Latest / Ethical Machines / We Need International Agreement on AI Standards

Transcript

We Need International Agreement on AI Standards · Ethical Machines

0:01
[SPEAKER_00]: Pebra, I guess, is Dr. Joanna Bryson.
0:03
[SPEAKER_00]: She is a professor of ethics and technology at the Herdie School of Governance in Berlin.
0:08
[SPEAKER_00]: And she wants us to think about AI in two different buckets.
0:12
[SPEAKER_00]: One bucket is thinking about AI's product.
0:14
[SPEAKER_00]: So various kinds of companies make a product that's quote unquote AI driven or involves AI, or something along those lines.
0:21
[SPEAKER_00]: And then they sell it to the general public, blah, blah, blah, you know how that goes.
0:24
[SPEAKER_00]: Okay, so that's AI as a product.
0:26
[SPEAKER_00]: And then the other way to think about AI,
0:30
[SPEAKER_00]: So we have electricity, we have airplanes and we have international agreements on how airplanes should be created and maintained and operated safely blah, blah, blah, and maybe what we should do is we should think about AI as utility because it's becoming a piece of digital infrastructure that doesn't respect national boundaries, right?
0:51
[SPEAKER_00]: Anyone who has an internet connection, more or less can access these tools no matter where they're from.
0:56
[SPEAKER_00]: So maybe we should think about it as utility.
0:59
[SPEAKER_00]: The reason this distinction is important is because if you think about AI as a product, then you can think as Joanna points out that when various harms are visited upon people by use of AI, it's already covered under product liability law.
1:13
[SPEAKER_00]: We've got legal infrastructure, whether we're talking about the U.S., the U.K., etc.
1:19
[SPEAKER_00]: We've got the legal infrastructure for thinking about product liability.
1:22
[SPEAKER_00]: So that's great.
1:23
[SPEAKER_00]: Now we just need to enforce product liability a lot in those countries when AI is used to harm people.
1:29
[SPEAKER_00]: Okay, great.
1:30
[SPEAKER_00]: When it comes to AI as utility, though, it gets a bit different.
1:34
[SPEAKER_00]: Now Joanna thinks product liability law doesn't work.
1:39
[SPEAKER_00]: The kind of infrastructure that is being built.
1:41
[SPEAKER_00]: And so what we need instead are
1:44
[SPEAKER_00]: transnational standards that countries agree to abide by.
1:49
[SPEAKER_00]: The way she puts it is that we need something like a coalition of the willing.
1:53
[SPEAKER_00]: These are countries that go in for, hey, here's how we're going to create AI in a way that is used as an infrastructure and it's not going to harm people because they're going to engage in such and such processes, have such and such standards, engage in such and such methods for avoiding the bad stuff that could happen with the use of
2:14
[SPEAKER_00]: Interestingly, very interestingly, she thinks that this transnational community that creates those standards might not even include the US, China, and Russia, so it's three of the biggest superpowers on the planet, and that's because she thinks those countries haven't shown a great deal of interest in creating regulations around safe AI.
2:34
[SPEAKER_00]: So it might be that the coalition of the willing includes perhaps the EU plus lots of other countries beside.
2:43
[SPEAKER_00]: We talk a lot about this, we talk a little bit about what would it be like for such an organization to operate, and how is it analogous or disallogous to the way that the EU operates?
2:52
[SPEAKER_00]: We talk about whether such an organization can be sufficiently agile, given the fast pace of technology, and the relatively slow pace at which regulations transnational or otherwise are passed, can it actually keep up?
3:06
[SPEAKER_00]: And what's the probability of thinking that we can actually form such transnational community to create those international standards?
3:13
[SPEAKER_00]: So lots to talk about in this conversation, Joanna has lots of opinions and arguments that are super interesting, so I hope you enjoy it.
3:21
[SPEAKER_00]: That's what this episode's about.
3:23
[SPEAKER_00]: If you have questions, comments, thoughts.
3:24
[SPEAKER_00]: You can email me at EM at readblackband.com.
3:28
[SPEAKER_00]: Also, as I mentioned, I think last podcast, I'm coming out with a book.
3:31
[SPEAKER_00]: My second book called The Ethical Lamer Challenge in April, 2026, 4 months or so.
3:36
[SPEAKER_00]: So there's going to be more about that.
3:37
[SPEAKER_00]: You can go to my sub-stack readblackband.substack.com.
3:40
[SPEAKER_00]: Subscribe there.
3:41
[SPEAKER_00]: You'll get updates about the book and, I don't know, some other stuff.
3:43
[SPEAKER_00]: Okay, that's enough.
3:44
[SPEAKER_00]: Let's go talk to Joanna.
4:01
[SPEAKER_00]: If I have you right, you think something like, we need transnational solutions for governing AI for making sure that AI doesn't go off the rails and what we've got now is utterly inadequate to the task.
4:13
[SPEAKER_00]: Do I have the view right?
4:14
[SPEAKER_00]: And if so, can you say a little bit about what it needs to talk about transnational solutions?
4:19
[SPEAKER_01]: That's actually pretty close to what I think.
4:21
[SPEAKER_01]: I wouldn't say it's utterly, but it's just they utterly not up to the task, but I think we need to go further.
4:26
[SPEAKER_01]: That's just the nature of regulation.
4:29
[SPEAKER_01]: So in the EU, we, your opinion, we have actually built some pretty interesting transnational solutions.
4:35
[SPEAKER_01]: They're under enforced, but they're still providing some hope.
4:39
[SPEAKER_01]: But there's a really interesting question is to how much you can do without the US.
4:45
[SPEAKER_01]: And the US doesn't even seem interested in domestic solutions.
4:48
[SPEAKER_01]: And so that's a challenge.
4:49
[SPEAKER_01]: That's not only it's not all about American uniqueness, after World War II in 1945.
4:54
[SPEAKER_01]: the U.S. was more than half the world's GDP.
4:57
[SPEAKER_01]: And even though the economy keeps improving, the liberal project has been successful.
5:02
[SPEAKER_01]: So the rest of the world is becoming less poor, and the U.S. is becoming a smaller fraction of that GDP.
5:09
[SPEAKER_01]: So I think we're in a situation now where the U.S. is not, there's like these sort of giant companies, a few giant companies that are more or less down south in the U.S., but it's not really a problem
5:23
[SPEAKER_01]: And I'm not sure it can, which might be part of what's selecting the governments that are interested in handling it.
5:28
[SPEAKER_01]: It's part of what it's empowering them.
5:30
[SPEAKER_01]: So I think it's really interesting to look at both ways.
5:32
[SPEAKER_01]: However, I want to make another caveat, which is that there's a lot of AI that is not in these companies that my client calls hyperscalers.
5:40
[SPEAKER_01]: I don't want to under, under state either in the US or outside of the US, the importance that's just the ordinary applications of people, you know, ordinary applications, like, you know, I don't know, calendar, but also a much smaller companies.
5:53
[SPEAKER_01]: And some of them are good and some of them are bad.
5:55
[SPEAKER_01]: Booking.com has recently been declared by the EU to be a platform on that scale.
5:59
[SPEAKER_01]: I remember when it first came out, it was amazing to have something where you could
6:05
[SPEAKER_01]: That was just great.
6:07
[SPEAKER_01]: So I just, I don't want to say this is all of AI.
6:10
[SPEAKER_01]: I just think that most of AI, we can regulate nationally because it's normally nationally scaled.
6:16
[SPEAKER_01]: But these hyper-scalers, I think are a new problem that was under-recognized.
6:20
[SPEAKER_01]: And I could go on for a bit.
6:22
[SPEAKER_01]: I want to hear about that.
6:23
[SPEAKER_00]: But first I just want to make sure that we've got the thesis down.
6:25
[SPEAKER_00]: So,
6:27
[SPEAKER_00]: What does it mean to have transnational solutions?
6:30
[SPEAKER_00]: Oh, are you thinking of something like we've got the international bill of declaration of human rights?
6:35
[SPEAKER_00]: Is it something like that?
6:36
[SPEAKER_00]: Universal, yeah, that's human rights.
6:37
[SPEAKER_00]: Thank you, yes, Universal, that's human rights.
6:39
[SPEAKER_00]: We've got that, is it something like that?
6:40
[SPEAKER_00]: Or is it something, are you actually something different?
6:43
[SPEAKER_00]: I think Europe has GDPR to that, that a protection regulation, OS has no such thing.
6:49
[SPEAKER_00]: You might think that's not as a bad thing.
6:51
[SPEAKER_00]: It's not the focus.
6:53
[SPEAKER_00]: Yeah, there's a California SCPA, whatever, but Illinois, Illinois, Illinois has had a lot of data privacy stuff in bio metric.
7:00
[SPEAKER_00]: Yeah, yeah, yeah.
7:02
[SPEAKER_00]: Okay, so you might think we don't need a transnational solution for the product for data privacy.
7:08
[SPEAKER_00]: We just need countries to do their do their thing.
7:11
[SPEAKER_00]: Okay, the EU did GDPR, great.
7:13
[SPEAKER_00]: We need the US to have a federal data privacy
7:19
[SPEAKER_00]: that doesn't feel to me like a transnational solution that feels to me like a sort of patchwork of solutions and individual solutions that create an overall protective barrier.
7:30
[SPEAKER_00]: Are you so I guess my question is something like are you pushing for each individual country to feel like do what they're supposed to do with regards to AI governance or are you asking for some kind of transnational coordination such that there is if you like a single document, a single
7:49
[SPEAKER_01]: Okay, yeah, good question.
7:51
[SPEAKER_01]: And to an extent both.
7:53
[SPEAKER_01]: So absolutely individual companies, individual countries have to do enforcement.
7:58
[SPEAKER_01]: That is one of the use of agreements.
8:00
[SPEAKER_01]: That is pre-wend or something called the Vest Follion Convention or whatever.
8:04
[SPEAKER_01]: Where they just decided that rather than defining power as like a city that's a capital that expels.
8:11
[SPEAKER_01]: our as an empire, rather we're going to draw these lines and try to eliminate war.
8:17
[SPEAKER_01]: We're going to draw these lines and say whatever happens in there is your business.
8:21
[SPEAKER_01]: So actually part of that was set up because people didn't like how much influence the catholic church had over the countries.
8:27
[SPEAKER_01]: So they were looking, they were trying to defend themselves against transnational entities.
8:31
[SPEAKER_01]: So, yeah, backing up a little bit into what you just said, don't forget that the EU is not a country.
8:37
[SPEAKER_01]: The EU is 27 countries that used to flatten each other in, like, just a flat housing.
8:42
[SPEAKER_01]: Like, if you look at Rotterdam, immediately, I think whenever the got bombs was actually in the 30s, it was a desert.
8:48
[SPEAKER_01]: It isn't that unusual to create a complete wasteland since the 20th century.
8:52
[SPEAKER_01]: So, these are 27 countries that historically did not like each other.
8:56
[SPEAKER_01]: And they found a way and the big question in my mind actually is, can we get countries that are not terrified of going back into that kind of dynamic to agree to give up?
9:08
[SPEAKER_01]: You give up some sovereignty, but then you gain skill.
9:11
[SPEAKER_01]: And that was the thing that even the largest country, Germany, I think is currently the third largest economy because a bunch of other countries have shot themselves in their feet.
9:19
[SPEAKER_01]: But anyway, it's only if you don't count that you itself as an economy, which is should, because economies are defined by their currency.
9:28
[SPEAKER_01]: And so not the whole of you, but rather the Eurozone, you should think of as an economy, and that would be around the second largest economy now anyway these things fluctuate.
9:35
[SPEAKER_01]: In fact, in 2007, it was the largest economy in the world, it was ahead of China and the United States.
9:40
[SPEAKER_01]: because it was least impacted by a mess up that the US suffered.
9:45
[SPEAKER_01]: Yeah.
9:45
[SPEAKER_01]: I don't want to get too boring here, but I hope for all geeks.
9:48
[SPEAKER_01]: I'm not sure who listens to your podcast.
9:50
[SPEAKER_01]: If we're geeks, it's really, trust me.
9:53
[SPEAKER_01]: You can be a programmer and when you start understanding how like the law works and all these companies.
9:58
[SPEAKER_01]: So the interesting thing is that
10:00
[SPEAKER_01]: countries are necessarily decentralized.
10:03
[SPEAKER_01]: If there's any place for anarchy is happening, it's at the super-nat national level.
10:08
[SPEAKER_01]: Okay, what's going on with, so you could say, if we do something super national, why don't we just have Microsoft or Google or somebody do the ones because they're super national?
10:16
[SPEAKER_01]: But they aren't the kinds of things that we tend to trust with enforcement.
10:19
[SPEAKER_01]: They don't have that kind of legitimacy.
10:21
[SPEAKER_01]: So basically, we define, I've talked to people in the U.S.
10:24
[SPEAKER_01]: They're like, oh, a government's going to collapse or whatever.
10:27
[SPEAKER_01]: The point is that a government is the way that a nation that is defined by a geography has gotten together to coordinate itself.
10:35
[SPEAKER_01]: Basically, you will always have something like that.
10:37
[SPEAKER_01]: Maybe that there'll be larger and smaller things.
10:39
[SPEAKER_01]: Things can fall apart like the U.S. has started.
10:42
[SPEAKER_01]: But nevertheless, you're going to have something that is going to have the executive force within a region.
10:46
[SPEAKER_01]: And that, hopefully, be coordinated by the local population, eventually the most stable solutions and they do that.
10:52
[SPEAKER_01]: Sometimes you have my face or whatever, but sure.
10:55
[SPEAKER_00]: That's good to say.
10:56
[SPEAKER_00]: But it does sound like what you're envisioning is something like.
11:00
[SPEAKER_00]: an organization that's international and character that articulates what proper AI governance or AI networks, whatever we want to call it, what that looks like, and everyone subscribes everyone who's a member of that organization.
11:16
[SPEAKER_00]: Abiza, and that's the kind of thing that we need.
11:19
[SPEAKER_00]: Yeah, largely, yes.
11:21
[SPEAKER_01]: So the thing is, I don't think this is special to AI.
11:23
[SPEAKER_01]: So that's what I was trying to say before, depending on how you define artificial intelligence, I tend to define it very broadly.
11:29
[SPEAKER_01]: So coming from the Minski school of even a thermostat is a little intelligent or a plant-artist, a little intelligent, right?
11:35
[SPEAKER_01]: But even if you're talking about generative AI is a lot of people, unfortunately, are, and they think that's the only kind of AI.
11:40
[SPEAKER_01]: Yeah.
11:41
[SPEAKER_01]: It isn't really that there's only four companies making that.
11:44
[SPEAKER_01]: However, there are companies that I would call, that there's all this other AI, and we need to govern all of it.
11:52
[SPEAKER_01]: It just the same way we govern any product.
11:55
[SPEAKER_01]: It just has to be that if it messes up and makes society worse off,
11:59
[SPEAKER_01]: than people have to be held to account.
12:01
[SPEAKER_01]: And if what happened was unforeseeable, then they don't get held to so much to account.
12:06
[SPEAKER_01]: But if it was something, the other people in their sector knew that was a bad thing to do, then they are reliable for all the damages.
12:13
[SPEAKER_01]: That is ordinary product law.
12:15
[SPEAKER_01]: And one thing that you have specifically said, is look digital technology is a product.
12:20
[SPEAKER_01]: Yes, it can do a service.
12:22
[SPEAKER_01]: Yes, writing it is a service, but the digital product itself has a product.
12:26
[SPEAKER_01]: And that was hugely contested.
12:27
[SPEAKER_01]: And that's one of those things that just kind of snuck through that no one was, people were over fixate on the AI Act.
12:34
[SPEAKER_01]: Honestly, this one came out and the product, let's say, product liability directive.
12:39
[SPEAKER_01]: I can't believe I know this stuff.
12:40
[SPEAKER_01]: I never had any interest in the market.
12:42
[SPEAKER_01]: Now I've got all these, but this like little tiny thing, and it's so many of the problems we've had.
12:47
[SPEAKER_01]: Anyway, so the point is that's the vast majority of AI is handled by that.
12:51
[SPEAKER_01]: The exception is what, as I said, MacCon calls a hyper-skiller.
12:55
[SPEAKER_01]: And I think that those are analogous to utilities.
12:58
[SPEAKER_01]: So when the US was, I always wish we, I'm sorry, I have too many identities.
13:04
[SPEAKER_01]: But when the United States was in a very anti-Chess law, they were interesting keeping companies the scale and individuals, the scale to which a democratic government could govern them, right?
13:16
[SPEAKER_01]: And so that was in the late 19th, early 20th century.
13:19
[SPEAKER_01]: But one of the things they notice is that there's some companies that just are never going to be regulated by the market.
13:25
[SPEAKER_01]: So basically, they were trying to say, what is the minimum that a government can do?
13:29
[SPEAKER_01]: And let's have capitalism in the market and competition solve most of problems for us.
13:33
[SPEAKER_01]: Great.
13:33
[SPEAKER_01]: Well, that's nice.
13:34
[SPEAKER_01]: But sometimes competition doesn't work very well.
13:37
[SPEAKER_01]: For example, telephones, railways, water supplies, things like that.
13:41
[SPEAKER_01]: You tend to have consolidation because it's just more efficient to have one provider.
13:46
[SPEAKER_01]: So in all this slide, there was this thing about you're supposed to look out for new utilities because electricity was new, gas was new, it's something that used to get lights from gas, things like that, natural gas.
13:57
[SPEAKER_01]: And I think that we've just stopped doing that.
13:59
[SPEAKER_01]: For reasons that aren't clear to me, I think web search, add tech, maybe operating systems, that these are examples of a certainly Twitter.
14:07
[SPEAKER_01]: of local communication, that these are examples of, of, of, of, you told digital infrastructure data.
14:13
[SPEAKER_01]: Yeah, yeah, infrastructure.
14:14
[SPEAKER_01]: And that doesn't mean they had that the economy be won.
14:16
[SPEAKER_01]: Actually, again, if you look back at the older stuff, even if you had 10% of a market, people like wait a minute, we probably need more than 10 providers to really have a market.
14:26
[SPEAKER_01]: Yeah.
14:26
[SPEAKER_01]: You go famously has whatever 89% of its market.
14:29
[SPEAKER_00]: Okay, so I think that you think, look, there are two ways that we can look at AI, however we can show AI.
14:35
[SPEAKER_00]: One is to see it as a product, in which case there's product regulations that say, hey, if your product hurts people, you're liable, that's one bucket.
14:42
[SPEAKER_00]: The other bucket is it's not a product on its own, it's something that creates the possibility to make another product.
14:51
[SPEAKER_00]: So it's a utility.
14:52
[SPEAKER_00]: So electricity is not a product, but you can create lots of electric appliances or whatever and so we can do stuff.
14:58
[SPEAKER_00]: So, couple things, one thing that you said is let's start with the product bit, it's a product.
15:04
[SPEAKER_00]: One thing that you said is people have to be able to foresee that there's such and such a harm that's going to be visited upon people.
15:11
[SPEAKER_00]: If they can't foresee it, if it's not reasonable that they would have foreseen such a thing, then they can't be held liable.
15:16
[SPEAKER_00]: Something along those lines where they should have held that.
15:18
[SPEAKER_01]: they have lower liabilities.
15:19
[SPEAKER_01]: So in the U.S. this is called the Acts of God, for example.
15:23
[SPEAKER_01]: I used to live in Chicago and I saw when there was a particular building going off that people were building the floor of those seers tower, which was at the time the world's tallest building.
15:34
[SPEAKER_01]: The seers tower that was right at the top of that on a windy day, all the windows were rippling and then eventually they broke and things like couches, things
15:46
[SPEAKER_01]: and it was afterwards declared an act of God that nobody could have foreseen that that there was going to be a windy day.
15:52
[SPEAKER_01]: There's lots of windy days, but it was some interaction about what was happening with under construction building and an unusually windy day.
15:59
[SPEAKER_01]: Yeah.
15:59
[SPEAKER_01]: And they decided not to holding one liable in particular for that.
16:02
[SPEAKER_01]: And all the people had cart, fortunately, not a single person died, but a lot of carts were completely demolished by these giant chunks of glass that had fallen out of space.
16:11
[SPEAKER_01]: And it was just up to the
16:16
[SPEAKER_01]: But if you had said, look, somebody who was building that building didn't do the right thing and somehow I don't know if it's done, then it could have been their liability rather than the poor people whose cars destroyed.
16:28
[SPEAKER_00]: Yeah, maybe this gets too much into the weeds of what counts as foreseeable, but here's the thing about things like gender by eye.
16:34
[SPEAKER_00]: a huge explanation for why it creates the outputs that it does is the prompts that users give it.
16:41
[SPEAKER_00]: Is an open AI or Microsoft or a Google or a anthropic, et cetera?
16:45
[SPEAKER_00]: What could they reasonably foresee in terms of harms that could be visited upon people?
16:50
[SPEAKER_00]: And
16:50
[SPEAKER_00]: I go on to say two things, they should foresee that all the bad things that could happen to people might happen with AI being involved, but you can't expect them to project against all the things.
17:00
[SPEAKER_00]: Or look, they can't really know what people are going to get up to because people are going to do all sorts of crazy things.
17:05
[SPEAKER_00]: They're going to do, and they're in all sorts of crazy prompts with all sorts of crazy goals and dissatisfile sorts of crazy desires.
17:11
[SPEAKER_00]: So it feels to me that if we see it from the product perspective,
17:15
[SPEAKER_00]: Maybe downstream people who take a foundation model and bet it in a very particular product, they can really understand the context of the deployment, but given that the context of the deployment from the perspective of a foundation model company are literally all the contexts in which people ever act, then it's a get out of jail free card for them because they can't reasonably foresee the things that could go wrong.
17:36
[SPEAKER_01]: So a lot of things that like technologists like you and I I'm guessing you and that's certainly I think we're going to be a big problems.
17:43
[SPEAKER_01]: They aren't that big of a problem because there's been thousands of years of law and people have run into these things before and they have a set of solutions.
17:51
[SPEAKER_01]: So in particular, for example, they're if you're thinking like, well, what could generate a way I do want.
17:55
[SPEAKER_01]: I had that 2017 paper in science that everybody quotes, but never remembers, I had was in.
18:01
[SPEAKER_01]: That was about showing that the bias that's in AI, gender of AI, and also everything with word vectors and stuff like that.
18:08
[SPEAKER_01]: It's exactly the same.
18:10
[SPEAKER_01]: It's the bias in the culture of produces the data the implicit biases in the culture.
18:14
[SPEAKER_01]: So you first of all, a quick word about explicit versus implicit.
18:18
[SPEAKER_01]: you may be the kind of person that works really well with people regardless of gender regardless of the ethic background.
18:25
[SPEAKER_01]: Maybe you actually really seek diversity and love that.
18:28
[SPEAKER_01]: Nevertheless, you could have, most of the stuff you read could associate women with laundry and men with investment.
18:35
[SPEAKER_01]: Sure.
18:36
[SPEAKER_01]: You have these implicit biases that makes it, you're just a little faster reading something if the woman is doing the laundry than if the man is doing the laundry.
18:43
[SPEAKER_01]: And that's something that psychologists already knew about it was called implicit bias and then we showed that the AI is reflecting the same biases and interestingly that as with the laundry case that they that those implicit biases reflected lived experience so.
18:58
[SPEAKER_01]: So in particular, we were able to show that with respect to labor side effects and also awful.
19:03
[SPEAKER_01]: And so sexists think that women don't tend to be programmers, but then I turned out at the time, which was horrific to me as one of the many horrific things about my own paper, was finding out that only like 4% of programmers at that point were women, because I had been a professional programmer in like the 1980s.
19:20
[SPEAKER_01]: And at that time, we were like 12% but increasing.
19:22
[SPEAKER_01]: But apparently that maxed out around, I don't remember, somewhere in the teens.
19:26
[SPEAKER_01]: And then in 2000, it just crashed.
19:28
[SPEAKER_01]: And I'm not really sure what happened.
19:29
[SPEAKER_01]: Maybe it's not this.
19:31
[SPEAKER_01]: Anyway, the point is that now that is a known thing, everyone knows that they go through, and they even know that they don't have to eliminate.
19:38
[SPEAKER_01]: And that was like one of the hacks that got stuck into the AI.
19:41
[SPEAKER_01]: That was a bad thing.
19:42
[SPEAKER_01]: I think from external effect features, people like, oh, I have to have no bias on data.
19:47
[SPEAKER_01]: The data is biased.
19:49
[SPEAKER_01]: The whole point of data is that you're expressing regularity.
19:53
[SPEAKER_01]: So you need to know about that.
19:54
[SPEAKER_01]: The band bias that's stuff that our society is to say we don't want to keep it anymore is called stereotypes.
19:59
[SPEAKER_01]: And you are not going to find, as I just mentioned, not going to find a part of the world that has no such things.
20:04
[SPEAKER_01]: We're always trying to improve our society.
20:06
[SPEAKER_01]: So any data is going to reflect that.
20:09
[SPEAKER_01]: So anyway, but what you can do is say that you shouldn't be making the system worse.
20:13
[SPEAKER_01]: And so you compare it against the metrics of what was there in the data or what was there in the previous human resources solution or something and you say it has to be at least like it.
20:22
[SPEAKER_01]: And everyone knows that now.
20:24
[SPEAKER_01]: And before we wrote that paper in 2017, people didn't know that.
20:27
[SPEAKER_01]: And so that's just that's the kind of thing that once something gets out into the, not just the academic literature, but the trade literature.
20:33
[SPEAKER_01]: When people say, there's duty of care that you have when you get certain kinds of data from people.
20:38
[SPEAKER_01]: I don't know this, I've read this and I don't know for Shia and Bill and track it down to be sure about this.
20:42
[SPEAKER_01]: But there was a theory that people were claiming that most companies knew they had to leave care of the information they had.
20:50
[SPEAKER_01]: So when you went to them and said, hey, do you know anything about people that are in the seventh-feet tall?
20:54
[SPEAKER_01]: They asked me to, would you like us to deliver messages to them?
20:58
[SPEAKER_01]: And other things were saying to us, here's all the data we have on some foot tall people.
21:01
[SPEAKER_01]: And here's all their friends, too.
21:03
[SPEAKER_01]: And it was like two or three obvious companies that were just handing out that data.
21:06
[SPEAKER_01]: or for money.
21:07
[SPEAKER_01]: Sorry.
21:08
[SPEAKER_01]: And there's other companies that were only selling access to what the data they felt that they had to be care over.
21:14
[SPEAKER_01]: But the concept of duty again goes back centuries.
21:16
[SPEAKER_01]: Like a baker, a corner baker, might know that there's some illness in the family, and they weren't supposed to tell everybody else on the block that information.
21:25
[SPEAKER_00]: OK, so I'm not sure that I get the answer to my original question.
21:30
[SPEAKER_00]: I'm sorry.
21:31
[SPEAKER_00]: No, all that's interesting.
21:32
[SPEAKER_00]: But my question is whether,
21:34
[SPEAKER_00]: If we rely on the product liability stuff, we accidentally let the foundation model companies off the hook because they can't reasonably foresee all the mischief that people can get up to.
21:45
[SPEAKER_01]: No, so we let them off a hook a little, but not very much, because we're smart and we start learning something.
21:51
[SPEAKER_01]: So the bias was just one example.
21:53
[SPEAKER_01]: Another thing that's going on right now is the people are saying maybe some of the mental health or maybe all of the mental health than we thought came out of COVID
22:04
[SPEAKER_01]: And if that turns out to be true, then all of a sudden there's going to be some kind of obligations about noticing if people are becoming unwell through over-interactual rampifications, some parts of their personality.
22:15
[SPEAKER_00]: It's a product liability stuff you were talking about.
22:17
[SPEAKER_00]: I thought the idea was not just that.
22:20
[SPEAKER_00]: harm was visited upon people by virtue of using some product.
22:23
[SPEAKER_00]: But also that the product makers could reasonably foresee that harm was visited upon those people.
22:29
[SPEAKER_01]: And what you're missing is that there's a threshold.
22:31
[SPEAKER_01]: So before people figure that out, you're not liable.
22:35
[SPEAKER_01]: And after they figure that out, then they start having obligations.
22:38
[SPEAKER_01]: And actually, in most legal systems, that how big they are,
22:44
[SPEAKER_01]: increases their obligations to do something about this.
22:47
[SPEAKER_01]: So like it's surprising what you can get away with the feeling like second or third in the market and there's like you don't have to pay attention stuff at all.
22:54
[SPEAKER_01]: But the people who are the biggest are the ones that wind up having to figure out how to solve things like that.
22:58
[SPEAKER_00]: So maybe this is a silly question or maybe it doesn't speak to all respective areas of expertise.
23:03
[SPEAKER_00]: But here's a question.
23:05
[SPEAKER_00]: I suppose that people use cue tips to clean their ears.
23:09
[SPEAKER_00]: Even though cue tips says don't do that, don't put these things in your ears because you could puncture any of your drum.
23:15
[SPEAKER_00]: So they tell people don't do it.
23:18
[SPEAKER_00]: It's reasonably foreseen that people will do it, but I hardly think that anyone is suing Q-tip or the makers of things that are like Q-tips for people poking their eardrums and going down for something of those lines.
23:30
[SPEAKER_00]: So is it the case that in all cases in which you could reasonably foresee someone using your product in a way that causes harm that your legally or ethically responsible for it?
23:39
[SPEAKER_00]: I think that Q-tip, the company, or I don't think that it's Q-tip, Johnson and Johnson is, I don't know.
23:45
[SPEAKER_00]: Yeah, it doesn't look like reasonably forcing that a home of yours that it upon people by virtue of use of your product is sufficient for being legally or ethically blameworthy.
23:55
[SPEAKER_01]: Yeah, I think that's a really interesting question.
23:57
[SPEAKER_01]: And again, my understanding is at least in our system, the US and UK systems with that kind of, that's called a lot that comes from previous cases, as well as lots written down.
24:08
[SPEAKER_00]: Yeah, come on.
24:09
[SPEAKER_01]: Yeah, it's not just case others, something else.
24:12
[SPEAKER_01]: Anyway, but the point is that there's a weird garden path between not only, what have you written down, what's been exposed, but how much damage is it worth prosecuting?
24:23
[SPEAKER_01]: What is this sufficient?
24:24
[SPEAKER_01]: So if they apparently, the amount of people that are having the amount of harm caused, it's concerning adequate redress that they've now put this.
24:32
[SPEAKER_01]: They probably used to say, you're cleaning sticks or something, and now they say sticks that are not for cleaning here,
24:41
[SPEAKER_01]: Yeah.
24:41
[SPEAKER_01]: So I think it really has a lot to do with what is even again current politics about what it's like to get enforced and things like that too.
24:49
[SPEAKER_01]: There's a lot of real politics to a lot.
24:51
[SPEAKER_01]: And it's and it's also playing catch up.
24:55
[SPEAKER_01]: And that's again people say how some passable to do this right.
24:57
[SPEAKER_01]: And it's yeah.
24:58
[SPEAKER_01]: It's not about doing it's more like evolution than it's so the thing about evolution is you may be the genetically put together the most amazing bird ever and then someone's each the egg before you even have it born it's evolution is not about finding the absolute fittest despite the survival of it is caricature.
25:15
[SPEAKER_01]: It's really about that on average.
25:18
[SPEAKER_01]: You might not just think into a direction such that they're likely to survive, so that there's these kind of progressions.
25:23
[SPEAKER_01]: And so it's more like that.
25:24
[SPEAKER_01]: It's just, it's various kinds of happenstance, smart people and gauge citizens, all these kinds of things that keep advancing society.
25:33
[SPEAKER_01]: That's what we're always trying to do.
25:35
[SPEAKER_01]: And incidentally, we're going down this direction, which I think is really interesting.
25:38
[SPEAKER_01]: We can talk more about gender of AI.
25:40
[SPEAKER_01]: But I do want to come back and answer one of your questions about how do we do this internationally?
25:44
[SPEAKER_00]: Well, yeah, I'm so about that transnational stuff.
25:46
[SPEAKER_00]: So yeah, let's go back to that for a second.
25:48
[SPEAKER_00]: Okay, if most of AI liability is already covered by product liability, why do we need to wheel in something so big as a kind of transnational organization to which everybody signs up, blah, blah, why not just say, hey, idiots, you've already got product liability, like even the US has that.
26:04
[SPEAKER_00]: So we actually don't need this transnational entity blah, blah, blah, we just need countries to enforce their product liability a lot and just leave it at that.
26:13
[SPEAKER_01]: Okay, great question.
26:15
[SPEAKER_01]: And there's two different answers.
26:16
[SPEAKER_01]: One answer is that some countries are too small and nobody can even bother to figure out how to sell stuff into them if they have their own lines.
26:22
[SPEAKER_01]: That's what they use, sort of about, that they create a common market and they give up a bit of their sovereignty in order to get more people buying and selling their stuff.
26:31
[SPEAKER_01]: And so that's one reason to harmonize not all the laws in the world, but a region.
26:35
[SPEAKER_01]: And so like now other regions like Africa and South America are saying do we want to do something like what the EU did so that people find easier to access our markets and we can make more money too and we can buy more stuff and it's all just easier and we can we can have more regulatory efficacy.
26:50
[SPEAKER_01]: So you can think about it.
26:51
[SPEAKER_01]: If you want to make sure that somebody doesn't come in and sell something that you don't think align to their values, then how much you can demand has to do with like how much repay and it would but is it to comply with their was and over how much money, how easy is it to make money out of coming to your market.
27:08
[SPEAKER_01]: And so the bigger you can make your market and the simpler you can make your lot, the more people will come and work with you.
27:13
[SPEAKER_01]: So that's one reason.
27:14
[SPEAKER_01]: But the other reason was this idea of utilities because things like GPS, airlines, the and web search are things that we don't need that many versions of.
27:26
[SPEAKER_01]: We probably need, it was important that the EU Airbus Boeing.
27:30
[SPEAKER_01]: But Airbus isn't beating Boeing.
27:32
[SPEAKER_01]: Now that Boeing has serious trouble because it's under regulated.
27:35
[SPEAKER_01]: We still need the capacity of both companies.
27:38
[SPEAKER_01]: but it's just fortunate that we have something that's given some example, it's a little better better for together right now.
27:43
[SPEAKER_01]: So anyway, that kind of thing that how are we going to do that?
27:46
[SPEAKER_01]: And that's the thing where I think we do need sort of a coalition of the willing, or in this case, willing means still believes in the role of law.
27:54
[SPEAKER_00]: But what does that look like for, okay, I stop some questions about product liability, but I have a few other people are tired of us talking about product liability.
28:00
[SPEAKER_00]: So I should
28:02
[SPEAKER_00]: Well, I'll just say one thing.
28:03
[SPEAKER_00]: Why not just think that, okay, the US has product liability laws.
28:06
[SPEAKER_00]: The EU has them the UK blah, blah, blah, blah, blah.
28:10
[SPEAKER_00]: So for those smaller countries, if they want to do business with the larger countries, why not just say, hey, listen, if you're going to sell them to our country, you've got to abide by our product liability laws.
28:19
[SPEAKER_01]: If you can do that, a bunch of companies might say, no, and then you don't have as much stuff.
28:25
[SPEAKER_01]: So that's why some people think it's worth giving up some level of sovereignty, it's not like giving it totally giving up, you just cooperating.
28:31
[SPEAKER_01]: Look at the UK.
28:32
[SPEAKER_01]: They pull out of the EU and they're kind of implemented.
28:35
[SPEAKER_01]: And they're still being influenced by the EU, but they no longer have a way to add your stuff.
28:40
[SPEAKER_01]: We'll all back.
28:41
[SPEAKER_01]: So the idea is that that's a G7 country.
28:44
[SPEAKER_01]: So even for G7 country, it can be worth it to work with other smaller countries, even bigger-ass, smaller countries, and come up with something that where more people will be willing to come to the table and negotiate with you and work with you about what it looks okay.
28:59
[SPEAKER_01]: It was a legitimate product.
29:01
[SPEAKER_01]: After the EU Digital Services Act went in, all these companies reorganized their architecture of their products, so that, and now I'm not entirely agree with you about what not calling you until the product, but that's okay.
29:13
[SPEAKER_01]: Anyway, their service might want to call that.
29:16
[SPEAKER_02]: Yeah.
29:16
[SPEAKER_01]: If you reorganize the architecture in a way that they have been lying and telling American academic organizations and American regulators that they could do, but now there was a large enough entity there that they're like, oh, maybe this is worth doing.
29:29
[SPEAKER_01]: And it was just like, it was literally I've been in the chat room and all the things, so people say those same people told us there was no way to do this.
29:36
[SPEAKER_01]: And we had always, that was why part of the reason I was cyber PhD in AI from MIT, you know, and I've also worked as a professional programmer.
29:43
[SPEAKER_01]: And so I was one of the people that was just going into the EU, various people in the parliament, the commission, whatever.
29:48
[SPEAKER_01]: And just tell them, no, don't believe that part.
29:51
[SPEAKER_01]: Somebody has done something, it's made these points.
29:55
[SPEAKER_01]: Someone could keep these records and if they're not doing it, it's an advertising.
29:59
[SPEAKER_01]: And that was part of what we were trying to help with.
30:02
[SPEAKER_00]: OK, so we get the game together, right?
30:04
[SPEAKER_00]: So we get these countries together.
30:05
[SPEAKER_00]: And we say, OK, we need to transnational agreement, blah, blah.
30:10
[SPEAKER_00]: What does the content of that agreement look like?
30:13
[SPEAKER_01]: That's a really super interesting question.
30:14
[SPEAKER_01]: And I'm not one of the kinds of people that tends to do that stuff by myself.
30:19
[SPEAKER_01]: But the one direction I have really been thinking about a lot because of this past year of 2025, it's like, what does it look like?
30:26
[SPEAKER_01]: So the EU is working on a particular side of problems with in particular borders, and we have these issues with veto players.
30:33
[SPEAKER_01]: So you can pick off or with what?
30:35
[SPEAKER_01]: They're called veto players.
30:36
[SPEAKER_01]: People that can make, because anyone can do a veto, right?
30:39
[SPEAKER_01]: Because it was a new idea, people weren't sure about joining.
30:42
[SPEAKER_01]: and any country could just cancel things.
30:44
[SPEAKER_01]: So the GDPR Ireland, which are the big tech, just put all this investment into Ireland and finally, and Ireland never thought it was a good idea to hold them to account.
30:53
[SPEAKER_01]: And then they also start doing that, the big tech also start doing that with Blocksonburg, which is a tiny little country, but it's one of the areas where we go out of money.
31:00
[SPEAKER_01]: And so now Putin is doing this with hungry and Slovakia.
31:05
[SPEAKER_01]: And so did they capable of vetoing?
31:06
[SPEAKER_01]: All kinds of things like that.
31:08
[SPEAKER_01]: And specifically, the hungry keeps vetoing things about what sanctions we should do to Russia.
31:14
[SPEAKER_02]: I should.
31:15
[SPEAKER_01]: Yeah, so it's put in some cheap gas and then urban cancel substance.
31:21
[SPEAKER_01]: So I think,
31:21
[SPEAKER_01]: Well, one thing we need to do is not have, there's a concern because if you don't have something like a veto, then you have this tyranny of the majority thing where we can get bullied.
31:31
[SPEAKER_01]: But at the same time, if you do have it, then you have this capacity for a coordinated external enemies or in challenge.
31:39
[SPEAKER_01]: So one solution that you have in some of its law, but maybe not enough of it, is like you say you have to have representation of at least 20% of the population or
31:51
[SPEAKER_01]: So that could be one big country that's between or a bunch of little countries that's me trying.
31:55
[SPEAKER_01]: But it means you can't just pick off one little country, right?
31:58
[SPEAKER_00]: Actually, you can just be locked in burdened.
32:00
[SPEAKER_01]: Yeah.
32:00
[SPEAKER_01]: So that or even Ireland, Barney wants to update it in the last three.
32:03
[SPEAKER_01]: But it's worth it.
32:04
[SPEAKER_01]: I can't believe it learned so much geography, I didn't even hear it.
32:07
[SPEAKER_01]: Yes, some idea about how many people live in different countries.
32:10
[SPEAKER_01]: Yeah.
32:11
[SPEAKER_01]: But the point is that
32:15
[SPEAKER_01]: that Slovakia was interested in the role of law for a while.
32:17
[SPEAKER_01]: And then right now it's more bending towards the Putin perspective and it may be money laundering isn't a bad thing.
32:23
[SPEAKER_01]: And the US, you know, lying that lying very much up over the role of law, seems less interested in it right now.
32:31
[SPEAKER_01]: And so I, that's why I'm saying we really need it to be agile so that people can Brexit at relatively, it's costly.
32:38
[SPEAKER_01]: It costs everybody a lot to not just you.
32:42
[SPEAKER_01]: I heard someone told me like about how much a cost of this billions for the U.S. to change its entire internal system around the fact that the EU, that the EU keeps not getting the EU.
32:51
[SPEAKER_01]: We need to expect these kinds of things.
32:54
[SPEAKER_01]: Really, I, especially, I think this is a consequence, partly climate crisis, and everything else.
32:58
[SPEAKER_01]: voters will be dissatisfied and there's going to be these big oscillations about which governments are in power and some of them are not interested in the rule of law and they're just supporting billionaires and so you need to have a coalition of people to say if you want to be in here if you're going to be working with us.
33:16
[SPEAKER_01]: Then we expected this kind of revenue returned and this kind of investment infrastructure and these kinds of reports.
33:23
[SPEAKER_01]: And so we'll have to negotiate that.
33:26
[SPEAKER_01]: But I think it's very important that it's reasonably decentralized.
33:29
[SPEAKER_01]: So that's a transcendental thing.
33:31
[SPEAKER_01]: It's tend to be decentralized.
33:32
[SPEAKER_01]: But with clear costs and benefits for not only the companies, but also the countries.
33:39
[SPEAKER_01]: What are they doing?
33:40
[SPEAKER_01]: What are they contributing?
33:41
[SPEAKER_01]: Usually they're contributing infrastructure they're contributing end of course data and all kinds of things.
33:46
[SPEAKER_01]: A user's customer beats this and there's a lot of things we're contributing and they have to educate those people and make sure that they have adequate, like as an electricity and broadband then whatnot.
33:55
[SPEAKER_01]: Yeah.
33:55
[SPEAKER_01]: So what are the countries contributing?
33:57
[SPEAKER_01]: What are the countries charging?
33:59
[SPEAKER_01]: What are the companies signing up to and what are they?
34:01
[SPEAKER_01]: A lot of the compliance with, for example, the ESAI Act, as originally written, there's a little bit of stuff that got chewed more into the 2020-20 three years.
34:09
[SPEAKER_01]: Yeah, to accomplish that.
34:09
[SPEAKER_01]: Yeah, to accomplish that.
34:10
[SPEAKER_01]: Why are there some of the compliance was actually good DevOps?
34:13
[SPEAKER_01]: So the EU actually claimed initially that it wouldn't cost a complete it was doing a good job of whether it's software anything to do the AI Yeah, myself and might have a finish woman wrote a estimate it's actually more like between 80,000 to 100,000 euros and so like Yeah, always for a complete to comply
34:33
[SPEAKER_01]: which it's not zero, but it's also not some American think tanks for coming up with millions.
34:37
[SPEAKER_00]: It's like, yeah, yeah.
34:38
[SPEAKER_00]: So what's the, okay, I'm still not sure what the, not totally clear what the content is, but I get, I'd have a sense of the outline.
34:45
[SPEAKER_00]: Oh good, I guess the other question that was, what's the probability that we're going to get there?
34:49
[SPEAKER_00]: So, especially if you want to get the U.S. involved, do you want to get China involved, by the way, are they going to be part of this?
34:54
[SPEAKER_01]: Honestly right now, the thing is that this can change by the day.
34:58
[SPEAKER_01]: You know, but it's probably a lot of classes and things have happening, but if China and Russia and the U.S. continue to be dominated by strong men who seem to only want to do whatever they want to do with their money and that be constrained in any way, that I don't think that they're
35:16
[SPEAKER_01]: And the interesting thing is that there's already discussions of an alternative role trade organization that does not include those three countries.
35:24
[SPEAKER_01]: And it's not something that the EU is already in, but if the EU joins it, then I think it's originally around the Southeast Asia.
35:31
[SPEAKER_01]: And if the EU joins it, then it's going to be like, again, around half of the world's GDP.
35:36
[SPEAKER_01]: So it's going to be a sizeable.
35:37
[SPEAKER_01]: thing larger than, obviously, larger than the use since it's inclusive and the idea would be that's the kind of thing.
35:44
[SPEAKER_01]: To some extent, like I said, we know it's certain to the extent that the EU is still here, although of course, as people trying to destroy it.
35:51
[SPEAKER_01]: So it's certainly we can do this.
35:53
[SPEAKER_01]: The question is, can we continue to scale it, can the agile, can survive the various kinds of legal assaults and all kinds of assaults that look at the hybrid warfare?
36:01
[SPEAKER_01]: That's one of the big things people are really pounding on in the EU.
36:04
[SPEAKER_01]: I don't know how it is in the US,
36:07
[SPEAKER_01]: December we are at war.
36:10
[SPEAKER_01]: Russia is doing everything it can to dismantle our democracies.
36:14
[SPEAKER_01]: We need to recognize that and that was just I'm not saying that was the lessons that was what was coming through and meeting after meeting and even newspapers.
36:23
[SPEAKER_01]: I was just in the British parliament and I was in a meeting of the council of Europe where people were talking about this stuff and then I went because I never been there before so I went inside in the gallery and watch the house of comments and they were talking about that there too.
36:49
[SPEAKER_01]: And but I think I don't think that other system is sustainable.
36:53
[SPEAKER_01]: I think it tends to go along for a little bit and then it falls apart from its own corruption.
36:58
[SPEAKER_01]: And I think the more that we can show that, like, keep telling people, and that's one of the things that Trump administration doesn't want you to know how nice it is living in Europe.
37:07
[SPEAKER_01]: But the more we can show that was one of the things that, again, my content really well with the Olympics, just showing, like, oh, here we are.
37:13
[SPEAKER_01]: It's really beautiful.
37:14
[SPEAKER_01]: And guess what, all the US athletes are coming out of getting their
37:19
[SPEAKER_01]: That was like, it's just reminding people that things can be good and they, in fact, are good.
37:25
[SPEAKER_01]: And then there's a whole question.
37:26
[SPEAKER_01]: Is that echoed?
37:27
[SPEAKER_01]: Well, it's a gesture.
37:27
[SPEAKER_01]: We're still living off of the occult, which is still the parts of the world.
37:30
[SPEAKER_01]: But we keep trying to fix that.
37:32
[SPEAKER_01]: And again, that's part of the recent GDP of the rest of what was going up.
37:35
[SPEAKER_01]: And so the relative strength of the US is going down, even though it's still building capacity.
37:39
[SPEAKER_00]: Let me ask you one last question about agility.
37:43
[SPEAKER_00]: So you want these regulations, the transnational organization to be agile.
37:49
[SPEAKER_00]: That seems like a big ask.
37:51
[SPEAKER_00]: One is that one of the reasons that the pace of AI, at least as it's been the last three years, has been really bananas.
37:58
[SPEAKER_00]: We started with traditional machine learning, narrow AI, then generative AI, and now we're talking about connecting those LLMs or generative AI's to various tools, and we give it autonomy, and now we have agents, and the sort of level of complexity of these systems seems to be increasing exponentially.
38:14
[SPEAKER_00]: And who knows what they're going to think of next, right?
38:16
[SPEAKER_00]: Now I don't know, maybe we've plateaued, maybe we keep going.
38:19
[SPEAKER_00]: And there's another thing that we have to add to our AI inventory, that we have to figure out how to recover in this thing.
38:25
[SPEAKER_00]: So one question is,
38:27
[SPEAKER_00]: I don't know what the question is exactly, but how reasonable it is to suppose that we can create a sufficiently agile, transnational entity that can govern, at the fast pace at which technology is changing, and then second and then relatedly outside of AI, we have blockchain, which is increasingly adopted by banks.
38:43
[SPEAKER_00]: We have quantum computers, and who knows who's going to use those things assuming, of course, that we get to the fault, taller, and quantum computer, blah, blah, blah.
38:50
[SPEAKER_00]: There's just so many different kinds of technology that then technologists are stitching together.
38:54
[SPEAKER_00]: You can stitch together an AI with a blockchain or AI with a quantum or have a blockchain quantum AI engineer and system.
39:03
[SPEAKER_00]: There's just so many different kinds of technologies.
39:06
[SPEAKER_00]: There's so many different ways of stitching these things together that how can
39:12
[SPEAKER_00]: something like a transnational entity, yada yada, account for or is be sufficiently agile to account for what appears to be the sort of, I don't know, what's the word I'm looking for, not incessant, but I'll just say incessant changing of the technological landscapes towards more complexity.
39:29
[SPEAKER_01]: Okay, so I'm tempted to go and answer the first to a second part of your question and then I'm trying to add which way to go.
39:35
[SPEAKER_01]: I think I'll do the second part first.
39:38
[SPEAKER_01]: So let's talk about blockchain quantum and then we'll come back to AI.
39:42
[SPEAKER_01]: So I think blockchain is absolutely bullshit.
39:44
[SPEAKER_01]: It's just like a very high overhead way of doing distributed storage and rate was better.
39:50
[SPEAKER_01]: And I know there's a lot of hype, but now it's basically why I tell Tesla the elevations went up or whatever, but I don't, I'm not afraid of blockchain.
39:57
[SPEAKER_01]: I'm a little bit afraid of people that believe in it, because there is so much on balance, on equal power, on the on governs, so that I think of a crypto in particular, cryptocurrency in particular or blockchain more generally, both.
40:11
[SPEAKER_01]: It's basically there's limited applications for blockchain and I've really seen a lot of people working very hard to try to make money with an except for in the pyramid schemes I haven't been.
40:19
[SPEAKER_02]: Yeah, okay.
40:20
[SPEAKER_01]: So I think that on the other hand, quantum is real, but it's scaling very slowly and it's extremely costly.
40:27
[SPEAKER_01]: And so what we know is that or what we strongly suspect, I shouldn't say anything certain before it's done.
40:32
[SPEAKER_01]: But what looks like it's going to happen is that only very large companies in very large countries are going to be able to afford it.
40:39
[SPEAKER_01]: And then those ones, well, the everybody else won't have any secrets because all the old encryption we were doing right suddenly is completely penetrating.
40:47
[SPEAKER_01]: Yeah.
40:47
[SPEAKER_01]: And also, I would expect that AI, not all of the AI would quite a lot of AI, like I think the travel and save salesmen problem was one of the things that someone's got to quantum algorithm.
40:56
[SPEAKER_01]: Yeah.
40:57
[SPEAKER_01]: Sure.
40:57
[SPEAKER_01]: So quite a lot of, you know, traditionally, I was about solving logistics.
41:01
[SPEAKER_01]: Do not be too blinded.
41:04
[SPEAKER_01]: The general AI is cool and fun.
41:05
[SPEAKER_01]: There's certain applications.
41:06
[SPEAKER_01]: It's really useful for.
41:08
[SPEAKER_01]: But don't forget that what made Amazon, what it is, which is also telling us thing is logistics, solving logistics.
41:16
[SPEAKER_01]: That's a great thing we do with compute.
41:18
[SPEAKER_01]: And I think that's one of the ways we come back to your question about, how can we solve these really complicated problems?
41:24
[SPEAKER_01]: I don't think we've got any kind of grip on exactly how fast we can change now.
41:29
[SPEAKER_01]: We do have all these superpowers that come from all this compute and all this capacity to solve problems.
41:35
[SPEAKER_01]: And the thing is that if we start recognizing that
41:39
[SPEAKER_01]: Like, it might be, again, a very large deconstruction of national governments because you will suddenly have every little entity is going to say, oh, we could have our own drone's leader, whatever.
41:48
[SPEAKER_01]: And so I do expect that there's going to be a lot more entropy, but I also think there's a lot more capacity and a lot of people decide entropy.
41:55
[SPEAKER_01]: I didn't know how do you want to think about it.
41:57
[SPEAKER_01]: I still, so it's come up to me anyway.
41:59
[SPEAKER_01]: It's a lot of complexity, but a lot of capacity to handle complexity, but I do think that at the end of the day, it doesn't make sense to think about this things in other than human time skills because that's all grounded in our interests.
42:13
[SPEAKER_01]: But going back to the AI part of this.
42:15
[SPEAKER_01]: Again, as I said, things are going to be changing, and there's a lot of complexity, but there's a capacity move fast.
42:21
[SPEAKER_01]: A lot of people think that you have to make legislation have to give up while this complexity.
42:24
[SPEAKER_01]: That's not only AI that's complex.
42:27
[SPEAKER_01]: Cities are complex.
42:28
[SPEAKER_01]: The jet fighters are complex.
42:29
[SPEAKER_01]: We have the advantage of how good technology we build all the time.
42:33
[SPEAKER_01]: And what we do is we just try to keep it in track of the human processes that go into building those things.
42:39
[SPEAKER_01]: The audits that are in both the digital services act and their artificial intelligence act in the EU, their process audits are saying, did you keep up with the standards that you're sector?
42:49
[SPEAKER_01]: Again, that legislator is going to keep up with it, the sector, the trade journalists, those kinds of things are what keeps up with it.
42:55
[SPEAKER_01]: So, did you keep up with what the smartest people made sure everybody knew was general knowledge?
43:01
[SPEAKER_01]: And that's why we all have obligations to do these kinds of things, like your podcast, to make sure that people understand, oh, this is what everybody understood at that point, looking even back in whatever this is 2025.
43:12
[SPEAKER_01]: People understood that was a problem.
43:14
[SPEAKER_01]: So, therefore, we had an obligation to try to address that problem.
43:17
[SPEAKER_00]: You know, it's usually just talk about process and I'll just ask one last question, wrap up, but one of the problems that I have with a lot of the ways that people talk about avoiding bad stuff with AI, bad outcomes with AI, is that it is very compliance driven.
43:32
[SPEAKER_00]: It's, did you do the processes?
43:34
[SPEAKER_00]: Did you do the things?
43:36
[SPEAKER_00]: And that's great if we know the processes secure a certain kind of outcome, but I don't or secure the avoidance of a certain kind of outcome,
43:45
[SPEAKER_00]: If we don't know that, and I don't know that we know the, for instance, for multi-agentic AI, I don't think that we have the processes, if so and so engages in these processes, then we won't have disasters that either result of an agent, an AI agent talking to another agent or something along those lines.
44:03
[SPEAKER_00]: There are certain regulations that are outcome-oriented.
44:05
[SPEAKER_00]: So, you know, your car must drive 50 miles per gallon by 2027 or whatever it is.
44:10
[SPEAKER_00]: And that's it.
44:11
[SPEAKER_00]: However, you get there.
44:12
[SPEAKER_00]: I don't care.
44:12
[SPEAKER_00]: Just make sure that whatever process you engage, you've got to get this kind of outcome.
44:16
[SPEAKER_00]: But with the AI regulations, it seems like it's far more process oriented where we don't know about the call of the connection between engaging in those procedures and securing the outcomes that we want.
44:28
[SPEAKER_00]: There are at least best guesses at what we think procedures are appropriate.
44:32
[SPEAKER_00]: But
44:32
[SPEAKER_00]: I don't think that in many cases we just don't have the procedures in place and part of the problem is that the technology is moving so fast that the innovators want to build more stuff new stuff before we have any idea what the best practices best procedures look like.
44:48
[SPEAKER_01]: Yeah, now some people are complaining about people releasing things without testing them and because it is just so easy to release and digital and so again that goes back to the conversation
45:02
[SPEAKER_01]: that give it how easy it is to release things, what can we do to try to get people to release things a little slower or with a little bit more due diligence?
45:09
[SPEAKER_01]: Again, if we establish norms and that's the thing, if we establish rooms, if we start understanding that's a problem, then maybe even for very small companies that they might want to sandbox or have some kind of a system side that they can check their ideas through.
45:24
[SPEAKER_01]: I think mostly people are much more worried about these utilities, these hyper-skillers, how do you want to call them?
45:29
[SPEAKER_01]: Because that's the stuff that has greater effect faster.
45:32
[SPEAKER_01]: Although I'll say that, I never forget about this.
45:35
[SPEAKER_01]: I don't know if you remember this.
45:36
[SPEAKER_01]: So it was something happening about 2018.
45:38
[SPEAKER_01]: And there was this huge, it looked like a hack, and it took down Netflix and Twitter, and everything went Google.
45:44
[SPEAKER_01]: And it turned out to be like three teenagers or maybe they were 25 of them that had a dark net bot net thing that was running on baby monitors and eventually I wanted to deal with everything but it was like it wasn't even Russia or anything.
45:58
[SPEAKER_01]: It was just three guys that figured out they could do that and then something ran away.
46:02
[SPEAKER_01]: So even very small companies that is one of the issues is trying to put in what are we firewalls, what do we do to stop those things from getting
46:12
[SPEAKER_01]: It's not that I'm saying that there's that there's some approval way that we can be sure that the world is going to be perfect and we're all going to live forever.
46:19
[SPEAKER_01]: I agree.
46:20
[SPEAKER_01]: Yeah, of course.
46:21
[SPEAKER_01]: You happily love her after.
46:22
[SPEAKER_01]: But it is that we do as long as this is what the role of law is about is that we keep being able to when we get good ideas, say, OK, now that we've seen this is worked in this context, at least we have an obligation to not do worse than that.
46:36
[SPEAKER_01]: Yeah.
46:37
[SPEAKER_01]: And even that, like I said, some of that slips one of your like, okay, we're in a context of four, we're in a context of play, then some of the things that normally work don't work.
46:44
[SPEAKER_01]: But even that is something we can understand.
46:45
[SPEAKER_01]: So yeah, I think it's doable.
46:48
[SPEAKER_01]: All right, let's, let's hope you're right.
46:51
[SPEAKER_01]: I do think what we do matters.
46:53
[SPEAKER_01]: And so, you know, everybody that's out there, we're putting good ideas for people can find them, they may make a difference.
46:59
[SPEAKER_00]: Yeah, that's hope so.
47:00
[SPEAKER_00]: Thank you, you're in so much.
47:01
[SPEAKER_00]: Really appreciate you coming on.
47:03
[SPEAKER_01]: Sure.
47:03
[SPEAKER_01]: Thanks for having me.
47:06
[SPEAKER_00]: Thanks everyone for listening.
47:08
[SPEAKER_00]: I hope you enjoyed the episode.
47:09
[SPEAKER_00]: And I want to thank a few people who make this all possible.
47:11
[SPEAKER_00]: Delano Alexander, he's my producer.
47:13
[SPEAKER_00]: He does such an amazing job on everything that he touches.
47:16
[SPEAKER_00]: Garrett Lane is the sound engineer.
47:18
[SPEAKER_00]: That's honestly, he's just, he's too good for me.
47:20
[SPEAKER_00]: Authamia Jairam has joined us regionally.
47:22
[SPEAKER_00]: He's helping us find the best guests to bring on to this podcast.
47:26
[SPEAKER_00]: And Ashley Wasselenko helps us organize all this.
47:28
[SPEAKER_00]: Get people on the schedule, get people's bios, headshots, it really keeps the trade on the track.
47:32
[SPEAKER_00]: So thanks guys, really appreciate all the hard work you do.
47:46
[UNKNOWN]: Thank you.