🎂 Builders and Innovators: Conversation with Azeem Azhar

Well, hello Azeem. It's great to have you here today. Today I'm joined by Azeem Azhar, the founder of the Exponential Review Podcast and a lot of other things that we'll get started with.

It's great to have you here today. Chris, it's fantastic to be on CFTV.

And we're going to get to that because I want to talk about how our listeners can get to know your podcast.

So make sure we get a chance to talk about that.

I'll just give you a little bit of background on Azeem. Started out in journalism, The Guardian, The Economist, BBC.

Ended up wandering in and spending a lot of time in the media and tech space, ultimately founding and selling a company, a tech company as a CEO.

And from there, you've gone to do a lot of advisory work. Organizations like Accenture, Harvard Business Review, the World Economic Forum.

In addition, on this side, you have this very successful media empire with the Exponential View.

And what more can you tell us about the Exponential View, which is your latest?

So big. It's pretty much this room that we're in at the moment and a colleague of mine who's in New Hampshire right now.

So it's more like a little hamlet, shall we say.

I mean, it's been a really interesting journey, but I was reflecting on the sort of excitement of talking to someone from Cloudflare, which is, you know, it's a network infrastructure business and we don't think about those companies that often anymore, right?

We think about the applications companies, the Facebooks and the Googles and the Amazons.

But when I first got to know the Internet, which was nearly 30 years ago, so 91, it was really all about the network infrastructure, because even though the Internet was this designed as this end to end thing, a lot of stuff didn't work as planned.

And you kind of had to understand how networks worked to get access to the resources that you need.

And of course, now we find ourselves 30 years later, it's really, really seamless.

And, you know, you like I get our ears chewed off if the so much as one packet drops during a key moment of a Netflix show.

So the world has really moved on in the last 30 years.

What was your introduction to the Internet back then? In 91, that was Mosaic.

What was the, how did you get access prior to Mosaic? It was prior to Mosaic.

Yeah. So I was a university student and we could just about get access to the sort of Unix machine where you had an account and you could timeshare.

The first websites were available, not very many of them, to be honest.

And you did most of your work using something called FTP, the file transfer protocol, and Usenet newsgroups, which ran on the NNTP protocol.

And that was really, really it.

But even then, this idea of being able to access a resource in another country and another organization and download the recommended restaurants for, you know, computer scientists at Digital's research facility was pretty exciting.

I mean, pretty useless if you're in the UK, but exciting nonetheless. I remember everybody competing for the Sunspark back in the day in the labs.

They were the machines.

You were in, early in your career, and when you were in the journalism side proper, what were the early applications of the Internet and how did you see, and you have a very interesting and privileged perspective.

How did you see, if you look back now, how was the infrastructure at that point of time for the Internet in a media context and how has that changed, especially where we are today?

You know, it was so interesting when I started to get involved because you weren't allowed to conduct any commerce on the Internet.

And the NSF, National Science Foundation, which kind of ran the key part of the backbone, had this sort of sense that commerce was not allowed and that changed, Al Gore changed it in sort of 93, 94.

And so, at the time, we were really driven by the ideas that came out of people like Marshall McLuhan and also the work that had been done at the Whole Earth Electronic Link, that this was about creating a global village where you could exchange information and you could collaborate and coordinate over great, great distances.

So, we can talk about the first bit of e-commerce, which is quite a funny story anyway, but before we get there, the thing is that even at that time, really amazing things were going on.

So, in 1991, there was a coup in Moscow and the coup eventually, you know, petered out, but our UNET, which was the sort of Russian University Research Network connected to the Internet, our UNET operators, even though there was a traditional media blackout, were able over Usenet and IRC, Internet Relay Chat, to relay what was going on in Moscow and keep the flow of information going out.

And I remember being a kind of spotty undergraduate, sitting on a green screen terminal in the basement, where we had like five machines we could access, observing this at a distance, and that was like a penny drop moment of, wow, this thing could be really, really different.

But, you know, I think people had realized that already early on.

I mean, you'd had Wired Magazine coming out of San Francisco, starting in 1991, this idea that something interesting was about to happen.

And as you pull that string forward, how has the infrastructure and the Internet that we know, how has that changed as media has sort of progressed?

Hmm. Well, we, I mean, there were some very interesting milestones.

The first websites that I put up were in early 1995. And people had no real idea how we were going to make money on this, because the Internet protocols themselves are pretty incredible.

And they are, you know, they're almost like a Conway's Game of Life, where there are these really simple rules that attach to the protocols, from which all of this complexity can emerge.

And we essentially use the same ones that we did, you know, 50 years, 45 years ago, right, when they started to get documented, but they didn't document any kind of transaction or embedded sort of finance element to it.

So, when those first Internet services launched, we didn't really know how they were going to be funded.

And Wired was one of the first to innovate around Wired Magazine with their website called HotWired, because back in those days, you never named the website the same thing that you named the print publication, right?

So, Time Warner, Time Magazine's website was not called the Time website, it was called Pathfinder.

Why was that?

What's the story behind that? Why was that the case? You know, I just, I don't think we understood that there would be this fusion between cyberspace and real space, right?

At the time, the dialogue was very much cyberspace meets space, whether it was coming from William Gibson, or coming from kind of Jaron Lanier in virtual reality, or Nicholas Negroponte, who founded the MIT Media Lab, who talked about the Negroponte switch between bits and atoms.

So, we constructed this dichotomy.

And in amongst all of that was the sort of governing political culture of the Internet, which was a sort of libertarian, permissionless modality, right?

John Perry Barlow, who was the lyricist of The Grateful Dead, and, you know, sadly passed away, was with the Electronic Frontier Foundation, and he presented the Declaration of Independence of Cyberspace.

So, there was an idea, it was a different place.

And I think because of that, we constructed these alternative brands from the ones, the thing that we had in the physical world.

And, you know, it was very true. So, Wired had HotWired, The Guardian, where I was, had GOTO, G-O-T-O, which is a kind of complicated name anyway, to use.

And as you pull that string forward, if you fast forward, at that point in time, you could see problems that were unsolved.

What problems do you think were, at that time, what problems did you think were hard that turned out to be easily solved?

And the problems that you thought were going to be easy but turned out to be hard?

That's a really interesting question. So, I think looking back at it, the things that we found difficult were some of the really basic fundamentals, like our discs would run out of space from the log files, the HTTP access log files, or there weren't load balancers in those days.

I mean, I'm talking to a Cloudflare audience and you're thinking, what, there are no load balancers?

You know, at the time, what you would do is you would kind of take this open source proxy cache called Squid and you'd reverse it and have a bunch of machines sitting behind the system.

But actually, that was too complicated for us. So, when I launched The Guardian's website, you are looking at The Guardian's first load balancer.

It was me. It was not a piece of code. There were servers and two Ethernet connections.

And essentially, what I would do was I would reboot one server when it got overloaded and there was a memory leak and it could no longer kind of handle the traffic, and I would physically move the Ethernet connection from one, that machine, into the next server and we'd lose kind of a bunch of users and their sessions would time out.

And then about 30 minutes later, one of the other two servers would have run out of memory and slowed down and I would pull the cable out and I would plug it into the other one that had rebooted.

And all this time, I was in the basement in this tiny server room with these Halon fire extinguishers overhead thinking, if something goes wrong, I'm going to be locked in here and essentially asphyxiated.

But, you know, needs must. And I was the load balancer for The Guardian for a few days.

So, those types of things actually ended up, of course, being solved very, very well, you know, first by better software, the stack got better, Moore's law made things faster, and then companies like Akamai came around and, of course, then Cloudflare a few years later.

And so, I think that problem has been solved. Curiously, the thing that hasn't been solved that I thought would be really quite easy to solve has been how users comment and participate on these platforms.

So, we had comments and, you know, there's Godwin's law, which is that on any comment forum, at some point, someone will call someone else Hitler.

And that was true in 1996. And it's still true now, only it's magnified far, far more across the social network platforms.

And actually, I didn't think that was going to be an issue.

I just thought, like, we're going to have great conversations, and we're all going to be sort of these Socratic, wise people using the global village to sort of further human knowledge.

Yeah, it's one of those soft underbellies. Do you think anybody in the space that has a dependence or a reliance on user comments, you think anybody is getting it right?

Or do you think that there's anybody that you think has the responsibility to get it right?

Well, I think there's a couple of different questions.

I think you do clearly see pockets of where comments work very, very well.

And that would be kind of wrong to sort of avoid those.

I mean, I have a we've I'm a member of private WhatsApp groups and private Slack channels, which are really, really kind of great, great spaces.

But quite often, they're the moment these things start to get bigger and bigger, you run into real issues with with how people engage.

And, you know, that's before you even get into the kind of business model aspect of needing to appeal to disgust in order to get people to hit reload, which is a sort of race to the bottom of the brainstem.

But you know, there are, there are people trying to innovate in this area.

I mean, every couple of years, half a dozen bright Silicon Valley people or New Yorkers, or whoever it is, comes up with a new startup, whether it's path or telepath, or whatever else to try to say, let's fix the commenting problem.

I don't think it's it's that straightforward.

And I think, you know, I mean, your other question is who's responsible?

I think that's a different that's a different question.

Do you have a point of view on that? You know, I do, I do have a, I do have a point of view that is kind of grounded in.

It's grounded in the fact that, that where we are today, the these big, for our Facebook being, you know, the biggest of them, are not going away anytime soon.

And they provide capabilities and service that is needed.

widely, the, the, there is, there is clearly a fundamental governance problem that is that exists in in Facebook.

And that governance problem is, it is not is one that we are familiar with for hundreds of years, which is that when you have an absolute power, which Facebook is, it will act with discretion and in arbitrary ways where we're unclear of its incentives.

And, and so you know, a good example is four years ago, in a wreck in a retrospective of war photography, a Norwegian newspaper published a photo of Kim Kim Phuc, who was a nine year old Vietnamese girl who was being burnt by napalm after a South Vietnamese Air Force raid went wrong.

And she's naked. And Facebook said, Well, we don't put up pictures of naked children.

So this is coming down.

And in fact, they even banned the Norwegian Prime Minister. And then they changed their mind 24 hours later.

So that is essentially the exercise of arbitrary power.

And since then, you know, there have been 1000s of other examples, but I choose this kind of very narrow one.

And so I think there is always an issue where there's an exercise of arbitrary power without there being checks and balances or an appropriate chain of accountability that needs to be tackled.

And what has happened over the course of the last 30 years is that we have moved.

We've moved the world into software digitally onto the Internet.

So everything we need to do is mediated by private companies like Cloudflare and Amazon and Apple.

And in doing so, we have not brought the governance and the mechanisms with us to give all of you firms the right clarity around how you should behave.

And fundamentally, what ends up happening is that things end up on, you know, Matthew Prince's desk, if there's been enough of a kind of public outcry, and that's no way for anyone to kind of run this kind of key public infrastructure.

So I think we do have to tackle this governance deficit in some way.

It feels to me like we've got here not because of any particular individual actions, but really as a function of the pace of technological innovation rapidly getting out of sync with the speed with which the public and governments were willing to move.

So I always give long answers.

You have, I think it's great content and context. If you, if just to put some context around the discussion, you get some of the most interesting minds involved in the exponential view, and the kinds of mind expanding content that is shared with your audience, it's great to hear it sort of digested and then put together in a way around the questions that we're talking about or the topics we're talking about today.

If I were to pick you up a little bit and just sort of reframe the dialogue, if you look at sort of contemporary Internet and the uses of it, where is it, where is the Internet working well?

And sort of where is it broken?

And those are, as you referenced a moment ago, there's two different questions, but how do you think about those?

Yeah, well, I mean, it's working well in so many incredible ways.

It's hard to even put your finger on it.

I mean, I'll give you one of my favourite examples is about two years ago, I was in Dublin, in Ireland, at the BT Young Scientist competitions.

This is a national competition that all Irish high school students can enter.

And it's incredibly popular.

And if you're ever in Ireland in January, it's a must go. And I met a 17 year old student there who had developed a piece of computer vision AI that could look at a cervical smear and judge whether it had cancer or not, with a higher, a better false positive and false negative rate than an expert human oncologist looking at the same material.

And that's incredible. The thing is that that was her first programming project.

And she used a an AI technique called a GAN, which was created by a guy called Ian Goodfellow, that was only three years old out of academia.

And normally things take like 20 years or pre Internet, they take 20 years to go from academia into into industry.

So here is a 17 year old 16 when she started borrowing her dad's computer, who's built this thing, teaching herself over the Internet using open source and archive and YouTube and Coursera.

I mean, that's, that's incredible.

I mean, that is that is an incredible outcome. And, and completely, for me encapsulates the very best of the Internet.

I mean, you're, you're solving a real world problem.

And you're doing it in a way that's unexpected.

That pattern you hear repeated. That is a great example. Yeah. But I think that that example, Chris also leads to where the Internet, where the Internet doesn't do well, which is, you know, fundamentally, what we talked about in that example is the elimination of a bunch of historical points of friction, things that would have slowed down that process, like peer review of Goodfellow's work, because he pre published it, and peer review of the code, which went on GitHub, and it was open source, and so on, right, all these things that might have just slowed the pace of this change.

And where we look at the the downsides of the Internet, it's exactly the same thing.

It's that stuff came out really quickly. Because we had this permissionless model, we didn't ask what are the second or third order effects, perhaps we didn't even see what those were.

And so you kind of end up with this digital world that is riddled in cybersecurity terms with incredibly complex and convoluted attack surface.

So anyone can now produce an IoT connected light bulb, which means that we've all got this new open port in our homes that can be sequestered by a botnet and attack our IoT connected lock and lock us in our house until we pay a ransom by Bitcoin.

And, and nobody nobody thought about that. Arguably, the the danger of of user generated content and polarization and extremification that you see on YouTube and Facebook is also an attack surface because it's increasingly used that way by, you know, any nation's adversaries.

And the reason that attack surface is as big as it is, is because the same things that allowed Laura to learn how to develop an AI system to read a cervical smear have allowed us these these companies to get really, really big really, really quickly without anyone asking the question.

A little bit of caution is required, we see some emerging risks.

It's a good cautionary tale, you're right, the friction, I think the way you framed it around the friction not being there in a positive and also in the in the sort of the cautionary cases, that's a smart way to think about it.

If you if let's fast forward into the future 10 years, and we look back on this period of time, how do you think we will describe it?

What what what will be sort of the the meta tags that we put on this time as it relates to the infrastructure and the technology and the behaviors that we have on the consumer side, all the things we've been talking about?

10 years is a really interesting time frame that you've chosen.

We will at some point get to a norm around this, this question of who's responsible for this infrastructure.

And that will also include the question, the the sort of vexatious question of the global Internet run by these, this set of protocols, and how does it overlay on territorial sovereignty, the states.

So at 10 years, in 10 years time, I would hope we would have come closer to answering particularly that second question about, you know, how do you overlay national boundaries on this thing that's sort of inherently global.

And, and we will look back at this period as hopefully a period of of generative discussion around it as we move to figure out what where those norms are.

Because today, what's happened is that essentially, I mean, I'm going to you, this is your cloud players business more than it's much than it's mine.

But as a sort of sort of distant observer, what I see is sort of a real lack of commitment by national governments to come together to tackle the the sovereignty question.

Because the ones who can, who are needed at the table, essentially, the US and China, are the ones who have the best offensive capabilities, as a way of using this, this network to as a form of sort of soft and hybrid power to achieve national outcomes.

And so they don't really have the incentive to say, Hey, guys, let's come together and have a peace convention, and come up with new sovereignty rules over the Internet, because in a way, they're the ones, they didn't need a nuclear non proliferation treaty, the US and Russia until more countries looked like they were going to get nukes.

And so we have this sort of strange situation where the incentives are not aligned from a kind of national security perspective to tackle this problem, the chaos is actually helpful, the lack of rules is actually helpful.

But I think that that will start to change, because at some point, the trade off between the chaos, and the impact on economic welfare starts to change.

And then you start to say, Listen, this is a bit silly, we need to figure out, you know, what is the equivalent of the UN or the World Trade Organization, for trusted networks and trusted data flows?

And what should that look like? Because if we can't trust the kind of core elements of the technology, it becomes very hard to do business in a world that is mediated by the technology.

Are you still there, Chris?

I think, yeah, I think the Internet let us down there for maybe 10 seconds.

But please, I'm going to give you one more prompt. I think everybody can see you.

Okay. Am I still here? I'm going to put it in chat. Okay, chat. Yeah, no problem.

Unfortunately, Internet is letting me down today.

Home connection, I was chatting you, what prediction do you have?

So if you look forward, what do you think will over and again, I'm using 10 years as a reference point, because it's 10 years, 10 year birthday.

And if you think about sort of the forward looking on 10 years, what, what are the principles, if you look back today, like we heard it today?

What is sort of a likely outcome that will be meaningful?

So I think that's a great way of phrasing it.

And the good thing about you, having bad Internet is I can give a better answer second time around.

So here we go. So I think Cloudflare and other infrastructure companies, in which I would include someone like Stripe, and I would include the big four platforms, maybe big five, are going to become an essential part of how our world will run, because it will increasingly move on to the Internet.

And right now, there is no suitable mechanism for you guys to be accountable, and also be able to sit at the table and drive sensible decisions for the benefit of the Internet for its health and its hygiene.

So I and I don't see the ITU or these other governance mechanisms necessarily doing that today.

So I would imagine that in 10 years time, we will have figured out some reasonable way for the core kind of infrastructure companies to be able to, to talk to nation states, and come to agreements around the governance and the operation of a global Internet.

And that makes me uncomfortable for one reason, because you guys are not democratically accountable.

And you're not globally accountable. Fundamentally, at the end of the day, the is the duress and the state force of the US that can determine what you choose to do or not or don't do.

But I think we will find some way of making sure that the key platform companies stay at the table and are able to participate in a healthy way in the governance discussion, which I think is a very is a very important one.

You have a, as I referenced, you have a very privileged lens.

And the, the amount of content that you share is, is really appreciated by many, we're coming up on the bottom of the hour.

Okay, one of the things that is, happens in my home, and I feel very lucky to be hosting you today, when we're on long car trips, we listen to the Exponential View.

So from my, my wife, to my 22 year old son who just graduated from university to my 12 year old that is in sixth grade, like, we all have it on you have quite a range.

How do people subscribe to or get access to the Exponential View?

Yeah, they can just look it up.

If you go to exponential view.co, or if you go to exponential view.co slash CF TV, there's a kind of a special link for all you Cloudflare aficionados.

Thank you for being part of the Cloudflare 10 year anniversary.

Thanks for sharing all of your thoughts on the Internet and data sovereignty, data sovereignty and the Internet in general.

And I wish you well. Thanks again for being part of this. Thanks very much, Chris.