Introduction to the AI Policy Tracker

uh this amazing tool of the ai policy tracker but here at mindstone live 2.0 we're able to see ray's second um amazing tool or sorry the second one i've come across so i'm sure you might have other stuff in the cupboard uh note to map um so ray if um you take from there yeah brilliant thank you joe and hi hi everyone nice to meet everyone um apologies if i have red eyes right now because it's uh i just woke up and it's 3 a.m in the morning from my hands so um but yeah this is a great um great uh cozy group to be speaking to um i'll i'll

I'll just, I'll share, I'll briefly share the AI regulation tracker that Joe briefly mentioned and share the note to map because these two are sort of related in the way.

So I'll just share screen. Sorry, I just accidentally, I just accidentally went off stage. So let me just show, yeah.

The Global Regulation Tracker

all right it's brilliant so this is what the um so this website that i've built tracks ai laws and regulations around the world all in one place on a world map so you can see the link here it's just techyray.com slash global regulation tracker and this tool is as easy as you can get so there's a you're on the landing page it's just a map and you just click on any country and it'll display a one pager of all the latest developments sector developments bilateral developments and the relevant acts and bills principles framework standards further reading my own personal commentary and also um related groups or organizations that the country might be in or in relation to ai regulation um i've been running this website for almost two and a half years and i've now at the point where i've covered every single country and territory around the world so you can see the entire map is colored in green

um for example you know click on the us there's a summary of the us's ar regulation position on the uk which i think most of you are based in right now um also as quick sign right there so um there will be there inevitably going to be some territories and countries which have more dots than others also for example if i click on cuba the profile is going to be very brief due to the fact that it's a regulation position is not as at the same stage as, for example, a country like China.

So this is like the main. So 1this regulation tracker is has been widely used within the global AI community just to keep abreast of policy developments and regulatory developments. And there are also other side features within the tracker to further support your research.

Additional Features of the Regulation Tracker

So one of the other features is an AI governance library. So I've collated a catalog of templates, guidelines and playbooks that are available publicly online. I've just basically consolidated that into one place.

You can search these three sources by the title. For example, I want to search for framework you'll give me you know some framework suggestions or um you can if you're a organization and you want to get find us a resource that's specific to your profile you can use the recommended for you functions for example i am a bank looking to implement a policy on general gen ai

And you can click on get recommendations. And then based on your profile, it would then suggest you which resources you can look at in the first instance.

There's also a compare countries function. So to the extent that you want to compare the policy positions between two countries, this is where this is for. So for example, let's say I want to compare the positions between let's say Singapore and Indonesia. Click on compare. And it will then produce a one paragraph quick summary of how these two jurisdictions compare. So for example, right now it's just giving me a quick summary of that.

There's also a live newsfeed function. So this will collect news over the past five days. And it acts like a regular newsfeed. So here you can just find some other cool news around AI regulation.

I've also recently released an API that lets you tap into the underlying data of the tracker. This is quite useful for those who, let's say, want to start a newsletter on AI regulation within their own internal organization. So like if you're part of a legal or a compliance function.

You can also, if you're interested, create a private chatbot or news alert on AI regulation. Or if you want to build your own educational website or public repository on this topic, this is where this API might be useful. And yeah, there are various tiers to this API depending on the scope of license.

Special Features: API and Analytics Dashboard

And finally, there's this new feature that I launched not too long ago, which is the analytics dashboard, where basically, based on the tracker data that I've been collating, I've now applied analytics to show different patterns and insights of this topic. So for example, different legislative approaches to regulating AI, the influence of the EU AI Act.

So for those who might know, the European Union has an AI law called the EU AI Act. And this has inspired other sort of similar models around the world. What sort of areas regulators are looking at these days, and also like, legislative and executive activity, news activity, and other activity by organisation.

So I guess the key thing here is that if you're a business or start-up just trying to understand what the market is, or if you're a student or researcher or academic who wants to write an essay or a paper on AI regulation, or if you work in government and you want to like do a study on different on different policy approaches or if you're let's say a lawyer who needs to write advice for on this topic um this is the website that can be used as your starting point to to do that research

Introducing 'Note to Map'

um over the past two years as i was building this and and people were using it they're also quite interested in this map feature and they were they're interested in trying to create their own trackers and this is like a segue into my other platform which is called note to map so it's linked inside here just build and launch your own interactive map tracker as the name suggests and as a description suggests this is basically a template where anyone can create their own trackers and publish that onto public link that can then share with others to view so for example

I'm going to, you know, create a sample map. And then this map demonstrates to map.

So they only say I want to create a tracker that just so let's say I want to hear you just click on a country, add entry. For example, Australia, I am from Sydney. And basically, just you just add countries based on whatever topic you want to create.

So you can create, you know, a world history map, or you can create a tracker of like, you know, the Amazon forest, anything that can track on the world map, this is basically the tool for it. And let me just create one, let's just create a few countries.

And also, you can change colors on the tracker in case you like color coding. So I'm just gonna save that.

And once you've done plotting your tracker, you can then in a publish map, And once it's published, it will then create a public link for you that can then copy and open. And then I'm going to paste this link into the chat so that everyone can see here.

So you can search for it yourself. So I just pasted the public link in the chat. But this will then display your tracker and people can interact with it.

Applications of 'Note to Map'

um so the as i said the use case for this is is very broad so here are some examples so for example i've created this world history map which attracts um you know a summary of the history of every country in the world i've also created a another tracker for you know just summarizing the legal systems of certain countries in the world whether they're civil law or common law or sharia law

and you can also just anything that as i said you can track on the world map this is a tool for it um yeah that's about it it's pretty um pretty simple pretty simple tools yeah i'm happy to you know take questions and discuss yeah

Tremendous.

Audience Interaction and Q&A

So obviously keeping a big eye on questions, questions in the room. I think it's just questions from .

This is so cool.

I'll steal a question from the previous session. Did you use any AI tools to help make these platforms?

So yeah, so for the first one, I do use a bunch of tools to help collect news sources. So in addition to this news feed that I just displayed, I also have used a bunch of scraping algorithms and news APIs to source news, and I use generative

AI services like Perpexily and Gemini to help summarize news articles, which I then review and then post onto the tracker. And so as you can see here, this tracker is updated every day. So I do this routine like every morning, every night to make sure that this tracker is up to date because at the end of the day, a tracker is only useful when it is up to date.

So yeah.

Totally true.

Trends in AI Regulation

Onto substance, have there been any trends in regulation in AI that you think caught you by surprise that people aren't aware of?

Yeah, yeah, yeah.

So this is where the dashboard feature can come in.

So I guess I'll talk about two main trends, just the first trend being like the style of regulation that we're seeing. And you can see here I've plotted the style regulation ranging from vertical to horizontal. So vertical regulation is where government tries to regulate AI based on the application by application focus or domain by domain focus. So for example, regulating deepfakes specifically or regulating content recommendation algorithms specifically. The other end of the spectrum is your horizontal regulation, which is one law that regulates all AI systems across the market.

And so When it comes to AI regulation, there's actually no one right answer. We can see there are actually, there's no strict consensus on what the right approach is. Each country seems to be taking its own sort of philosophy and there's no dominant approach so far.

Perhaps what's more popular and what's the majority these days is the horizontal AI regulation approach, So that's, for example, the EU AI Act. And that, as I mentioned, has inspired quite a few variants around the world.

And speaking of the extent of that inspiration, you can see here, I've plotted out how each of these horizontal AI regulations compared to the EU AI Act. As you see here, the closest examples being from Brazil and Turkey, and the least close examples being Japan, South Korea, Mexico, despite their act also being of a horizontal regulation nature.

So these are maybe just two insights I'll share today. Of course, that's interesting.

Okay, I'll try to formulate this as a question.

But first, Eddie, as you aggregate the regulatory landscape, do you get a sense of how the focus of regulation is changing over time, i.e. can you extrapolate about what regulators think the big threat will be?

yeah so that's a very good question it's actually the focus of the third diagram here so um so i think just to do a timeline um ai regulation started it it was emerging as a topic since let's say 2015 2016 under the field of ai ethics so back then um really the only big thing that many uh ethics and regulations were concerned about were self-driving vehicles. So the big tech companies were trying to develop self-driving vehicles. So around that time, the laws were really focused around product safety, product safety in relation to self-driving vehicles. Then fast forward to 2019, 2018-ish, we start to see laws focus on social media, including content recommendation algorithms due to concerns around algorithmic bias.

So AI algorithms pushing biased content, pushing abusive content, to users or you know companies using your personal data to then try to spam feed you on advertisements and so we saw we saw regulations around that side uh a big turning point as you probably already know was in 2022 when chat gpt came out and that started the whole attention around um generated AI. And from there, we start to see different subtopics being regulated. So as you see here, privacy was the first big area that regulators were focused on. So ensuring that AI companies they use and handle your personal information appropriately and in accordance with the law. And if not, we saw governments plugging the gaps around privacy.

The next big area was around trust. So basically reliability and safety of an AI product ranging from testing standards to audit standards to risk assessment frameworks.

and another big area was around cyber so a lot of of the technical standards that were seen around ai are focusing on the cyber security and resilience of an ai system and also another big area is deepfakes um especially since last year last year was a big year because many countries held their elections and there was a lot of concerns around ai or around AI being used to create deepfakes or misinformation that could distort voting preferences.

Yeah, so these are the so I've been sensing these sort of regulatory for exchanges over the past few years. That's, again,

Job Losses and AI Regulation

Another question from Ed, which is it doesn't look like there's regulatory focus around job losses. That feels like an imminent threat.

Yeah, that's a good that's a that's a that's an interesting point. So I haven't yet actually haven't seen that many actual regulations talking about trying to protect job losses. I have seen some but they've been they've been actually been very few examples.

Most of the job losses have been mentioned in the context of let's say, economic strategy papers or just general minister speeches, but A lot of that hasn't actually been encodified into the law because it's not really an area that can be solved easily by legal measures.

So it's more of an economic issue rather than a legal issue. I think that's the sense I'm getting from the data.

Yeah.

AI Nationalism and Global Influence

There's a book coming out called AI Empires, which has a female author whose name I'm just, or Empires of AI, by Karen Howe. One of the theses of that book, as I understand it, is that Europe, particularly the EU, will remain a kind of regulatory supremacy that will remain mostly influential. China will remain influence of infrastructure and infrastructure supremacy as long as as long as a lot of supply chain and the USA will retain supremacy for emergence of technologies obviously you have a very unique and detailed perspective on that regulatory picture what's your feeling of whether that's the right description of what's going on and whether if so whether it's likely to stay that way

Yeah, that's a very good point. I am seeing a trend of what some people call AI nationalism. So some of you might have heard of the term AI sovereignty, which is the concept that a lot of countries are trying to configure the policies to boost domestic AI capability, whether it be domestic AI manufacturing capabilities and trying to improve self-sufficiency and the resilience of domestic supply chains to be able to service AI internally.

But I'm starting to see that AI sovereignty take on a more aggressive character, which is AI nationalism, which is the same thing, but the policies are a bit more competitive. Well, not competitive, they're a bit more trying to undercut another country's AI capabilities. So the case in point would be the U.S., export controls that restrict the export of critical AI chips to other countries, notably China. That's one example of AI nationalism.

And so I see that aligning closely to Karen Howe's thesis in her book. whether that results in an outcome of you know us retaining supremacy in most areas in china being number one in the infrastructure layer that i still think it's quite early to tell um it might i what i'm seeing right now is that um the direction of both countries air policies are the most sort of um expansive and the most um most economically supportive of industry as they can be right now. So I expect the US and China will have very similar bases in the near future and as jensen huang um recently said it's a it's going to be a long race it's not going to be something that you know we'll find a winner in 2025 or 2030 it's going to go like a long-term gradual grind all the way to the even i'd say 2 100 2 200 so yeah um i think eventually every country will probably have will be split into two camps.

The camp one being those who can actually build AI from scratch, which I call the developer states. And countries that have to rely on other countries' model capabilities in order to service AI, I call that the deployer state. That's already sort of started to emerge. the developer states being US, China, and to some extent, the UK, the deployer states being sort of the rest of the world. I think that distinction will still remain sort of true for the next five to 10 years. But I think that the deployer states will start to catch up and move into the developer states.

Once a specialist say open source models become mainstream, and it will find new techniques to develop you know, powerful language models in much more cost efficient manner.

So yeah. Oh, for sure. Although I know, I think that that distinction is also in the EU AI Act, or I'm not sure it's the phrase. Yeah, develop. Yeah, it's, it's, it's, it's, it's a distinction used to help distinguish responsibilities and liabilities between different stakeholders. So that's at a micro level.

I'm also seeing that development deployer analogy being expanded on a international level, in terms of describing countries. So yeah. I think that there is a Eurocentrism there though, isn't it? In terms of it's a EU piece of legislation that's now framing how we see the international relations of these different AI developing countries.

Yeah, the European Union definitely has had the first... They're technically not the first in legislating AI, but what they have produced being the EA Act is sort of like the poster child of AI regulation that many countries are sort of referring to these days. And yes, the EU AI Act has established a lot of constructs and concepts that are being borrowed and used around the world for in this topic.

Yeah. Tremendous, right?

Conclusion

If you are keen to stick around for my talk after this, obviously you're welcome to, but goodness knows, I think it's 4am now where you're presenting from.

So well done. Well done you. Well done you.

Thank you.