- 1 week ago
In Episode 14 of Overlap, we discussed—from the perspective of the Global South—the AI and the Global South and its impact in the world. teleSUR
Category
🗞
NewsTranscript
00:04Hello and welcome to one more episode of Overlap, these conversations from across the world.
00:11Tonight we have the great pleasure of being joined by Jeff Jiang. He is the Secretary General of the
00:19Global South Academic Forum and also AI expert. And we're here to have an important conversation,
00:25such a pressing conversation about AI at this moment. Thank you, Jeff, for joining us here in
00:31Overlap. You're welcome. Happy to have this opportunity. Of course. And you know, we use
00:36this time in order to get a little bit deeper into some of the most pressing issues and bringing
00:42this perspective from different sides of the world. And of course, China is such an important
00:47partner in this moment in order to understand AI development. And I just wanted to start this
00:54conversation by thinking about how polarized the discussion about AI seems to be right now. For
01:02some, it seems that AI is going to solve all productivity issues and will bring all solutions
01:09we haven't found yet. And for others, it's the great enemy, the great danger that is putting really
01:15humanity at risk. On that spectrum, where should we put AI right now, particularly from the
01:24Global South? How would you start that conversation? I think I can understand the concerns, the
01:30worries about AI. Of course, first of all, we all know the potential. It can accelerate the speed of
01:40science and social science research. It can help people do many things. Robots can perform difficult
01:49physical work for people. But I totally understand people also have the worries, especially if you
01:57look at what AI, particularly in the United States, what AI being used in last year. It's been used for
02:08wars. It's been used to monitor and surveil people. And companies are giving their information about
02:18individuals and organizations to intelligence organizations like the CIA. And they don't even
02:26hide that. And the building, the expansion of data centers are leading to environmental and ecosystem
02:36worries. And of course, there is also a worry to employment. So I think it's people have good reason to
02:47be
02:47worried about AI. But I would say that is more a political issue than a technical issue. It's about how
02:57the
02:58technology being used and how it's being regulated. And that is a political decision. Well, you can say in
03:06some cases, in some countries, there is a lack of political guidance. There are companies making their
03:12decisions without political regulations without political regulation. And that per se is also a
03:18political decision. So I think what we're seeing, the increasing worry of AI development, showing us the
03:29lack of a good political guidance in some parts of the world, especially the United States. So then the question
03:40to the
03:40global south becomes, is that the only approach, the only solution we could ever have? Based on that point of
03:50view, we
03:51would say that's why the experience of China, the model of China becomes so important because it shows there is
04:00not
04:00only single approach for AI development. There are alternatives, at least there is another alternative, which is a socialist approach
04:12developing AI.
04:13So, Jeff, I definitely wanted to get a little bit into that. So what you were saying about this being
04:20not a technical
04:22conversation, or at least not only a technical conversation, but also political. It's so interesting because we usually
04:30see this discussion being held worldwide about AI development, AI risks. And we're talking about AI as if it were
04:38a single thing, a single, like, almost
04:41neutral technology that can just be developed or halted. And we're not talking about the different models, the different
04:49ways in which that can happen. And I assume from what you were saying that you are putting the perspective
04:57on
04:57that position, right, that there could be different ways. And of course, we're talking about the United States and China
05:03China, because they're not the only partners, but they are the major partners in this AI race that is happening
05:10right now. So how
05:11would you describe that? If what is the, if there are alternative models, what are the main differences in terms
05:19of the approach to AI, the way of thinking about uses for AI, the production itself, the, what are the
05:29main drivers and differences
05:31between how the US is tackling this issue and how China is thinking about this issue?
05:37In the last 18 months, we see some important development of AI, both in China and in the US. I'm
05:45more familiar with the
05:46China side, and I think that's, that could be a more important topic for the global south. So I will
05:52explain more from the
05:53China's perspective. The first important development, of course, was last year, January, when DeepSeq had its R1
06:04release. It was a powerful model. And since then, the Chinese companies like DeepSeq and GLM and Kimi, they're
06:15developing one generation after another generation of powerful models. Just two days ago, or today, DeepSeq released its
06:25latest version. It's another very powerful and cheap, cost effective model. But what really mattered was DeepSeq, since R1, was
06:38a
06:39open sourced large language model, meaning it opened all the parameters in the model so that, in theory, anyone in
06:49the
06:49world can just download the model and host it by themselves and they can use it. And in the last
06:5618 months, we are
06:57observing the Chinese companies, we're seeing a pattern that, together, those companies are creating an open ecosystem of
07:08technology. Instead of trying to monopolize the market and, and the rain fence users in their own closed software and
07:19and AI ecosystem, they try to create an open ecosystem, in which large language models can be alternative to each
07:27other. Users can make their own decisions to which model they use. They can build their applications for themselves. So
07:34I think
07:34that openness is the first and very essential difference. And the second one is the AI plus strategy. So it
07:45was
07:45clarified last year in the 11th document of the country. It encourages the usage of AI in all the other
07:55sectors, not
07:57not taking the development of technology as its own purpose. Instead, China tries to encourage using
08:06technology to enhance the productivity and effectiveness of every other sectors, being in
08:14manufacturing, agriculture, education, healthcare, governance, everything. So what we are seeing is an open ecosystem
08:24reduces the cost of every customer, every sector to adopt the technology, and then it encourages the
08:34innovation not only in AI industry or ICT industry, it encourages innovation from every industry and creates a wide
08:45wide adoption of the technology in the whole society. And I think the third difference is that the
08:55regulation of the, from the government to AI is pretty strict. Let's use that word. There is a regulation for
09:09more generative content. There is a regulation for, let me see, privacy and data security. If we make a
09:19comparison, right, open AI and Anthropic each had a open letter published on their website saying they saw some
09:29users from some governments using their AI for some purposes for some topics. They seem to be quite proud of
09:39reading those individual conversations and expose the usage. As an ICT expert in the industry who worked for
09:4720 years, I found that quite astonishing to me because in the industry, in the ICT industry in China, large
09:57-scale
09:57user data and privacy data is heavily protected. It's just not imaginable that the company runs the
10:07service also just goes into the database and reads individual conversations. It's something I think that from that point
10:16of view, the level of regulation of the United States has a lot to be improved.
10:23Exactly. So you were touching on some very important topics. And first of all, because we're trying to bring this
10:30argument in order to make for a better well thought discussion of these topics online for a general audience and
10:41a general public,
10:42who are a lot very concerned about everything that has to do with the AI right now. When you're talking
10:48about the open source model,
10:51how would you explain that for someone who has no idea in terms of what goes into in terms of
11:01the production and the programming side of the AI?
11:07How would you describe what an open source model is that you are linking to a Chinese way of developing
11:14AI? And what is it pulled against? What would be the alternative?
11:19Interesting. Thank you for giving me the interesting challenge.
11:26So in many ways, AI we're talking about today are very similar to human's brain, right? Human's brain have...
11:36Our brain is basically a neural network with hundreds of billions of neurons connecting to each other.
11:45And that connection makes intelligence possible, right? So AI we're talking about today in many ways are very similar to
11:56that structure.
11:57So there are hundreds of billions, sometimes already trillions of neurons connecting to each other.
12:06They are not biological neurons. They are parameters. So they are parameters in computer, in data, in computer science terms.
12:16And you can imagine when a company first developed such an advanced AI, such an advanced model,
12:28it can choose to close that model, to hide that information from anyone else.
12:37They hide the parameters so that they can monopolize the intelligence. Then if they are the only company can do
12:46that,
12:47they can basically decide how much premium they want to, how much profit they want to earn from the monopolization,
12:54right?
12:54This is economic 101. And that is exactly what OpenAI and Google and Anthropic are doing.
13:03You don't see they have open source models. They hide the parameters. Only they themselves know how their brains are
13:12made.
13:14The Chinese companies, led by DeepSeek and followed by other companies, selected a different approach.
13:22They decided when they developed an advanced AI, an advanced artificial brain, they opened all the parameters.
13:31So it's a big file. It's going to be a few gigabytes to dozens of gigabytes size.
13:39But in theory, anyone in the world can just download the parameters and replicate the model on their own machines.
13:50It's not cheap machines, okay? It's still expensive machines, but they don't monopolize it anymore.
13:57Other companies like OpenRouter, for example, or Alibaba, they can just download those models and host them.
14:05And so they start to offer the same level of capacity to the world.
14:11Of course, you can imagine a country, for example, Venezuela government can decide,
14:18let's build a data center and host that model in the data center so we can offer,
14:24even though the country is being sanctioned, but we can still offer the capacity to our people.
14:31Or a university can do that as well. So they don't have to rely on any particular company to offer
14:41them the competition power to use the intelligence.
14:45And more importantly, they are not bonded anymore to the monopolization.
14:51The company cannot get extra high profit out of them because they have choices now.
14:59So that is, when I say an open ecosystem, that is what it means.
15:06In such an open ecosystem, you have a few different models at least.
15:12As of today, you have DeepSeq, you have GLM, you have Kimi and Q1.
15:18They are all pretty good. I'm not saying they are the best today.
15:22They are a few months after the best, but they are pretty good.
15:27And with that options, an organization can choose from those models and decide,
15:35okay, this one fits me better. And six months from now,
15:40if this model lagging behind, we can choose to another one.
15:45They are not monopolized by any particular vendor anymore.
15:49So in my concept, I call it the commoditization of larger language models.
16:01Basically, you make AI a commodity.
16:03AI is high technology, is advanced technology, but it doesn't have to be super expensive.
16:10It can be a commodity just like electricity.
16:13Electricity is also high technology, but it's not super expensive.
16:17The same to artificial intelligence.
16:20You were talking about the possibility of, for example, Global South,
16:25Latin American government of having the decision of taking this open source models
16:31and using them to foster their own capacity and their own challenges.
16:39What do you see in terms of the great opportunities that AI taking in this way could open,
16:49particularly for Global South countries like Latin American and Caribbean countries open right now,
16:56considering the great dependency that we have of general economic structures, the gap differences,
17:05and the technological gap that is always putting this sector, this Global South,
17:13just behind general global supply chains and everything.
17:18Is there a key there in AI that we could tap into?
17:22You know, in the last 20, 30 years, there has been always a structural insufficiency of supply of software to
17:38the Global South.
17:38In many cases, Global South users are not only relying on the Global North for infrastructures,
17:47but also relying on the Global North for ideas of what software should be built and what users should be
17:55served.
17:55We have to, in many cases, we have to follow the Global North's model to decide,
18:03OK, because there are existing software and there are software companies building those things,
18:09so we better adjust our operation model and management model so that we can benefit from using the software.
18:19It happened in the 1990s when ERP getting popularized.
18:26It happened in the 2000s, Web 2.0 being popularized,
18:30and then when people talk about Web 3.0 and decentralized internet,
18:39it's always the Global North had some ideas.
18:42Let's be more clear, it's almost always Wall Street and the Silicon Valley had some ideas of how information should
18:54be used and managed in the society,
18:57and then Global South follows those ideas and the software.
19:02Because we didn't have that many experts who can develop software for our people,
19:08and also because our people have less resources, less money to hire those highly paid expertise to build software for
19:18ourselves.
19:18So let's be more broad information tools, OK?
19:24Like Excel is an information tool, or a spreadsheet, a Word template is also an information tool.
19:33It's larger than just the software.
19:35But anyways, my point is there is always a structural insufficiency of information tools supplied to the Global South.
19:46Global South, especially the Global South people, are not well served for information tools.
19:53We have a lot of situation, we need information tools to help our people,
19:59but we don't have such tools because we don't have the resources.
20:03I think a great opportunity that AI brings is that now we have the opportunity, we have the possibility to
20:12build information tools for our own people.
20:15Let me give you an example.
20:17I recently worked with a TV station, Pan-Africanism TV, in Ghana.
20:24You probably know them. They are a small TV station, much smaller than you are.
20:29They have about 80 people in total. Many of them are just volunteers.
20:35And they don't have the capacity to have world-class commentary for emerging news.
20:45They had to rely on some experts to write comments for them.
20:50So thanks to the development of AI, after 10 days training, the editors of PA TV was able to build
21:01such a, we call it agentic systems.
21:04So it's an intelligent system composed with a few agents.
21:11So agents are independent sort of units can perform some particular task.
21:19For example, one agent is a deep researcher.
21:21It can go to the internet and collect any news, any sources about a news thread, things like that.
21:31So they created such a system. It has six or seven agents.
21:36And then those agents work on a larger language model.
21:42Now they are able to, with any emergent news happens, they are able to write a world-class commentary.
21:51Of course, still they need to interview some experts to get a few sentences comments,
21:56a few very short but sharp opinions.
22:00Then they can write a world-class commentary in a few hours.
22:04I think that level of capability is the biggest opportunities that AI brings us.
22:11Think about PA TV, right?
22:14Who would build a software for them?
22:17Nobody. Nobody builds software for them because they don't have money.
22:21They barely have money for their own salaries.
22:24So the software companies would never go to them and try to build a software commentary,
22:30a news production system for them.
22:32But now, with the help of AI, they can build such a system for themselves.
22:37I think that kind of story in the grassroots organizations, medias, think tanks, universities, social movements,
22:47they should be able to build a lot of information tools for themselves.
22:51They don't need to rely on anyone and nobody can cut their supply of information tools.
22:58I think that is an important opportunity.
22:59So what you're saying is that in these cases, used in the way that you're presenting it,
23:08AI could be useful in order to bridge certain gaps regarding investment, technical capacity,
23:18in order to do things that otherwise these grassroots organizations, projects could not be able to do
23:25because now they have this technology at their disposal.
23:28That is super interesting to keep in mind.
23:31Also, a little bit of optimism as well.
23:34We hear so many terrible things about what's coming.
23:37So it's good to have that in the horizon.
23:40Now, of course, everything that we have on our conversation, we could continue talking.
23:45I think we could do a podcast on every one of these items.
23:48But you touched on communication and the way this is being used.
23:52And there's a lot of other risks that we are seeing and direct impacts that we are seeing right now
24:00of AI in communication of maybe other uses that are having more negative impact.
24:08We have communication services using AI and so reproducing Western stereotypes that are just repeating what main Western media is
24:21saying
24:21because they are learning from those contents themselves.
24:27So that is making the battle for ideas from the Global South so much harder in that way.
24:35We are seeing AI also collaborate in the spread of misinformation, for example.
24:43Well, a lot of that has to do with political campaigns and how the far right has used it.
24:48So particularly in communication, because I know that you have this perspective of seeing everything that could be done.
24:56What do Global South communication outlets need to keep in mind in order to stay away from the main risks
25:05and also use AI to their advantage?
25:08This is a difficult one.
25:10Because the media sphere is so complicated and it's so weedy.
25:25It's not a technological issue anymore.
25:28It's a political issue that the social media platforms are not being regulated.
25:37When they are being regulated, they're regulated as like you cannot talk about supporting Gaza and you cannot support Cuba.
25:50And then when other people are spreading fake news on those...
25:55This is pre-AI, right?
25:58This is before people use AI to generate those content.
26:03Look at Donald Trump.
26:04Look at Donald Trump.
26:05Donald Trump is making fake news every day and then manipulate the stock market.
26:11But nobody says anything to it.
26:13So it's already not a technical issue.
26:17But of course, having the technology in hand makes them more capable of creating those fake news and disinformation and
26:27flood the internet with hatred and bias.
26:32I don't know the fundamental answer.
26:34You know China is able to prevent itself from those misinformation by creating great firewall.
26:45By not allowing West so-called mainstream media and social media just propagate in the internet of China.
26:55I believe, I personally believe, the regulation of the internet is part of the sovereignty of a country.
27:06It doesn't make sense to say that cyberspace is a free open space, that there is no territory in the
27:14cyberspace.
27:15Because when you say that, it's basically saying as a national state, you should give up any regulation and give
27:23the power to Google and Facebook.
27:26So my opinion is national states should regulate internet space.
27:31I know it's a difficult task.
27:34There is a lot to do to achieve that.
27:37But that is, I think, the fundamental solution to that.
27:41Then back to Global South Media.
27:45I think an urgent task for us is we need to learn the technology and to use it properly and
27:55use it wisely.
27:56Because the problem you mentioned, almost all the mainstream large language models have a very clear pro-West bias.
28:07Or be more specific, a pro-West liberal bias.
28:11When you talk to those models without any fine tuning, you feel that you are talking to an American Democrat.
28:19But that political tendency can be adjusted.
28:24And it's quite easy to be adjusted, I would say.
28:27What you need to do is just set up your ideological and political framework for the AI to follow.
28:36For example, in my own case, I have a very thorough Marxist framework because I'm a Marxist.
28:45So I tend to use Marxist theory to understand and to analyze the world.
28:51For example, I would always start from historical dialectics and materialism.
29:00So I created a seven-layer Marxist ideological framework.
29:07And I always tell, before any task, I always inject that prompt.
29:13So it's basically a prompt.
29:15I always inject that prompt to the AI that says, you are a Marxist theoretical analyst.
29:22You should think, follow this framework.
29:25And then here is a task.
29:27I found it's quite effective.
29:30It can suppress a lot of those pro-West ideological bias.
29:36And it fits well for my task.
29:38Of course, I'm not here promoting Marxism.
29:40I'm just saying there are options.
29:43People should learn how to build their own ideological framework and use that to adjust the behavior of AI.
29:50That is clear.
29:52And also, I was thinking regarding what you were saying about how sometimes this conversation is being had in terms
29:59of lack of regulation or the need for regulation in terms of how much a state is regulated.
30:07And there has been a lot of campaign from the West regarding what you were just saying, right?
30:13Internet is a free space.
30:15And a lot of things that we know are not such, like what you were saying right now regarding the
30:22shadow banning and the just neglecting of all the posts regarding Gaza.
30:27For example, the ongoing genocide and Cuba, of course, Venezuela.
30:32So we know that there is regulation, that it's not a matter of having or not certain regulation of what
30:40is happening on that cyberspace.
30:42But rather, who is taking those decisions, who has the capacity to implement them.
30:49And somehow in that discussion, they are being able to impose a narrative where what we are discussing is not
30:57really what is happening.
30:59We are sometimes entertained in media discussion regarding the importance or not of regulation.
31:07And from a Western perspective, the regulation before you talked about how strict China is regarding certain aspects of AI
31:18development.
31:18And sometimes Western media tries to present that in a bad light as if it were an over-regulation and
31:26presenting the lack of such regulation as a positive thing.
31:29We're seeing in the spaces that Western operation is also heavily regulated.
31:35It's just heavily regulated with other principles and other actors in space, sometimes prioritizing the monopolies and their needs and
31:47wants from this cyberspace.
31:49I recently had an article on this topic.
31:52So it's basically a theoretical debate with Antonio Legri and Hart.
31:59It's about how do you identify the roles, the players in political dynamics, right?
32:07Antonio Legri insists that in order to fight against the empire,
32:14OK, we all agree that the empire now already has a good combination of the United States as the government,
32:25it's military, it's intelligence, now it's technology.
32:29But the approach to fight against the empire, Legri insists, should be autonomous collective of the multitude.
32:41But I argue, as Robert Cox said, there are at least three important roles in the political space.
32:51There is empire, there is civil society, and there is national states.
32:58We should not ignore national states as so far the most powerful way of collecting and uniting people in the
33:08global south.
33:09It's an achievement of the national independence movement, right?
33:15So we should not ignore the existence and the power of national states.
33:22The civil society should find a way to have a unification, to have a united front with the national states,
33:30so that they can together resist the empire.
33:36So I think that is when people talk about countries, especially a few important countries,
33:45China, Venezuela, Cuba, Vietnam, DPRK, Iran, right?
33:51You know what I'm talking about.
33:52They try to figure those countries like an evil authoritarianism,
34:00and try to convince a civil society to get away from national states.
34:07But that will only weaken the power of resisting,
34:12and that will make the empire more easy to penetrate.
34:16Jumping particularly from what you were saying regarding the importance of strengthening the national state capacity
34:23of resisting these impositions in terms of the AI development as it's being done, for example, in the West,
34:34particularly from the United States.
34:36Thinking about how this is happening in Latin America in particular,
34:41being it's a territory with the historical dependency in terms of the US using the territory,
34:50the resources and understanding it as its own area of influence.
34:54I think that is pretty much determining what is happening right now and what will happen in the coming years.
35:00And AI will be one of the key matters of discussion in that sense.
35:06We've been talking about the possibilities for Global South countries and everything that could be done.
35:11But we are acting now, we are actually living in a world in which the US model for AI is
35:20the closest one right now,
35:22in terms of at least the will that they have to control this territory that is Latin America and the
35:30Caribbean.
35:30What should these national states be really looking for in terms of what are the true risks that this Latin
35:41American
35:41and Caribbean states are facing in this region and in this area?
35:47What should they be looking for?
35:49And I think that we know the answer because you talked about a unified front already.
35:53But can national states in Latin America tend up for themselves alone or is it necessary to think of some
36:02sort of coordination?
36:03That's a good question.
36:05I think the first thing we need to do is to have alternatives.
36:11I'm not even saying replacements, right?
36:15I'm saying alternatives.
36:16If we are relying on a handful of American companies, we know what is going to happen, right?
36:25They are not going to open their technology.
36:27So the countries will not have, eventually have control to the technology.
36:33When they decide to cut the supply, they can cut the supply.
36:38That already happened to a few countries.
36:40And they can control the price because you don't have alternative.
36:44So that means you will pay higher cost of it.
36:48They will extract your data.
36:50And as we know, data is a new oil of 21st century.
36:54It has economic value.
36:56But Google, I suppose, is not going to have a mutual agreement with the Brazilian government
37:03and to say, let's develop our data together.
37:07I don't see that as happening.
37:09So the first thing, I think, is to have alternatives.
37:12Just to have competition.
37:14And I believe even liberal economists will agree with me that competition is good for a healthy market.
37:24So where does that competition come from?
37:27From China, very clearly.
37:29And it's not from China as a country or as an alternative supplier of those technologies.
37:36It's from an open ecosystem.
37:39Because those American companies, what we see is they are offering a closed ecosystem.
37:45They try to bind you.
37:48They try to reference the users into their own ecosystem so that they can have higher profitability.
37:55The alternative is an open ecosystem.
37:58The latest large language models become commodities.
38:02And the users can freely choose which models they use.
38:07And they can build their own applications on top of it.
38:10So that's the first step.
38:13Then the second is, I think there is a lot to learn from the, please forgive my arrogance if it
38:23sounds like so.
38:24But sometimes I feel myself swinging in the middle of being a Chinese and being a Latin American.
38:32So I sometimes speak like this.
38:34And I think there's a lot of things we can learn.
38:36So this is, this is the right place for you then.
38:42Yeah.
38:43I think a lot of things we can learn from the China's AI plus model.
38:47It shows the development of AI is not for the sake of AI itself.
38:54It's not for the stock, the stock market, the stock price of a few company.
39:01If you look at the NASDAQ today, right, it's basically a few companies are leading the growth of stock market.
39:09That is not what China is looking for.
39:12China, through AI plus strategy, China is trying to make AI a useful tool for every other sectors.
39:21And to increase the efficiency of every other sectors and to help people from having to work on those boring
39:29and dangerous and heavy work.
39:33So I think an AI plus strategy can also benefit many global source countries.
39:40You don't have to have world class, large language model development capacity.
39:46To be honest, that is difficult and that is expensive.
39:49But even without that, you can use the technology to create the applications that benefits your own people.
39:59And then, of course, there are some security bottom lines you have to pay attention to.
40:05Like continuancy of supply, the ownership of data, and the right to control, and the right to regulate, and the
40:15right to decide which direction to develop.
40:18I think there is a tendency, overall, there is a tendency to overestimate the value of foundational research and development
40:31and underestimate the value of applications.
40:35Building applications that the people need, that can help the people, is also important innovation.
40:42We should not underestimate that. We should not undermine the innovation of building useful tools for the people.
40:51So you can see my T-shirt. This is from Allende, 1972.
40:58Allende said, we must create the technology proper to our own reality.
41:05He didn't say that as foundational research and development.
41:11Back then, when they tried to build an IT system in Chile, they didn't have their own computers.
41:17They bought two IBM computers to build the system.
41:20But Allende can see the value of building applications to serve the people.
41:26So he said, we must create the technology.
41:30He didn't say we must adopt the technology.
41:33Because to understand what the people need and build the applications, the information tools, to fit the people's needs, is
41:43also innovation.
41:44It's also creation of technology.
41:46So I think Global South countries should not undermine that.
41:51And we should bring the importance of making applications, making technology to support every other sector and to serve the
42:04people, to give a priority to that.
42:07Jeff, I want to touch on one of the most pressing issues regarding AI, at least on the conversation that
42:13is being had on this side of the world.
42:15That has to do with the impact on the environment.
42:20And in Latin America in particular, that is a great concern.
42:24Because historically, the deployment of resources driven by industrialization and everything that has come from that has had particular impacts
42:34in our territories.
42:36One of the most partially hit territories in terms of the effects of global change and how that climate gap
42:47also affects the global south.
42:49So there is this understanding that there is no AI development without very harsh environmental consequences.
42:59And I wanted to ask you your opinion on that and how should we understand and tackle that discussion.
43:06Yeah, I encountered that question a lot when I visited Brazil and Argentina.
43:13I tried to compare the situation of Brazil.
43:15You know, Amazon is building some new data center in Brazil and the issue of environment and the ecosystem becomes
43:24an urgent one.
43:25And people are angry of that polluting the environment and the spending of electricity and water resource.
43:35So I tried to think about it in comparison with what happens in China.
43:42And I think the difference is that in China we can see a national planning.
43:50So there is a strategy called data in the east and computation in the west.
43:56So you know the east coast of China is where the most population are and the most economic activities happen.
44:06So the most data generated in the east.
44:10And there is a national plan to build large scale data centers in the west.
44:17Particularly in Inner Mongolia, in Ningxia and in Guizhou province.
44:22And then to transfer the data from the east to the west to have computation and then send back the
44:30result.
44:30So that the environmental footage can live in the west provinces where less population live.
44:40Like for example, in the data center in Guizhou province, it hides in a huge mountain.
44:46So nearly nobody sees it.
44:49And the natural abundant water resource can be used for cooling.
44:54And then in the case of Inner Mongolia, they chose a different cooling solution.
45:02Because Inner Mongolia has less water resource.
45:06So they use mostly an air-based cooling solution to reduce the environmental footage of that province.
45:17And of course, in the whole process, there is the regulation and assessment from the Ministry of Environment and Ecosystem.
45:26So it shows a national planning in China's case, which of course doesn't exist when Amazon tries to build some
45:41data center in Brazil
45:42or in some county of the United States.
45:46I think that is, again, you can see it's a political issue.
45:51It's not because of the data center per se.
45:55It's because who gets benefit and who takes a cost and how democracy works in the whole process.
46:05There is no democracy when Amazon decides to build a data center in a small town.
46:12There is only business.
46:13They purchase the land and they build their data center.
46:16And then people start to realize, okay, they are consuming all electricity.
46:19There is no democratic decision process.
46:26So, yeah.
46:29I don't know how to solve this issue.
46:31I think this is the time national states should play their role.
46:37National states should become a platform of democracy to make sure voices of every stakeholder,
46:47since we are talking about multi-stakeholderism, right?
46:50Every voice of every stakeholder should be heard, should be respected,
46:54and then eventually have some national planning to take into account everyone's benefit.
47:02So we go back to it's not a technical discussion.
47:06It's a political discussion at heart.
47:08And from a Latin America perspective, I would add, of course,
47:12that the importance of national states, the capacity of enforcing regulations
47:18that are thinking about the needs of the population.
47:22In Latin America, because of our history of dependency in terms of what the Global North has imposed on us,
47:33that can also only be achieved through cooperation and integration among those national states.
47:39And that is what, of course, Western Global North is trying to hinder, right?
47:46There's just a lot of different threats to follow when we're talking about AI,
47:54and sometimes it seems that we do not know where to look at and what's the right next step.
48:03So you are fully devoted to this topic, and you've been touring Latin America.
48:09If you had to talk to the social organizations, the civil society that is concerned about this,
48:16and how to foster AI development that truly benefits the population,
48:20what do you think we should focus on right now, like in this year, the next five years?
48:28What should be the top priority for Latin America right now?
48:31I think an important development since last year is the maturing of authentic environments,
48:41meaning everyone can build your own intelligent agents for your own tasks.
48:48You don't need to rely on any existing software for your tasks.
48:52So I would almost urge the social movements of Latin America
48:58to start learning how to build your own agents for your own tasks,
49:05being it, I don't know, summarize meeting notes, right?
49:09Every time when I see an organization and somebody, after the meeting,
49:13spend two hours to try to summarize a meeting note, I feel angry.
49:17Because you are already short of resource, and you are wasting your time for those trivial tasks,
49:22giving it to AI, like political research, like communication, and education.
49:32We are working with partners in Brazil and Argentina.
49:38We are working with MST in Brazil to start those trainings.
49:44And so far this year, they have trained, I guess, about a thousand people from different movements.
49:51So I think it's very urgent. The movements need to start right now to learn how to build their own
50:00agents for their own tasks.
50:02As soon as you start, you will find a lot of more interesting possibilities.
50:10And I have a Substack column, so if you go to Substack, search Jeff Shon, you will probably find my
50:19column.
50:19I try to share the latest development, not the advanced technology, just applications in grassroots.
50:29I try to share those application stories from China to the global south, so you can see, okay, this is
50:36possible.
50:37Then I can use the technology to build something similar to this.
50:42Today, I'm writing a story about a small county in Guangdong province.
50:50They build so-called AI party secretary for each village.
50:56So each village has a virtual AI party secretary.
51:01People can ask her questions and get answers.
51:04So this is a very straightforward, very simple, but very useful way to use the technology to serve the people.
51:11So stories like that, I hope they can inspire our readers, our audience.
51:17They can find some interesting ideas from those stories, and then they can start trying to build similar applications for
51:26themselves.
51:26We definitely hope so, too, and it's all about opening those new horizons, the possibility of thinking about this future
51:34in which new opportunities can come from the grassroots, including AI.
51:40And we hope that's the way we are heading.
51:44At least that's the hope.
51:46Well, thank you so much, Jeff, for joining us here in overlap tonight.
51:50Thank you very much.
51:51So that was one more episode of overlap conversations from across the world.
51:56In this case, going to the world of AI.
51:59We will meet again for now.
52:01See you next time.