described as 'a beast that eats acres'" as contrasted against someone "who has worked close to the ground in the open air of the field," from Horse-Drawn Tools and the Doctrine of Labor-Saving, an essay so in tune thematically, though different enough, with Getting Along with Nature that I initially searched the latter for the exact phrasing.)
I think I've relieved enough of the pressure of ideas now to step away from the keyboard and return to the company of others. I hope I can synthesize all this into something more coherent soon.
One year later...
Still stuck on "what if we touched Wireguard LANs?" Not getting any wiser here, just more tired.
Need to do a review of this one of these days when I've actually gotten enough sleep.