Ivo Vegter - If Privacy Is Dead What Replaces It

The next wave of AI will not only change how we work, it will change what we believe. When deepfakes get plausible, citations can be fabricated, and “evidence” can be manufactured at scale, the real threat becomes an epistemic crisis: a slow collapse of shared reality where even experts struggle to verify what is true. We sit down with freelance journalist Ivo Vegter to talk honestly about that risk, and why the fix is not panic but disciplined source checking, clearer provenance, and a return to careful fact-finding.

We also push back on AI hype. Large language models can feel like intelligence, yet they do not understand what they output, and they cannot replace the grounded learning humans get from the physical world. That matters because it reframes the debate: instead of obsessing over sci-fi extinction stories, we should focus on the near-term consequences of AI-generated misinformation, rising surveillance, and the tempting convenience of always-on personal assistants that know far too much.

From privacy tools and multiple online personas to the politics of power, we connect the dots between big tech, government overreach, export controls, and the uneasy relationship between liberty and democracy. We even get into figures like Peter Thiel and companies like Palantir, and what happens when states become dependent on private platforms. If you care about AI privacy, deepfakes, digital ID, and the future of constitutional democracy in South Africa, this conversation is for you.

Subscribe, share the episode with a friend, and leave a review if it sharpened your thinking. Where do you draw the line between privacy and functionality?

https://x.com/IvoVegter

https://muckrack.com/ivo-vegter

https://dailyfriend.co.za/author/ivo/

Links:

www.bitcoinforbusiness.io

X: @gavingre

X: @BTC_4_Biz

Primal: GavinBGreen@primal.net

NOSTR: npub12qv07tpwk8x8fy2uuqczghpappap395npuxvsx8pgksh97pezv7s8r7qta

Transcript

Welcome And The Big Questions

SPEAKER_00 0:00

Hello everybody and welcome back to the show. This is the Straight No Chaser Podcast where we talk about human freedom through money, technology, philosophy, and economics. Today we have Ivo Vector joining us again. This is the second conversation I've had with Avo, and as always, it's a really good one. Ivo is a freelance journalist. We talk today about privacy. What the future looks like. Are we heading on a better trajectory or are we getting worse? What does worse even mean? And what does better mean? It's a great general chat about how big technology is moving, how that affects our privacy, and uh what we could possibly do about it. So great chat, thanks I as always it was good, and I know you guys will enjoy it.

AI And The Epistemic Crisis

SPEAKER_00 1:16

A lot has happened. But anyway, first off, welcome to the show. Good to have you back. Good morning, Kevin. It's uh great to be on the show again. Um, you have been um releasing a lot of information uh or articles um on the AI topic. Uh it does kind of seem to be the topic everybody's talking about. Um tech stocks in the US, anything connected to AI seems to be going through the roof. People keep on saying AR is the only reason the US economy is actually going anywhere. Um, so whether you're talking to tech guys uh or finance guys or um even marketing guys, business uh strategy guys, everyone is talking about this AI thing. Um so maybe Abo just kick off from your side. Um, where do you see where we are at the moment in terms of AI? Uh, where have we come from and where do you think we could be going to?

SPEAKER_01 2:15

Um, yeah, AI is a fascinating topic at the moment. Um I have a sort of weirdly conflicting view about AI. Uh on one hand, I think that it is in some ways pretty catastrophic. Um, and in particular, what worries me is um uh the sort of what I call the epistemic crisis. Um the fact that it just makes it very hard for us to know what is real. You know, if if AI can even produce scientific papers with fake citations and can then fake those actual papers that it cites, um, it becomes really hard even for experts to tell what is real and what isn't. Um, you know, in popular politics, you're gonna have deep fakes, you know, politicians saying things that they never said, um, doing things that they never did. This to me is a is is a major, major issue. And uh uh, you know, we've already seen what happens when people uh lose faith in the media and lose faith in in science. Uh and this is just gonna make things worse, which means we need to go back to some sort of uh um modus operandi where we fact-check things, where we have trusted authorities, that we know what the sources are of various you know claims and so on. Um it it um it's really throwing the cat amongst the pigeons in terms of how we form our knowledge and how we know what information is true. So from that perspective, I'm uh I'm of a pretty apocalyptic view. On the

Why LLMs Are Not Minds

SPEAKER_01 3:56

other hand, I think that we all overestimate uh AI significantly. Um you know, when we talk about AI, we generally talk uh in popular parlance, we talk about uh large language models. And uh large language models have very specific limitations. You know, they are basically statistical prediction engines of which words are more likely to follow other words, um, you know, given a certain context. Um the larger the context and the more sophisticated the model, um, the more that looks like human intelligence. But it isn't. Um AIs are not alive, they are not thinking, they are not uh they don't understand what they produce. Um the in fact the term intelligence, you know, it remember that it's qualified, it's artificial, it's not real. Um what large language models do is they reflect our own intelligence back to us because they're based on on training data, on on human-produced uh information. And you know, it's it's it's it's really just a uh vastly more sophisticated search engine to existing human knowledge. So it's not producing intelligence, it is reflecting our own intelligence back to us. And I think people overestimate what um what real intelligence is. Um to become intelligent, um, that's not impossible for machines. Um, but that would require much more than just a language model, it would require a world model. A sufficiently detailed model of our actual world in which the computer, the machine, could then draw inferences um and perform experiments and learn from the the physical world model uh in the same way that we learn from our physical world. You know, we draw inferences about you know what happens when we jump off a step, what happens if we throw something, what happens, what's the difference between dropping a tea uh the porcelain teacup and a plastic cup. You know, one will shutter and the other one will bounce. Um this is this sort of thing that we uh infer from our experience with the real world. So a computer to be anything remotely um human-like in intelligence would need a world model that is sufficiently sophisticated. And I think we're very far away from having the computing power um and and memory to actually create a world model like that. There's research in that direction, but we're very far from actual uh machine intelligence. So I I think you know the whole notion of human extinction and so on, I think that's that's way overblown. Um I think we do face an epistemic crisis about the provenance of information, about where it comes from, about how true it is. But I don't think that we face uh anything remotely resembling you know human extinction or machines taking over.

SPEAKER_00 6:59

Um I I was taking notes while you were speaking there. There were so many things to sort of work through, but uh you mentioned, you know, to get to a uh level of intelligence, we would need a detailed world map. Um there was an expression, um, I love the expression, not sure who actually coined it. Uh it certainly wasn't me, but uh the expression is the map is not the territory. And I popped into my mind while you mentioned, you know, we need a detailed world map. Um, you know, how do you get a map that is so detailed that it actually gives you all the information with without being the thing itself? You know, the the whole point of a map is to kind of um simplify uh in in in a uh you know uh an old school map, a geographical um map, your paper map. You would look at it if it simplify there's a mountain, if it's a cadastral map, there's sort of mountain uh roads and or there's roads and mountains around, but it's kind of simple. I mean, it's the the mountain indicated on the map is not the actual mountain for obvious reasons. Um otherwise, you know, you don't need the map anymore. Um so uh yeah, building a world, uh a world map, I mean that would be incredible. Uh do you have any idea uh what that would even look like?

SPEAKER_01 8:15

Uh no. Um to be honest, you're quite right. Uh this is this is a and this is a problem with computer models in general. Um, you know, I've always felt the same way about uh climate models, for example. You know, to model a climate model, even the most complex climate models with you know the most uh uh billions of parameters, whatever, are still a vast simplification of the actual climate, the actual climate out there. You know, the actual climate out there goes down to molecular scale, atomic scale detail, which we cannot model because you know we just don't have that kind of detail uh in our models. Um the same with economic models. Um, you know, we we have models about how people might behave uh in response to a particular economic policy, but we simply don't have the underlying information, the underlying detail um, you know, of people's subjective, of individuals' subjective wants and needs that um that could validate that thing. So when you know when I say a world model um is far away, I really mean it is far away. Um it it's I don't know at which level of detail a world model would be good enough for a computer to to match humans in terms of you know its ability to deduce uh um uh facts about the physical world. Um I don't know when that will happen, but I I don't think it's anywhere it's anywhere near close because yeah, you're right. Um the the only the only sufficiently detailed world model is the actual world. Um, you know, and then then you're talking about how now you need a machine that actually has enough sensors and observation capability, same as us, that you know, it can see the real world and understand the real world in that sense.

Surveillance Data And World Models

SPEAKER_00 10:13

So, you know, in order to get the amount of data that's required to build this map, which will maybe lead to this intelligence, um, you know, we do seem to see in the world we live in at the moment that surveillance seems to be on the increase as well. Um, whether it is uh cameras in suburbs that are documenting vehicles that are moving or facial recognition as people walk in and out of buildings, uh, all this move on to digital IDs, uh, talk of central bank digital currencies, all these technologies that can uh record individual human action and maybe through that infer wants and needs and that sort of thing. Do you think that these might be going hand in hand? That the increase in the surveillance will lead to an increase in data, which will lead to an increase in uh capacity?

SPEAKER_01 11:12

Yes, I think so. Um and this goes to a very fundamental uh conflict in the whole debate around privacy and so on. Um, you know, I've been very resistant to using personal assistance, for example. You know, I've never used Google Assistant on my phone. Um, you know, I don't like uh or sort of Siri on Apple phones. You know, I I simply don't like a computer that second guesses what I'm doing and tracks everything I'm doing. Um, in the very early days of of when these things first came out, you know, I got sort of a very uncomfortable feeling with my phone sort of, you know, I was a way I just woke up at a hotel and my phone tells me, oh, listen, you've got this car coming and your flight is at 11.40. Um, so you need to get moving. What the hell? My phone is now managing my life. I uh how does it know this? And it deduced this from my from my email inbox. You know, and I I felt very uncomfortable about that, and I've resisted um surrendering that level of detail uh to the big tech companies. Um I I just don't like it. Yet by doing so, I've also surrendered or given up a lot of actual useful functionality. You know, the idea that a phone can act as a personal assistant um is really quite appealing. You know, it saves time, it makes you more efficient, it means you don't forget things, um it gives you ideas that you otherwise wouldn't have had. You know, the whole thing of, oh, I see you're in uh you're in greenside. Um would you like to know what uh great local restaurants there are? Well, yeah, it beats me by you know, toddling down the street and looking at the at the at the restaurant frontages. Um it's just a more efficient way of doing things, um, and it gives you new ideas. So I've given up a lot of functionality in the pursuit of a privacy uh that I'm not even convinced I can maintain. You know, um we're on Google Meet right now. I have to be logged in via Google. That means that Google knows that we're talking, Google knows that we're connected, Google knows what we're saying because I guarantee you are I don't know, maybe Google isn't uh eavesdropping on the conversation and transcribing it. But they could be. Um I don't know. You know, so I'm not even sure that your pro your your privacy in that sense is that it's still possible to protect that privacy. Um I like to think so. I mean, but I go through I also go to significant lengths to you know to try and hide what I'm doing. I use a VPN, I use ad blockers, I use um tracker blockers, uh I have an encrypted email facility. I prefer to use Signal uh rather than WhatsApp or or um other IM platforms. Um I don't use a single cloud provider, I use I use several. So I I'll use one for cloud data storage, which you know synchronizes my my my desktop with my laptop and so on. And I use another one for uh email and another one for VPN. So you know, there's no correlation sort of attacks there on on my privacy. And even so, I feel that I'm not doing enough, that it's it's a futile effort. Um and now with the rise of AI, I'm also starting to think that, well, you know, the the the capabilities of these AIs, the the opportunities that they create in terms of doing things more efficiently and and uh you know, doing things that you never could would have been able to do before. Um how does that square with my instinct for uh privacy preservation? And it's a it's a very difficult thing, you know, it's it's it's it's it's the old conundrum between um you know freedom and and and and safety. Um you know, in this case it's privacy and functionality. Um how do you where do you draw that line? Um and I'm not sure that I really have an answer for that.

SPEAKER_00 15:37

Yeah, it's a very interesting one. I I sometimes wonder if um we're not going to be finding ourselves in a world where we have two personas. Uh one is the one that's plugged into the system, as we call it, all your Google accounts and your ARs and your, you know, all these sort of things. Um and that is almost like your your your Joe soap, um, your card-carrying member of society, uh, voting, blah, blah, all that sort of thing. And then you may end up having a secondary uh ultra ego, where you know everything is in an alias and it's heavily protected and runs through all these you know VPN and and privacy tech software tools. Uh and and one you're free to speak your mind, and the other one you you know you wave the flag, you card-carrying member, so on and so forth. Almost a bit like those old World War II movies, you know, where someone was a member of the Nazi party, uh, you know, the factory owner or something, and he had to produce stuff for the military, but on the side he was busy smuggling people out of the country or whatever the case may be. Uh, I I I sometimes wonder if we we don't have something like that in our future.

SPEAKER_01 16:51

I like I like that concept actually. And it made me think of um, you know, as a as a journalist. I've been a journalist now for more than 30 years. Um, and that also that always means maintaining a sort of a public image and a knowledge in my mind of what the public does and doesn't know about me. You know, what they are allowed to know and what they aren't allowed to know. Um, you know, you aren't allowed to know my address. Um you know, I don't like to speak about my family. Um, there are certain things that I simply don't that that are not part of my public persona. Um you are allowed to know that I live in Naizna. You know, I don't have a problem with that. That it it influences what I think, how I think about life, think how I think about South Africa. Um you know, I I don't consider that particularly dangerous because there's there's 70, 80,000 people in Nizna, and you're probably not going to find me even if you do try. Um so you you have this private life, you know, what I read is pretty much private. Uh my hobbies are private, uh, my relationships with my family and and and uh physical friends are are private. You know, they are not a part of my public persona. My public persona is uh quite real to some extent. You know, I'm I'm honest as a journalist, but it's curated, it's it's a part of who I am, not um not all of who I am. And I think the same is gonna end up uh happening in the in the online world that you you have multiple personas um and you try to create them in such a way that no single entity can easily build up a complete picture of who you are. Um and then there's the other thing that you have an online persona, but you have an offline persona. Yeah, that's that's very different. You know, you you should never have all of your life online. I mean, this is I I still believe very much in a in an offline private life that you need to keep separate from your online activities. So, yes, in a sense, I think privacy is dead. Um you know, it takes significant skill to preserve privacy, and I don't think the average person has that. Um but on the other hand, there are things you can do to minimize the impact of that fact. You know, and that's uh, you know, like I said, keep multiple personas online. Um use multiple big tech companies rather than just one that has access to all your data. Um you know, use different email addresses, you know, for different things. Um I have uh different logins, different email addresses, uh and and different passwords for every single thing that I log into. You know, I have a password manager that manages that. Um you know, so you you're gonna have to go to some lengths to actually connect all my different online personas, all my different online pass uh email addresses and and build up a complete picture of me. Um can a state level actor do that? Absolutely. If the you know, if the state is after me, I don't think anyone's gonna be able to defend against that. Um even the best, even the best people in the world, you know, the um uh aren't aren't entirely capable of defending against state level attacks. Um so but but I you know, unless it's I I don't see a good reason for the state to come after me. So I'm fairly comfortable that I'm um that my privacy is well enough protected.

Deepfakes And Source Checking

SPEAKER_00 20:39

Um, Abba, I'd like to come back to something you said earlier, uh just talking about one of your concerns about AI is this issue of um deep fakes, uh trusted authorities, that sort of thing. Um I I agree with you. Uh it's very concerning. Uh every time I see something, my first assumption is uh this is probably not real, and I'd have to fact check it first before uh you know passing it on or even commenting on it. Um I I read a great thing the other day that said, you know, if you read a newspaper headline and your first response is that your blood pressure goes up immediately and you get irritated, the chances are it's just a fake news thing because that's what it's designed to do, is just to get you all emotional and hot under the collar. Um but with you know, how do you maintain your trusted um authorities or your industry experts or or or subject matter experts? Uh, if there is this issue of deepfakes, and you know, would you even notice over time if uh some content is being changed and what the person normally says slowly starts changing over time? Um, how do you guard against that sort of thing? Or do you have to keep on just being vigilant the whole time?

SPEAKER_01 21:58

You have to keep vigilant. vigilant and um and you have to to consider your sources you know if if something arrives to me on on a whatsapp group you know or or uh you know some someone sends it along forwards it along then the chances are much higher that it's fake than if I find uh say a video of Trump speaking on the official White House website you know um I've seen so many fake Trump tweets or truths you know on truth social first thing I do every single time I go to truth social and I check that it's true. It's a bit sad that all of these fake ones are pretty plausible um and sometimes they do turn out to be true. But you know I always go to the source. I have to go to Trump's account on Truth Social and see did Trump actually say this. And you have to do this constantly.

SPEAKER_00 22:58

With scientific papers for example I think we're going to get to a point where A we need to trust certain uh top publications to do some of the legwork for you um so we'll learn to trust say nature uh nature journals but not trust um you know some of the smaller predatory journals that you've never heard of um and then if you still unclear if it still sounds implausible then you need to actually certainly as a journalist you need to actually check with the individual researchers a do they exist b did they write this paper C did this paper say does this paper say what your copy says it says um you know are are the are they are they legit and then you can go back to the the the journalistic practice of saying well you know Bob said this which means that if it's still not true then at least use you you have plausible deniability that Bob is lying you know um because you you've we've never been able to guard completely against that um you know you could I suppose you you know you can check with multiple sources but ultimately if someone really wants to defraud you they're going to but as long as you've done the legwork that shows that you've done that you've taken reasonable precautions to make sure that what you see and what you hear is true what you repeat is true um I think we're going to be a lot better off but that's becoming harder and harder um so yeah I think vigilance it's it's you know the old the old thing freedom requires eternal vigilance um we're saying in this case uh knowledge and true information requires eternal vigilance yeah it's probably not a bad thing that um maybe we've just become lazy uh and we kind of just you know we when the newspapers told the truth and the the 6 p.m news was the truth or that's probably a world I grew up in uh nobody seemed to really question it too much. It was just oh well that's the news and you maybe I was younger I just assumed that there is a single truth and the news has just told me what that is so you know you don't argue. It seems uh as you get older there you start finding out well there's uh you know different slants on the same story and it can be spun a whole bunch of different ways. Um

Peter Thiel And State Power

SPEAKER_00 25:27

I spent a bit of time in the US and quite interestingly when you search uh for most topics uh they have a news from the left news from the right option uh on the search results um so you know that's sort of in your face that there could be a uh two stories on the same source uh which is super interesting um but you you mentioned uh something about things slowly changing over time and it it uh you did an article that you titled what's the deal with peter teal um and I think in your in your opening line there you said you know how did Peter once a libertarian uh seeking escape from democratic socialism come to build the tools of state coercion um what is happening on that side over there um I mean he he's one of the famous uh what do they call them paypal uh PayPal mafia guys uh goes back a long time um but seems to be libertarian but now is um doing unlibertarian things yes um he is actually I think it's quite dangerous in many ways um I I find you know for someone who claims to be libertarian well first he he he uh completely lost faith in democracy now I understand that as a libertarian um libertarians have a very uncomfortable relationship with democracy on the one hand they believe everyone ought to be equal uh before the law um on the other hand they are faced with the very real issue that you know they don't agree with the tyranny of the majority that the the whole majoritarianism aspect of democracy doesn't sit well with the idea of individual freedom you know why should I be subject to a government for which I didn't vote you know just because a majority of my countrymen voted for that government um this is a a fundamental philosophical um conflict in in you know libertarian thought uh the idea that that uh democracy is not conducive to individual liberty um this is why I've always I've I've long argued because I I can't think of a better way of a fairer way um to run a country than by democracy and I I don't see on which grounds I can deny someone a vote um I also don't see from a utilitarian point of view a system that is less likely to produce um revolutions you know if people feel that they have a stake in the government and they voted for the government then they're not likely to revolt.

SPEAKER_01 28:17

If they're run by an elite and don't have a say in the government then they are more likely to revolt. So I don't see an alternative to democracy. But that's why I've always insisted on um constitutional democracy. I insist on a constitutional that protects individual rights and individual liberties. And I think that's a that's a distinction that people don't often enough make. And I think the the idea that a constitution ought to preserve um you know classical liberal principles I think that's very very key to democracy. Now to go back to Peter Thiel he rejects democracy altogether and he wants to establish sort of free entities outside of the control of governments.

SPEAKER_00 29:08

You know he was into the seasteading movement uh sort of 20 years ago um when that still seemed like a a reasonable possibility um he would happily buy an entire country and then run it as a corporate entity um which is you know it's entirely private property and anyone would be subject to the rules of the private owner of that land um rather than to a government what I find confusing is how he um is now working so closely with the US government in in you know sort of building the tools of state surveillance and building the tools of of of uh making war and so on and that that really baffles me um and I have a feeling that where he's going is he feels that if he uh controls those tools if he controls the information that governments have then governments become reliant on him rather than the other way around um and it's a way uh it's a way for him to to buy the control that he wouldn't otherwise have over governments and subject governments to to to to his will so he's gone from being a libertarian in a way to being uh a a a pretty authoritarian fascist um which is very worrying you know and it's it's if you if you not if you're not careful and you're not uh really dedicated to the to to individual liberty um strict libertarianism can actually lead to that because you know once you've once you think of everything in terms of corporate relationships and and uh private property relationships well then the richest guy makes all the rules um and I'm not sure that that's a world that I you know that I would be very keen on um I I mean governments don't seem to have a great track record of just stepping back and allowing some wealthy individual to call the shots um so it would be interesting to see what happens uh down the line if the govern US government decides he's too powerful. I mean you know if China has uh famously um subdued individuals that have sort of spoken out of turn or don't seem to be waving the flag hard enough you know when the camera's on them or whatever the case may be um so I wonder if we may find other so-called uh democratic countries starting to do uh the same sort of thing if they feel threatened especially if it's by one single wealthy individual you know quite easy to make one person go away Putin's also thrown a whole bunch of um oligarchs out out of windows um either that or russian oligarchs are just uniquely uh clumsy around um around windows um so yeah i you know that that is your risk you become once once governments feel that they that they no longer hold the power or that these oligarchs aren't useful to them anymore then you're gonna start seeing that sort of uh very ugly real politic playing out um you know so it it it depends i mean but that's that's I think why teal supports Trump because he knows he can manipulate Trump you know he knows he can keep Trump um on his side um Trump is very vulnerable to to flattery and to you know to gifts um uh and and bribery so um trump is for to for someone like Teal and the same goes for for um some of the other oligarchs you know like Musk um he's an I he's an ideal head of government I think they'll have more difficulty with uh with people who are less well inclined to the the oligarchy um but it's concerning you know it's uh when when when everything depends on on Palantir and the whole government depends on Palantir um well it becomes very difficult to to say okay well we're gonna rip and replace and get someone else in here um it becomes a he becomes very hard to dislodge so I'm not sure exactly how that's gonna play out and I think on those lines we've seen a similar sort of confrontation between the US government and anthropic recently um and that seems to be about AI capability in military applications um and then there's been a whole lot of public forwards and backwards between the two on that so there there seems to be another confrontation that's just you know looming something's waiting to go wrong there.

Anthropic Export Controls And Futile Bans

SPEAKER_01 34:10

But you see there um I don't think they they relied on anthropic you know they were going they were going I mean they wanted the Pentagon to run on Claude you know they wanted the Pentagon to run on Anthropic's AI specifically um to the exclusion of the of of other AI companies um but they hadn't built that yet you know they hadn't become reliant on on anthropic so when anthropic said okay you can do this we'll we'll help you with all of this stuff but we will not tolerate our AI being used for autonomous weapons and what was the other clause they had um anyway they they they they wanted humans to be involved in the in the chain of command um they wanted a human backstop and the Pentagon said no and then they kicked Anthropic out um can they do the same with Palantir? Well I'm not sure that they can I think Palantir is now so interwoven in the in the Pentagon and the various other government departments it becomes very hard to kick Palantir out once if if teal starts making demands um I think the Anthropic thing is is interesting because I mean it's it's now gone to the point where they've just declared their their their frontier model um to be subject to export controls which which anthropic cannot comply with um other than by blocking access to that model altogether. It doesn't know which of its users are American citizens and which aren't um and seeing as even foreign nationals within the US and foreign nationals working for Anthropic are barred from using those frontier models there's no way that they can make those frontier models available to American citizens without making them available to foreign nationals. So that effectively bans um Anthropic's frontier model so they throw away all the the capabilities of a frontier model all the advantages it has all the the the lead it gives American individuals and companies over their Chinese and Russian rivals um just because they worried that you know a foreign agent could use the model to to conduct cyber attacks um it's uh it show it shows a very um authoritarian uh instinct in the government that they'll do that and I'm i'm I'm not sure that this is really about the possibility that uh you know that the model could be used to conduct cyber attacks because the other models can also be used to conduct cyber attacks to to to conduct cyber research you know um security research and and conduct attacks um they are either equal to sometimes better than the the mythos slash fable model of of anthropic um or they are not very far behind and will soon be there and that goes for models like deep sea Chinese models too you know so I'm not sure that they really are going to be effective at all in in keeping a weapon out of the hands of foreign adversaries um so the question then becomes is this targeted at Anthropic specifically because anthropic is on the wrong side of the Trump administration um and I think that's true. I think that uh you know that's this might this might just be revenge for not going along and and and allowing its model to be used for for autonomous weaponry.

SPEAKER_00 37:59

Yeah it it kind of has the the smack of um uh controlling nuclear technology you know in the old days sort of trying to apply the same rationale to this new technology and saying well this is as potentially lethal as nuclear weapons were but you know the big difference is we don't have to smuggle uh physical quantities of uranium across borders um with the you know the AR stuff it's it's ones and zeros with nuclear proliferation they went to great lengths to try and uh carve out a way for people to use nuclear for peaceful purposes um you know with rules and inspections and and and you know great levels of control you know you build a nuclear reactor that's fine but the IEA IAEA comes and inspects you several times a year um you disclose exactly how much uh uranium you have the enrichment levels what you do with the nuclear waste etc um all of that becomes very controlled because they want to keep nuclear weapons out of the hands of um of of rogue states with um software though it that that is futile that it really is futile um you know if if anthropic doesn't develop that model then someone else will um and if it's not an American company well then it's a Russian company or a Chinese company um you know there's nothing about mathematics which is really all that's uh behind these models um it all it requires is mathematics which is publicly available and access to very large amounts of information which is also publicly available so um it's futile to try and you know stop software in the same way that you applied to nuclear proliferation you know they tried it with the DVD encryption protocol I don't know if you remember those days you know people walking around with the with a number on their t-shirts which is basically a hash of the DVD encryption protocol um proving that the government cannot control that you cannot you know security by obscurity is simply not possible um and and the same thing happened again with intrusion detection systems uh and and in pen testing systems that they tried to block under the the Vasanar uh arrangement in when was that mid mid 2000s so mid-2010s um when they realized that they ended up banning or or blocking their own security researchers the guys who are responsible for the defense against these tools because these tools are are inherently dual use you know you need mythos to protect against mythos class attacks you know if if um if you're gonna block your legitimate researchers from from access to mythos or fable um then only the bad guys are gonna end up having access because I can guarantee you that your average Chinese hacker is not going to ask for a license you know they then they don't care about US export controls.

SPEAKER_01 41:15

As soon as they have a model that is sufficiently capable they're going to be using it. And then and then how how do you defend against that? So it's uh they're trying the impossible um and you know it's it's the belief among governments that they can somehow control a software environment or control ai for that matter um is is is it's borders on the fantastical. You know Asimov wrote about this back in the 40s he he he wrote the three laws of robotics why not because he was proposing three laws that would keep that would make robots safe you know that would keep humans safe from robots but he he wrote those three laws in order to spend the next 20 years writing short stories showing exactly how those three laws could be violated. How they did not keep people safe you know how they had internal contradictions that there is no way to to to defend um algorithmically against you know sort of software based attacks. So it's it's you know the AI can't be controlled. And I think that goes to a different subject now but you know it it can't be controlled. We need to find a way to A live with it live with its consequences and B use it responsibly and and this is I think the challenge for for people in the next 20 years is going to be learning to use AI responsibly um and intelligently and learning to deal with its consequences um the genie's out the bottle you can't change you can't you can't fix you can't restrict this you can't restrain this you can't ban the frontier models you know it's none of this is going to work you can't control them.

SPEAKER_00 43:07

I mean

Government Overreach And AI Regulation

SPEAKER_00 43:08

I I think this is a great way to sort of bring up this uh incredible story recently about AI regulations drafted by our government and turns out half of or parts of it were AI generated um which is I mean it's just it's a uh it's laughable at that point but uh that actually leads me to something that I I really want to ask you about and that is um this government's overreach issue uh I mean you've just clearly um explained why trying to control software is basically impossible. Um so why why is it that governments seem intent on trying to control everything and it seems to be getting worse not better. Um you know you you would have thought that by now in this 2026 we've got all the laws we possibly ever Need? Why do we keep needing new laws? Why do we need new regulations? But more importantly, why do governments feel the need to interfere? It seems to be a global thing. It's not just a South African thing. What is your view there, Abba? What is behind this and why is this happening? And more importantly, do you see this turning at any point in the future?

SPEAKER_01 44:25

No, I don't see I don't see it turning. The answer to that, I think, is actually quite complicated. It has in part to do with a genuine desire to try and keep people safe from real or imagined threats. And this is a populist um sort of thing. You know, people people demand this of their governments. You know, they they always, you know, they look at something that's wrong with the world, um, and they and they they turn to government and say, do something. You know? Um and and this is uh who was it that said if you do that you hold freedom very cheap. Um you know, it's people demand to be kept safe, and they demand that off their governments. So when people see that something is not safe or poses some sort of threat perhaps to their children or whatever, they demand uh safety from their government. Um governments use that in order to amass and increase and and uh retain their power, consolidate their power. Um it was Mencken who said that the you know the the practical purpose of politics, the practical aim of politics is to menace the population with an endless series of hobgoblins, um, you know, none of them real. Um, you know, to to give them an excuse to legislate in order to keep the people safe. So this is sort of built into our political system in a way. Um, and that's why the instinct, you know, even I mean Malazi is a DA minister. Um, you know, sorry, Malazi, he he supposedly believes in free enterprise and and uh you know small government and so on. He takes one look at AI, and besides for ignoring the fact that half the paper was written by um by an AI, he thinks it's perfectly okay to create seven new regulatory institutions in order to govern AI, in order to keep people safe from AI, make sure that people use AI responsibly. You know, when frankly A, it's none of the government's business, and B using AI responsibly, well, that's that's sort of built into our our system of private property anyway. If you put an AI in a car, you know, I mean, because one of the things they're concerned about is not just your use of a chat bot or whatever, but um AI in in factories and in machinery and so on. If you put an AI in a car and a car then makes a choice at some point, do I have a head-on collision with a bus or do I pile into a bunch of school children? All right, those those those are the two options. Avoid the school children and have a head-on with the bus or kill the school children. This is a decision that I don't think an AI can can can reasonably make. Um I don't think a human can make that decision either. But what's gonna happen is this is gonna go to court. And the maker of that AI, the guy who put the AI in that car, is going to have to answer for the for the choice that the AI made. So all of these things about AI safety, this is very easily gonna be worked out in the courts. Most product safety, um, a lot of product safety is is uh simply because of lawsuits. You know when when something you buy says don't eat on the label, right? That's not because the government said put don't eat on the label, it's because someone ate it and then sued. You know. Um most of these warning labels, some of them are quite ridiculous, but they all come from some idiot who did that and then sued. And I think this is a far more free market way to deal with with product safety and and safety of things like AI than uh trying to regulate it in advance and restrict it in advance. Um because by doing by restricting AI in advance, uh A it's very difficult to do because because of the way LLMs work, you know, they're not predictable algorithmic things. You can't just give it a rule and expect it to abide by that rule all the time. It's a statistical inference engine. It's it's it it might um it might abide by that guardrail, but we've we've you know we've shown often enough that those guardrails are not impermeable. Um but secondly, one of the great um features of AI is the fact that it's fairly unpredictable, that it's it's uh you know it's it's it's not algorithmic. It it's what appears to us is originality and the ability to create um you know new links between different types of human knowledge and science. A lot of the usefulness of AI is going to get regulated away if you regulate all the all of the safety in advance. Um so you you're sacrificing the usefulness of AI for some imagined safety. And you know, I it's it's a futile exercise. Um and governments instinctively do that, they instinctively want to control everything in the same way that they instinctively want you to do business only if you have a license, only if you have permission from the government. Um but I really don't think that A, they can and B they should do any of this stuff. Um they're not competent to do it. I mean, I'm sorry, Sully Malatzi, you don't understand AI. I mean, I studied some of this stuff, you know. I've I've got computer science and applied maths in my background, and I spend a lot of time looking at AI. I don't fully understand AI. I wouldn't consider myself qualified to regulate it. Um it's crazy that governments think they they are qualified to regulate it.

SPEAKER_00 50:41

Uh I can imagine a scenario where um the creator of the AIs on the stand can't answer the question, so they put the AI on the stand and the AR runs rings around the person asking the questions because the AI just has access to a greater body of work uh instantly than the person asking the questions. That would be very amusing, yes. Totally. Um

Decentralisation Constitutional Democracy

SPEAKER_00 51:03

so just to coming, you know, to rounding up the conversation, coming back to something you mentioned on democracy, because I think all of these things tie back into this idea of democracy. And I agree with you. I think as South Africans specifically, we've had a great um uh uh uh understanding of how democracy works when you have a government you haven't voted for. Um some people did vote for the current government, some didn't. Uh it's been 30 odd years of current governments, and no amount of voting seems to be solving that problem, um, hopefully in the next election or the one thereafter. Uh, but yeah, you can easily find yourself trapped uh for many decades uh and not able to vote your way out of it. Um do you think uh something like a more direct um form of democracy that's broken down into sort of more municipal level, a little bit like the American system where they have their federal governments but then their state governments and then they have their counties and the people there seem to vote for the chief of police and the county commissioner and that sort of thing. And if you're not happy with them, you just simply don't vote for them the next year because I think every year they have to stand for re-election or every two years or something. Um, do you think that that might help balance the scales in South Africa?

SPEAKER_01 52:26

I think it would. Um I've I've I think decentralization of power um is inherently a good thing. Um and one of the reasons I think so is because if you combine that with the fact that we have freedom of movement, it allows you to choose the government under which you want to live. You know, if you if if you're you know, if policing, for example, in Joburg and Cape Town are run at the local level, then you can choose to live in the city where policing is the most effective. Um, and if you if you if if if if in one town or one province things aren't going well, you move to the province where things are going well. So that way people vote with their feet. And I quite like that idea because it um it rewards good governance and it punishes bad governance governance. You know, bad government governance means you're going to lose your population. If you lose your population means you lose your tax base. Um, you know, that ultimately you lose what you're trying to govern, and that means you lose power. So I quite like this idea of competitive politics, um, in a sense, you know, that is as as decentralized as as you can get. Um, it doesn't solve all problems though. Um, there are certain things that need to be run on a on a national level, um, some even on a supranational level. Um you know, I I wouldn't want something like um uh uh pandemic control, for example, infection control. I wouldn't want that to operate on a local level because you know infections don't don't respect borders. Um it's but that's one of those things. Epidemics is one of those things where you really need cross-border cooperation um and you need some sort of cross-border action uh to be taken. So this um this yeah defense is another one where you know you you can't really defend on a province on a provincial scale or on a on a city scale against a foreign adverse adversary. Um so yeah, it's a good thing. I think um a more federal idea is a good thing. I think the the you know the Cape promise asking for for decentralized policing power. I think that's a that's a great uh a great example. Um but it doesn't solve all of our problems. And I think ultimately we still need to rely very heavily on a constitution that guarantees individual liberty. Um and the South African constitution is okay in that respect, um, but it's far from perfect. Yeah. Um so so you know, in my in my ideal state, I would have a classical liberal constitution and then uh a decentralized or federal sort of um democratic system with fairly direct democracy. You know, I also have a problem some some questions are not really suitable for democracy, you know. You can't have a democratic referendum on every single single decision that the the government makes. Um a lot of people have you know they they have misconceptions about a lot of things, they're they're wrong about a lot of things. So you can't go too far and say, well, you know, people should decide by referendum everything that the state does. Because then you're gonna get a whole lot of stupid stuff. You know, you're gonna get anti-vaxxers that are gonna be voting against vaccinations, you're going to get um, you know, anti-GMO people voting against modern agricultural technology and food, which means you're gonna end up starving. That's um you know, you it's you've got to find a balance somewhere, you know, with state defers to experts and um where where decisions that cannot be taken at a local level are taken at a at a higher level. Um, but yeah, at the moment I think the the bias should be towards increased decentralization and increased federalism um in South Africa.

SPEAKER_00 56:49

Yeah, you know, I I agree with you. Everyone's got their own opinion, and you'll never get a bunch of people in a room to agree. Uh, I mean, boardruns have known that for decades. You know, you can't uh the more people you have involved in a decision, the less likely you'll ever get to one. Um so yeah, uh for me personally, I think uh, you know, governments uh should be small and efficient. And uh I would enjoy seeing governments produce guidelines instead of uh endless regulation that they're trying to control everything. Um it would be nice for them to say, hey, we've got uh you know, there's GMO food, this is what you need to think about, there's non-GMO food, that's what you need to think about. Now, off you go, uh, because I think I can make up my own mind. And if I'm I've made a mistake, I'll go, okay, well, that was done. Uh, I'll go do something else. Uh, it would be great for us individuals to have more decisions over what we're going to do for ourselves uh without affecting anybody else. That's a great suggestion. Yeah, I like that. More more guidelines, fewer regulations. Yeah, 100%.

Closing Thoughts And Where To Follow

SPEAKER_00 57:55

Um, Ave, thank you so much for your time. We're we're sort of pretty much up on time. Um, I'd like to ask you for um closing thoughts. Uh anything you'd just like to add that we maybe haven't covered.

SPEAKER_01 58:08

Um, I just think that you know, we're facing, I mean, we spoke a lot about AI. Um, I think we're facing an extraordinary revolution in the economy. Um, and the impact is going to be disruptive, it's going to be transformative. There are immense upsides. And I think we all need to learn how to use AI and to take advantage of the of the upsides. Um, there are also tremendous risks that we need to learn how to mitigate um and how to deal with when they do occur. So it's um it's a it's very interesting times. Um, I really think we're on a on the brink of a revolution that is as big as the internet, as big as computers themselves, um, you know, as big as the industrial revolution. It's going to change a lot of how we how we do things, uh, how we work. I don't think it'll steal our jobs, but I absolutely think it'll change uh the kind of jobs we do. You know, in the same way that we we did we ended up not using clerks and typers so much anymore after computers came around. Um it didn't lead to mass unemployment, but it did lead to change in the types of jobs that that medium-skilled people did. So, yeah, we're gonna we're gonna see a lot of change. And um I think it's uh yeah, it's an exciting time. It's a bit scary, but it's it's uh it's also very exciting. And there are great opportunities and great upsides to be had.

SPEAKER_00 59:41

Yeah, I I completely agree. I think it's one of the most fascinating things. Um, I mean, me as a non-technical person have been able to just get lost in that a little bit, uh build tools. Um we have software for my business that I have built, uh, Android apps that I have created and put in the app store. Uh it's it's quite amazing. Um, what you would have to get a team of people to do. Uh, you can now, in many cases, um, do it on your own. Uh, I think it makes you it gives it a bit more power, it makes you a little bit more sovereign. Um, but yes, uh, we do have to keep our eyes and ears open, our brain switched on at all times, check the work, don't just assume that because you're playing for a top-tier model, uh, that you know it's the gospel truth right there. Uh, but certainly I think it's going to be a roller coaster ride going forward, and I'm really looking forward to it. It's amazing. Um, Iba, if uh people want to follow you and reach out to you, what's the best way to uh follow your work and get in contact?

SPEAKER_01 1:00:44

Um follow my work. Uh at the moment I write mostly for dailyfriend.co.za, um, which is the journal of the uh Institute of Race Relations. Uh and I've been on X for many, many years, and I'm still there. Uh and it's just at Ivofechte.

SPEAKER_00 1:01:01

I V-O-V-E-G-T-E-R. Excellent. I'll put that in the show notes for anyone that wants to reach out. Um, Ave, thank you again for your time. I hope you have a great uh short week ahead with the public holiday yesterday and the comrades and all that excitement. Um so all the best for the rest of your week. Thanks for your time. Thanks for and thanks for yours. Cheers. Bye.