264 – Earthkeepers versus AI Empires (Part 2)

[00:00:07] David: Hi, and welcome to the IDEMS Podcast. I’m David Stern, one of the founding directors of IDEMS, and I’m here with Kate Fleming, another director, and I’m delighted to keep going on our previous discussion about this rather inspiring event, Earthkeepers Versus AI Empires.

[00:00:24] Kate: Yeah, I wanted to really dig into the specifics with you. So who was in the room? What was the agenda? What were the takeaways? The takeaways for attendees, but also for you, how did you come out of that? I think you came out very galvanised and inspired and feeling quite validated in a lot of things that we’re working on.

And so, yeah, to start, I guess maybe who was in the room, who convened this? We talked a little bit about it last episode.

[00:00:51] David: I guess the lead convener was a guy called Jim Thomas who has been involved in advocacy and in the food sovereignty space for a long time. He did some amazing work, for example, on teff in Ethiopia. He was, if you want, the main convener, and it’s maybe worth starting with some of his framing and what he brought to the table at the beginning of the meeting in terms of the framing.

This was prefaced by framing in terms of indigenous communities. There were a number of indigenous communities brought in from around the world, and their perspective was brought out first. And this was important because this was the earthkeeper part. These are communities who often have a history of earthkeeping in different ways across generations, which are to be valued and recognised.

So they laid that perspective, and that included, as we discussed in the previous episode, how some of those perspectives are being challenged and put in danger by the data centres. And then Jim did the framing, which I suppose really introduced the AI empire, which, as we discussed, Karen’s book explained and presented, and just how large this apparatus has become, highlighting that the wealth accumulated by these companies, there’s only one time in recorded history when companies had accumulated more wealth, and this was related to the slave trade and colonialism.

And that parallel in terms of percentage of the world’s wealth was an important part of the framing of this event. It’s a framing that I’m not used to, but it is something which is really striking when, actually, the numbers are put up of just what a large percentage of the world’s wealth is in their hands and how this compares to nations, nation states. And not only just nation states in terms of disposable income, but actual total income within large countries.

The wealth that these organisations have is of an order of magnitude which means that just as they can buy out other big tech companies, they could, in theory, buy out countries. That is the order of magnitude of their wealth, the total wealth of large countries.

This framing was an important initial framing. And actually one of the terms that came out from some of the participants, which I hadn’t heard before, was the “artificial wealth of the AI empire”. Because, unlike in other cases where that wealth is tangible and you can say what it is, in a lot of these cases it is perceived wealth, it is because people believe in the importance and the wealth that should be allocated to them, rather than very clearly what those assets are. This is why there is the discussion of it potentially being a bubble. And that is a danger for society as a whole. What happens if suddenly that wealth, that perceived wealth disappears?

So that framing was part of the initial framing and played out through the three days of the meeting. The meeting then had four tracks. They were associated with the four major rivers within Zambia in a way which was very visual. And those tracks I’m going to summarise them as 1) related to the mining, the extraction, and the issues around the minerals, 2) related to the data centres and the actual infrastructure which is getting built, 3) related to the, if you want, food sovereignty and the implications around what’s happening for agricultural biodiversity, for all these other elements of the food system, and 4) related to power in a way that I wasn’t really familiar with and this was framed in a lot of different ways.

I expected to join the food sovereignty track because that’s why I was there. I was there because of our work related to agroecology and food sovereignty. This is where I expected to be. And as the first morning was taking shape, I discussed it with Jim and he convinced me to go to the power track.

And I’m very grateful that he did, because this is where I was actually confronted with things that I really had never thought about as deeply, and with the broader question of where is this line between what is the technology versus what is, as Karen has framed, the Empire. And that power element of that is something which I’m just not used to thinking about. So I spent a lot of, well, all the sessions which were dedicated to the tracks in the power track.

[00:06:19] Kate: Can I interrupt you and say: I think you think about this all the time? I feel like every conversation we have is about it. I don’t think we frame it that way, I think mostly we’re thinking about communities, agency, recognising that agency enables you to deliver better impact.

And actually this is where I would say all business is political. When someone says, “oh, don’t get into politics”, businesses are highly political, it’s why they lobby, it’s why they advocate for things. They are advancing worldviews and politics, they are just centering them, and then everything else is positioned as the outlier point of view while you’re just against progress.

Whereas I think we’re very much in pursuit of progress. We just think about how progress happens very, very differently. And I think we think about progress as something that is.. and we actually talked about this yesterday when we were having a conversation about what does innovation look like when it happens at the speed of trust? And you were talking to a venture capitalist who was like, “nothing ever happens”. And you said, “that’s not true at all”.

And in fact, I think the argument we make is, well, when you build trust, things percolate more slowly at the beginning and then they start to cascade in ways where you have all kinds of growth because you’ve got all these networks of trust that can be used and activated. Versus what often happens when you’re trying to get to scale quickly and build proxy systems for trust is that when you reach the limits of that trust, you suddenly find the ways you can expand are quite narrow. You’ve built this very linear, very kind of single-minded track to scale and you have to keep beating at that, beating that dead horse, beating that drum. I don’t know which is positive, whatever it is.

So, anyway, I guess I just bring that in as I think we think about power all the time, we just don’t frame it in a way that I think an activist community would frame it. We are just thinking about it as “this is what is useful for impact”. This is actually how you start to really solve these problems in ways that are..

[00:08:20] David: No, you are spot on. And just as Jim helped to sort of direct me towards that, it was clear that I am thinking about areas of this, but never at the sort of political framing of the power of it. It’s something that I’ve not tried to engage in before. And actually it was really refreshing to do so, and to hear the other perspectives, and to actually be in opposition with some of those perspectives, and to recognise that that tension is not bad.

There was a lot that came out of this where my view was challenged and other people’s views were challenged by my views. And that was healthy, that we have certain ideas around how communities can be put forward by technologies, and to be actually in the room with communities who could benefit from this, and to hear about their experiences of actually people saying they’re going to do similar things but not serving them.

That was really healthy for me, to be grounded in the realities of what it really looks like. And I think it’s very interesting, I’ve never really worked or engaged with the indigenous American communities where elements such as wealth has become, in some cases, granted to communities in ways that have still destroyed the communities.

And some of these, the histories around that, which I am not totally ignorant to, but certainly not knowledgeable about, this is something where, to listen and hear those perspectives and how they’re thinking about what technologies mean and what it would mean to have things which are really indigenous and their own, it was insightful.

[00:10:14] Kate: That’s an interesting example, I think, as the American in the room, the Native American experience of having all of your wealth taken away, which maybe was a different conception of wealth, it was land, it was culture, it was all of these different things.

And then a lot of the way that communities, indigenous communities have regained wealth, I think in the US is often through casinos or things like that, have just been totally destructive. So it’s this one way that you see you have value – and I’m sure there are other examples too, this is just the one that comes to mind – but it’s not elevating the community, it’s not really tapping into the social, you know, the pro-social networks behaviours, the long history and tradition, all those things that create meaning, a lot of which people were deliberately cut off from. I think that’s a global thing that happened where you break historic memory to reset and channel people into something different.

So yeah, I don’t think you have to have deep American knowledge necessarily to understand that that is something that happened everywhere that was colonised. You see that across different African contexts, you see that in any context where the organised history and culture and traditions and systems of power are an impediment to doing what you want to do or to doing it at the pace you want to do it at.

That’s mainly it. You could have done it, but you would’ve had to do it collaboratively, you would’ve had to share the wealth, you would’ve had to build very different models. You know, systems, governance, everything would look different if you were doing it collaboratively versus, well, we’re just coming in and we’re gonna do this exactly the way we wanna do it and you’re just an impediment. You’re either with us or you’re in the way and then we’ll just bulldoze over you.

[00:12:05] David: Well, and what’s so interesting, of course, is the parallel there and exactly how concrete that parallel is with what’s happening with the AI empires and the narrative of “you are either with us or against us, this is inevitable”. And that is a narrative and a story rather than a fact. That is not true. There are multiple ways in which we can move forward from where we are, which we know from a technical perspective. And we’ve been fighting that, if you want, on our own, very isolated.

This is what I was so astounded by, that, actually, before this meeting I felt very isolated in what we’re trying to do. There was nobody else there trying to do what we’re trying to do, but I don’t feel isolated anymore because I feel seen, I feel heard. And I feel that actually I hear you as well, you’ve got a different track you are going down, one which is not the track that I know or necessarily will follow myself. But the concept of allies working together to actually say, well, we are strong together because there are a lot of people who are seeing different pieces of this and what it actually means, and it is not inevitable to do it this way.

[00:13:22] Kate: One of the things that Karen’s book really brings out – and I think we have always thought this way, but she brings it to focus in the current model of AI – is that there is this narrative that AI is just this output, it’s a product, it’s what tech loves to do. You just create this consumer friendly facing product and you hide the system underlying it, which is just treated as business. That’s not treated as choices and all these different variables.

And so I think one thing that she does very well is that she breaks out those variables about governance, about energy, about resources, about how data is collected, about how data is cleaned and made sense of, and who’s doing that work, often for very little pay.

These are systems. And we often talk about a system and we just sound chaotic. It’s like, what are you trying to do? Why is this a system? You don’t question when ChatGPT or OpenAI as a company has this system behind it. It’s just, you see that as like, oh, well, that’s the subcontractor who’s doing this work. You have this totally different narrative for that system.

And so I think what you stepped into, it sounds like – and I wanna talk a bit more about who was in the room – it sounds like you stepped into a space where you got to experience it directly. We do this all the time, but often through partners, but you got to directly sit with people like, oh, you’re interested in governance, you’re interested in labour protections, you are interested in protecting the environment. And so in this single room about power, you actually had the convening and representatives of the system and really hearing someone who’s an expert with their own thinking and systems of accountability and meaning making or whatever that is, as opposed to this is convened by OpenAI and then everyone is just “this is what you will be doing”.

[00:15:10] David: Well, exactly. And this is the thing, what was so interesting is that everybody, just as we’re struggling to tell our story and to actually say what’s going on, everybody is struggling to do this. But what was so powerful about being in that room – and I was definitely an outlier, there were no other mathematicians there – was that I was welcomed and I was accepted as being different from how others thought, but everybody was different. Because as you say, different people have their different perspectives.

And what was so interesting was that it really brought into perspective just how big the resistance to “this is inevitable” is. We’ve had a narrative for quite a long time, that we think we can do things differently from a technological perspective, in terms of how we build with community ownership and all of these things.

And yeah, there was no one else exactly on that path, but there were so many other people on the same path in terms of actually saying it is not inevitable that this is the only way that we, as a society, need to evolve, need to develop, progress. We are all in favour of a form of progress, but “what” progress does matter. And who holds that power. And with that power, the society that emerges, this is where we want real intentionality.

There was a wonderful exercise of actually looking back and looking forward, drawing out insights from the room about what does the past look like, what does the present look like, what does the future look like? And imagining what a future could be was really inspiring to hear. What if tech was sort of like tending a garden? You know, it actually requires human effort to build and maintain the technologies that serve you and that you have some control over in that way. It was a vision, which I’d never heard before, but I really quite liked that idea that, actually, tending a garden, you are controlling nature, you are sort of influencing, but you are doing so in a way which is balancing.

What if we thought of technology, and digital technologies in particular, in a similar way where there were real community choices and individual choices in how that garden flourished and what that looked like, and it took human effort.

These were things in which, I think, there were many different visions of the future in different ways that emerged, but it was really inspiring to have such an incredibly diverse community looking backwards, looking forwards. One of the things that was really inspiring, and it’s the reason for looking backwards, is to say that this isn’t the first time there has been real forces at play capturing wealth, which is a view of what is happening.

And in the past when these have happened , this is how we emerged from that. And history then shows how that is not inevitable, and that existed and then it was not sustained, a new future emerged.

[00:18:17] Kate: Well, and I think part of what was energising for you – and I find this energising for myself too, and we had a bit of a conversation about this – is that activist spaces often find themselves in the position of being reactive, because they are often community representatives, they are people who are interested in policy, they are not the builders of things. So the most you can hope for is that you’re shaping regulation, you’re shaping the direction of travel.

But it’s very hard to conceive how you actually build something different and you can actually kind of step off, or the long arc of trying to curve what exists as AI in the direction that will be good. That can feel very nearly impossible. And I think you found yourself in the quite rewarding position of being able to say, oh, well, you just don’t have to build it that way, you could build it quite differently. These are built for this goal – and we can talk about this in another episode, I think, about what these models of AI are.

But I think, one of the things that we have always been aware of, and it was our point of commonality that brought us together, is that awareness of data and that existing systems are not built to gather the right data. And I use data very broadly there not as analytics, but information and meaning making and all these different things that are how you create meaning as a human, and societies create meaning, and all of those different things.

And I think we were both so aware of how you can’t possibly address these hard problems unless you’re designing systems that people trust, that they have control over, where they’re going to share things that are really high quality, valuable information and data. You literally cannot solve problems or begin to move them in different directions without accessing that information.

And so that is fundamentally a data problem that we have been thinking about for a long time. And now we’ve kind of arrived to this point where it’s like, whoa, maybe this is our moment to shine. Because a lot of the problems around AI are data problems. We’re like, “well, have I got some ideas for you?” And so I think that also for you was very rewarding to be in this room.

And I think Karen, I know we keep referencing Karen as if she’s our best friend, but…

[00:20:28] David: She joined us remotely, so I didn’t even get to meet her.

[00:20:32] Kate: But I think that she highlights a lot of these things, she gives us the lob that we get to smash now where it’s like, yes, that is a problem. And we have thoughts on how you address that and this doesn’t all have to look that way. So I guess, I don’t know, I don’t wanna get into it too deeply here, I think we should save it for another episode, but just as a snapshot, what did you feel that you were able to contribute out of that kind of historic and current work we’re doing and all that perspective that is the IDEMS kind of direction of travel?

[00:21:03] David: I mean, it is this key point, which is if you can separate AI as a tool from AI as the infrastructure which is getting built or as the narrative of actually what it’s doing, then almost all the problems people care about and want to solve are solvable using tools that already exist, with data which is not that big, and with computation which is really achievable.

And so the advances that have been made are enough to be able to be transformational. And the transformation that people care about and that could be really valuable of using these AI tools to do so, well, it’s just a question of actually building that out and actually making it happen, and building the structures to do that in ways that respect governance and ownership and safety of the data in different ways, and actually have the right data going in.

The whole point is, the expression that’s often used is “garbage in, garbage out”. If you are building with all the data in the world, you have no control over what goes in. Whereas if you actually have a community and you care about and you are able to control what goes in as information and how that’s used, then that will almost certainly serve you as a community.

This is sort of globally, it’s not about the general AI dream. That is something where there’s possible future visions of other advances that could happen in the future. But the point is they’re not needed for the things that could really help communities, help society right now.

[00:22:41] Kate: It is such an interesting thing because the rallying ambition that put so much energy into AI was this AGI, artificial general intelligence, this future state where it exceeds human intelligence, whatever that means, and we just solve all these problems. 

But the reality is that it has become like a false god and a diminishing return under current models. And if anything, it should be the thing that’s a nice research lane, it should be that lane because that is far less practical, far less achievable. The scope and scale, is it even what we need? There’s so many questions that people should be asking about that.

If you actually just take your head out of the grandiose talk and all of this kind of “hype”.

[00:23:29] David: Yeah.

[00:23:30] Kate: There are so many really practical, very efficient, small data sets, where there could be people who actually understand them and are experts of. But perhaps the biggest issue is that they are not conducive to monopoly power, they are not conducive to aggregation of resources. There are all of these other forces that are continuing to drive that push to AGI that aren’t necessarily relevant to the day-to-day problems that most of us are seeking to solve.

[00:24:05] David: That’s so spot on because it is about this element that the actual problems that we could solve, there’s just so many, we now have tools to solve them in ways that even just a few years ago was not possible. And the research to do this well should be on how you do it with less data, with less compute. That’s where research would be really useful right now, because it would now say, well, you can solve that problem actually cost effectively.

[00:24:39] Kate: Well, I would argue research and also development. I don’t even think, based on what we’re seeing, things have to be in research that long. And even at the end of Empire of AI, she gives an example of preserving a language. And the actual data that’s needed and the work, you need AI that we have now, this generalist learning AI, to get you started.

And we are going to do another episode because I think this is really important. We’re gonna do a few more episodes, but one is going to be about that transition from large language model AI to these more specialist, where people can control their data, all of these issues. But I think we see that these are not pipe dreams, these are really achievable, they just need some funding and you need some set of people who are expert in how to do this kind of things, so it’s not everyone’s trying to figure it out on their own.

[00:25:24] David: Yeah, watch this space, let’s come back with a few more episodes around this topic because it really is energising to see that we are not alone. There is a community, there’s people who are thinking about different parts of it. We have a contribution to make, and I think it could be a really important contribution because it is something where actually I would have hoped that the contribution we want to make should have been made by academia.

But as is, again, illustrated in the book, a lot of the financing of academia and the way that’s happened has led to the fact that actually the academic effort has been derailed. I’ve been talking about this for years in different ways, and it’s Karen who actually explained it better than I’ve ever done, how that academic effort, which was really diverse and looking at all these different approaches, has been narrowed down into actually a small set of problems, which are not the right problems that we know need to be tackled.

[00:26:31] Kate: Thank you, David. It’s very interesting. I’m sorry I wasn’t there, it sounds like a great event, super interesting.

[00:26:38] David: Well, there may well be a follow up and I’m sure that I would put you forward as well. You would add a lot of value and you’d be able to sort of articulate certain things that I’m not particularly gifted at articulating.

[00:26:49] Kate: Maybe we’ll figure it out on podcast episodes.

[00:26:51] David: Maybe we will.

[00:26:52] Kate: Thanks, David.

[00:26:54] David: Thanks, all the best.