So, I mentioned this to Vint Cerf the other day, I sort of said this is a picture of the world we are occupying and it will increase in proficiency. And he said, oh well that's ubiquitious computing and Mark Weiser was talking about that in the '70s. And Vint knows that because, of course, he architected the internet, he did TCP/IP, and since 2005 he's been the chief evangelist at Google, where he is involved in spotting the best internet technologies and fostering the idea those products and those services can be best built in the largest possible systems. Marc Andreessen, of course, developed the Mosaic web browser, he co-founded Netscape, he started a web hosting company called Loudcloud back in the day where companies could just plug into computing like it was a utility. And now, of course, at Andreessen Horowitz, he invests exactly in this world of computational intelligence everywhere. Now until recently of course, Marc was off Twitter having had a minor career as the ultimate tweet stormer.
He came back recently onto Twitter leveraging a site called 21.co, where he answers questions for $100 a pop and the proceeds of that go to the organization, Black Girls Code. We love this idea. So, we decided at the session today, every time Mark answers a question we're going to contribute $100 to Black Girls Code. And, of course, in the interest of fairness, Vint's answers get $100 too. So, without further ado, Mark Andreessen and Vint Cerf. [APPLAUSE] [MUSIC PLAYING] QUENTIN HARDY: So, let's start. The cloud. That's going to be a big deal, yes or no? MARC ANDREESSEN: Yes. VINT CERF: Sorry, what? QUENTIN HARDY: That's $200. Keep count. That's $200. VINT CERF: But, I want $200 for that. QUENTIN HARDY: Oh. OK. Yeah well, develop it in the next one. So, we talked about this idea of computational intelligence everywhere in this enormous mediating influence. What do you think will be the most important barely discerned or unrealized dimensions of this, particularly where enterprise is concerned?
VINT CERF: So, my first thought is that we're moving from a space where the cloud is kind of homogeneous and uniform to a space where it can be and is becoming less uniform. It's becoming differentiated and so, the cloud's providers may be able to offer a variety of different kinds of services. At Google, for example, we're seriously looking at quantum computing. We're seriously looking at neural networks. We're trying to make those various capabilities accessible through publicly known APIs. The whole idea here is to give a variety of functional computational capabilities in the cloud systems. But, there's a second possibility that I'm curious about and that's the recognition that there are multiple clouds out there and there is a conceivable utility in being able to move data back and forth among the clouds. Even imagining doing computations partly in one cloud and partly in another and exchanging information back and forth. QUENTIN HARDY: How many big clouds do you think there'll be?
VINT CERF: It's hard to say. One thing that can happen is that companies will want to run little cloudlets of their own. They want to be able to interact with the major clouds. For example, keeping some data on there and the protocols for doing this however are not quite there. It's like the days of the internet. Before the internet, we had networks from IBM and Digital and Hewlett-Packard you could connect their brands of computers together over the networks, but you couldn't link them across brands. And so, the internet said wait a minute, let's see if we can find a way to do that. So, that's where I think Intercloud may eventually emerge but Marc you may have a different view. MARC ANDREESSEN: Let me come at it from a different standpoint. You're talking about the idea basically of intelligence permeating everything and then being connected up to the cloud. I'd just come at it from a different or maybe a more elliptical angle, I think this is something environmentalists should be very fired up about.
I'm kind of surprised they're not, which is leverage on the real world, leverage on physical resources. Let me give you just the obvious example people are playing with right now. The average car in the United States is utilized 4% of the time. So, 96% of the time it sits parked, by the way there are 1,000,000,000 parking spots in the US, and if you aggregate up those 1,000,000,000 parking spots, you have the state of Connecticut literally, in terms of size. This is just massive waste and inefficiency because the cars aren't in the cloud and so, 96% of the time they sit there. Then by definition, 96% of the sheet metal and the glass and everything else that goes into the car that's produced is just sitting idle. Take on the other hand the new world emerging of self-driving cars connected to the cloud and then run as a cloud-mediated network with rides on demand, and you probably can't get a car utilization up to 100% but you could probably get it up to 50%, 60%, 70%, 80%.
QUENTIN HARDY: Step functions [INAUDIBLE] efficiency. MARC ANDREESSEN: 10X increased in efficiency and therefore a corresponding decrease in the number of automobiles, therefore decreasing physical impact– QUENTIN HARDY: And in the creation– MARC ANDREESSEN: –of production. QUENTIN HARDY: In the creation of that efficiency, in itself, there are opportunities. Other things are freed up. MARC ANDREESSEN: From an economic– QUENTIN HARDY: As someone who grew up in Connecticut, equating it to a large disused parking lot is interesting. MARC ANDREESSEN: Exactly. I'll let you speculate on that. It also means you can transform cities. You can turn cities into parks because you can take out a lot of the streets and a lot of the parking lots but just the physical impact of the resource load on the environment. There are going to be all kinds of areas, whereas the world gets smarter and we can capture more data and software, bring it up on the cloud, and then lever the physical world through software.
There's huge efficiency gains that people aren't even thinking about today that I think also will go straight to a lot of the current environmental issues. QUENTIN HARDY: Vint, I want to dig a little bit–, sorry go ahead. VINT CERF: This sounds a lot like trying to model what's going on and more accurately, in Singapore they are building a model of the city. It's a city state. There are about what, 5,000,000 people who live there, but they're building a very accurate model to try to understand how the resources of the city are used by the people who live there. You guys remember the matrix? And no, I don't think I was the model for the architect, but anyway. [LAUGHTER] QUENTIN HARDY: You've been programmed to think that. VINT CERF: But, the important thing is that in the story, they were emulating the population of people in order to feed them ads to see how they would react. But, what the Singaporean researchers are doing is trying to do a very, very fine-grained model of the way in which transportation and other resources in the city are being used in order to understand if they changed things, if they could modify it, it would make a difference.
So, this modeling thing and ability to calculate based on data you received is a really powerful idea. QUENTIN HARDY: So they're going to model for an unintended consequence? VINT CERF: Well, that's possible. QUENTIN HARDY: Yeah, that's great. I wanted to dig a little bit deeper into what you were saying about different types of computing in the cloud, could be CPUs and GPUs, that exists now, neural nets and quantum, eventually, Google's doing research, this is something that you entirely foresee at this point. What are the use cases? What sort of world does that build? VINT CERF: Well, think about the kinds of problems that these various computing engines are capable of solving. The neural networks are not the sort of thing you do for payroll, for example, but yet it's a very powerful tool for machine learning for typical AI kinds of applications. In the case of pun computing, there are certain algorithms that will allow you to complete a computation much, much faster than you would with conventional methods.
On the other hand, some problems running on a quantum computer take longer, or at least no better than a conventional machine. The whole idea here is to get the problem to fit the computational capability. Now we have this broader range of opportunities to match the computing capability with the problem that we're faced with. And that, I think, is unexplored territory at this point. QUENTIN HARDY: It's really going to change the educational system on how coding is taught too, isn't it? VINT CERF: Well, I don't know whether it will change the educational system itself, but it might certainly force the educational system to help people learn how to compute using these different kinds of tools. QUENTIN HARDY: Think about use cases a lot. You could even end up in a regulated computing environment because quantum could be used to damage a cryptocurrency near and dear to your heart and as much as you can perhaps, hit the encryption of Bitcoin using a quantum formula. VINT CERF: OK but Marc would probably agree, this is ping pong, right?
At this stage of the game, it's well known that certain kinds of crypto are going to be made less secure because of certain quantum algorithms. Shor's algorithm, for example, that gets in the way of doing logarithm like algorithm. The solution to that is different mathematics and already people are exploring the whole new range of mathematics that will not break against, at least Shor's algorithm wouldn't break it. I don't know maybe you are looking into that. MARC ANDREESSEN: When you tell a venture capitalist there's a giant technical problem, we just get all excited. QUENTIN HARDY: Find me a smart guy. MARC ANDREESSEN: Find me a smart guy to go solve it and we'll fund it. There is new math underway now and there will be new ways to do that. QUENTIN HARDY: OK. MARC ANDREESSEN: But, it might be a major change. This might be significant and, by the way, it might be the world needs a step function increase in security. QUENTIN HARDY: Yeah. That would be a good thing right now.
MARC ANDREESSEN: For example, this week, perhaps. QUENTIN HARDY: Depends on how you feel about your democracy. Now, you were saying to me earlier that you are seeing a lot more, I hate saying conventional, but historic businesses, non-Valley businesses moving in and taking a really active role in learning about the cloud directly. CEO straight to the problem, learning about ML, how do they approach it differently. What do they want from this? MARC ANDREESSEN: I think there's a history of it. A lot of big companies got excited in the 1990s because the internet had arrived and IBM had e-business and the whole thing. Then, a lot of big companies in 2000, 2001, 2002 breathed a massive sigh of relief and said oh thank god that internet thing didn't work. Stick a fork in it, it's done. Everybody knows the dotcom thing was a bubble that was a joke, it's over and now we don't have to worry about it. They maintained that view until, I would say, about 2010, 2011. Then, I think it flipped pretty much directly into panic.
It's interesting it also coincides with the generational change of leadership at a lot of the big companies where they now have a younger generation of leaders actually running the companies, who grew up with computers, unlike the previous generation. In the last six years, we've seen a significant move and, more recently, a flood of big companies from incumbent industries coming to Silicon Valley and trying to learn what all this stuff means and trying to figure out how to use it to either improve their business or defend against new attackers. QUENTIN HARDY: Are they looking at–, sorry go ahead. MARC ANDREESSEN: Correspondingly, we've also seen the rise of entire new generations of startups, that are now using these new technologies, to directly go after incumbent industries that historically have never been threatened by technology and never been threatened by technology startups. QUENTIN HARDY: An Airbnb essentially models a hotel chain or an Uber models a taxi fleet.
MARC ANDREESSEN: The basic calculus here is that the tech industry historically has been very good at having a big impact on industries in which there is rapid technology change, rapid productivity change, that also turn out to be industries that 1, just aren't that big relative to the size of the economy and 2, aren't that regulated. Media and entertainment, consumer electronics, retail, there's just not that big as a percentage of the economy right. Most of the economy is health care, education, and construction, government. These giant industries that are just big and gigantic and regulated and with huge income and strength. It appears, or at least what a lot of entrepreneurs in the Valley believe, is that the new technologies are now so powerful and so potent, if you bring mobile, plus network access, plus AI, plus whatever, and you apply it into an industry like real estate or transportation or education or health care, you can now actually have a big impact. It wasn't possible in the past and so we're seeing the attackers also shop at the same time.
VINT CERF: I'm actually going to argue that big is not necessarily the only reason the cloud is of interest. The other reason is continuous computation, continuous gathering of data, continuous observation and analysis can make a huge difference in the medical world. For example, you only go to see the doctor when you're sick, so the doctors model of you is you're always sick and he never sees you when you're healthy. If you're monitoring your body for example, then you can get a baseline that says this is what's normal. Now they have the ability to say, uh oh something is not normal. But it only works because you have this continuous monitoring going on. That's probably also true for various kinds of business where you're trying to understand what's going on dynamically and being able to see things in real time makes a huge difference. We don't have time for it in some of our advertising mechanisms. We know in real time how much people are spending on a particular ad.
We can stop showing an ad when it runs out of the limit that the users have set but that only works if you have real time computation of billing. QUENTIN HARDY: I think just to circle back, it's interesting to look at this process of these new companies modeling an industry and then attacking it because when they do this, they don't look at it to talk about how different types of computing are. They don't look at the toolset in a siloed way if you look at a [INAUDIBLE] or an Airbnb or a Snap, there's mobility, there's AI, there's all these things and they're not thinking about them in a siloed way, they're thinking about a very continuous way with the objective of using the best capability towards their end. MARC ANDREESSEN: And they're not trying to defend legacy, investment, status quo. Business models are now trying to find big employee bases that maybe do need to fundamentally transform. It's asymmetric. The odds of success of the attacker are lower than the odds of the defender defending successfully but the challenge is easier, attacking is easier.
QUENTIN HARDY: And even from a risk point of view, if you learn something, you've got an asset at the end of it anyway. MARC ANDREESSEN: Yeah, right. VINT CERF: You mentioned business models and this raises an interesting problem. Some companies build their businesses around the model, which has been very successful. Newspapers here are an example, and then the world moves out from under them. One interesting question is whether we can use these computational models to sort of help you A) figure out whether not your business model is failing, or B) what alternative business models could there be. The idea of being able to do what-if simulations of a fairly large scale model of what's going on, could be a very powerful tool and only a cloud computing environment has the capacity to do that. MARC ANDREESSEN: I'll give you an example. My friend Sam Larson has this theory, basically says historically in the economy it felt like bits have always moved faster than atoms but in reality for most, the economy atoms move faster than bits.
Why do chain fast food restaurants exist? It's because if you're a hungry traveler at 9 o'clock at night and see McDonald's sign, you know what's there. And if you just see random Joe's diner, you don't know if it's any good or not, therefore, this massive rise of chains across the economy and this massive decline of independent restaurants and bars and everything else. What he says basically is, that was just because it's actually easier to get the atoms in place in the form of the McDonald's sign than it was to transmit the information about which local restaurants are good. In the new world, it flips. It's now easier to go on Yelp or TripAdvisor, or any of these sites, or Airbnb and be able to discover it. No, I actually want to stay in this apartment, I want to go to this local restaurant, I want to go to this local bar, this local coffee shop but I'm going to have actually a better experience. QUENTIN HARDY: It's changing the consumer, too, because they can sharpen their desires better.
It's not just the generic McDonald's, it's an interesting experience they can derive. MARC ANDREESSEN: That's right. And then you get the feedback. you collect the data and so you get the data back. You get the two-way ratings on Uber and Airbnb and eBay and you get the two-way ratings, so, you actually have, if anything, better quality control than the centralized management. VINT CERF: This raises a very interesting question about the accuracy and quality of the data and in some cases, the integrity of the data. Somebody who wants to mess up your business could conceivably go in and monkey with the data, giving you the wrong impression of how your business is running. So, I think that there are gotchas that hide in this infrastructure that we're going to have to work hard to protect. QUENTIN HARDY: Who knew in the new world it's still going to be hard to be good, right? The humans. To speak to the data problem and the large incumbent problem, the big incumbents come and obviously they have enormous amount of data they want to use, and they want to generate new data, and understand their customers better, what is the decisive thing for a company?
Is it going to be the amount of data, the quality of data, the reliability of data, or is it case by case? VINT CERF: My first reaction is that all three have a role to play here and I would not want to elevate one after the other although, the quality of the data and it's reliability is terribly important because otherwise it's garbage in, garbage out. But, I mentioned earlier this notion of being able to capture things on a continuous basis. That turns out to be almost as important as all the other parameters that you mentioned because that continuity gives you a sense for what's going on in a way that you wouldn't normally get if you have a rather slow collection of information. MARC ANDREESSEN: I would take speed. I think it's simply speed of adaptation because just look at the ways you deal with data today. They're just completely different than they were five years ago. The ability, for example, to apply deep learning to data is a brand new thing and you indicated, the young companies that are growing up with this technology are just doing it by default.
The older companies that have built up data sets and methods of processing data over 30 or 40 years aren't moving that fast on it. I think the algorithmic improvements and the ability to leverage the cloud and all of the things that are happening and just the huge new flood of data that's arriving, I think it's going to be a question of flexibility to be how fast can your technical workforce actually adapt and actually do things in a better way. VINT CERF: Now we need to arm wrestle a little bit because think about rapid training on the stock market. One of the problems with a lot of those algorithms is they all follow the same thing like kids playing soccer. The ball is over here if the market's going down, they all sell. If the market is going up, the all buy. And the market goes like that, you have a bang-bang control algorithm so, you have to be careful about fast leading to instability. My question to you, Marc, is how do you balance the stability question against the speed of response?
MARC ANDREESSEN: We assume that you're taking care of it. [LAUGHTER] QUENTIN HARDY: Well that's OK. VINT CERF: OK. MARC ANDREESSEN: So it is interesting– QUENTIN HARDY: Done. VINT CERF: Right. MARC ANDREESSEN: I mean most of our companies are so small when they start that they don't have that much of an impact at all. And so, the goal is to get into a position where they actually have an impact. VINT CERF: Is this a good problem to have, you're saying? MARC ANDREESSEN: Yeah it's a high quality problem to have once you get in position where you start to introduce that level of instability. For the most part, our company's self-identity as they're up against the man. If it screwed up it's because the man has already screwed it up. Now whether they're right or fooling themselves we– VINT CERF: We can hope that everybody who comes and uses Google Cloud Service will have that problem, because it will be big enough to do that. That would be good for everyone of you.
MARC ANDREESSEN: That would be good. VINT CERF: OK. QUENTIN HARDY: This is very much an "if" question because I know you love your job and I know you see 100 companies but if each of you were to start a company now in this world, what would you go after, what would be an interesting thing to build around? VINT CERF: I know what I would do. I've fallen in love with a book on microbiology by Bruce Alvarez and his colleagues. It's in the sixth edition. It's 1,766 pages of beautiful illustrations and explanation about what goes on inside a cell. It's Manhattan in there, it's not a bag full of water with chemicals rolling around. If I were to start all over again, I would want to dive into understanding what's going on in there and then extract the knowledge, which we are gaining over time, in order to do useful things. Our Calico company has noticed that people get old and they're trying to stop that of course, the engineers have a solution. You know that's not the solution [INAUDIBLE] have and at my age, we care a lot about that.
So understanding what's going on will yield a lot of insights potential therapies and other things that I'd be fascinated– and besides the whole thing is so interesting, it's unbelievably complicated. There's even a store and forward packet switching system inside there, moving proteins around that have little labels on them to say where they're supposed to go QUENTIN HARDY: Really? VINT CERF: And other little proteins that crawl along the miton tubules. They get grown and there's Pac-Man on this side and they carry the label proteins to the place where they're supposed to go. The problem is mother nature figured out packet switching 3,500,000,000 years ago, it just has a while to catch up. QUENTIN HARDY: –you can do this thing. And you? MARC ANDREESSEN: I think there is something happening. Building on what Vint said, I think there's something happening in science that's the one layer below what we've been talking about. We've been talking about the impact of computing and cloud on technology and business.
There is a layer below what's happening in science which is what we see everyday now, which is, we see world class biologists coming out of great research universities who are also top-end computer scientists and 10 years ago that wasn't the case and all of a sudden it is. It's a generational change. We're seeing top-end physicists who are software experts, we're seeing top end economists who now have a totally different way of studying the economy because they all use software and big data in a totally different way than their predecessors did. I think computer science and software and information theory is colonizing the rest of science the same way tech is colonizing the rest of business. I think biology is actually front and center and that would be a very interesting area to go into and then I think economics is going to be super interesting, of which cryptocurrency is one example but there will be many intersection points. Then, I think there's probably another half dozen after that are equally interesting.
I'd be trying to bring computer science in other fields and then build things that there's been no way to even conceive. We have a company we just funded as an example using deep learning to do cancer biopsies via blood draws. If you talk about the long term implications not only for diagnosis of cancer but also, turns out for treatment of cancer to actually have a complete map of tumors in the body with complete analysis from a blood draw is just light years ahead in terms of the outcomes that you could generate and to us it's a software company. It's a deep learning and data company. QUENTIN HARDY: Thank you cloud. VINT CERF: There's a couple of problems here that we should probably recognize. Just knowing genetic sequences is no longer enough. We have to know all about epigenetics, we have to know about the microbiome in our gut because that's how our systems have evolved together. Our immune systems are not just human, they're also microbiome as well. So that's one issue.
There's still a lot to be done in this space. QUENTIN HARDY: OK. I want to say, by the way, we just passed the $2,000 mark and the– VINT CERF: Been keeping track. QUENTIN HARDY: We're going to go to lightning in just a second with a bunch of questions that we got off Twitter before we had this session. Very quickly though, I want to take a quick break and then why don't you tell them about the invention no one knows you two created. VINT CERF: This is where I get to blame him for everything. Back in the day around 1994, he's off with Jim Clark doing Netscape communications and I am at MCI, and we decide we're going to build the MCI mall, not mail, I did that back in '83, this is the mall in 1994. MARC ANDREESSEN: It is the Amazon, pre-Amazon. VINT CERF: Pre-Amazon and you know that well, yeah '94. So, we decided, where is the highest point of departure for this and it's obviously the Netscape Communications server and clients. So we bought $7,000,000 worth of licenses for servers and yeah, well, you're welcome.
MARC ANDREESSEN: Thank you, thank you. QUENTIN HARDY: You're on your way. VINT CERF: Then I realized that there are going to be transactions on the servers. And there are going to be broken transactions because people will just disconnect or get halfway through and decide not to do anything. I'm sitting here thinking, my servers are going to be full of all this cruft, broken transactions, and I don't want all that crap because I don't know when to get rid of it. So, I go over to Marc and I say listen Marc, I don't want all this crap. Why don't you find a way to store it on the client side and let me get the state of the transaction when I need it, when they show up and say, I'd like to finish. So, he goes and invents cookies. If you don't like cookies, it's his fault [LAUGHTER] QUENTIN HARDY: There you go. Put two great minds together and you get cookies. MARC ANDREESSEN: Strategically though it gives them a really fun name so that nobody would ever get upset.
You may not know the other side of the story so back– VINT CERF: Oh, yeah. Well, you want to tell the other side? MARC ANDREESSEN: I'm not going to tell the side you want me to tell I'm going to tell another side that you don't know. We get literally the $7,000,000 check by the way MCI our second customer. The first customer is a little bit more creative application. I'll let you leave that to your imagination. MCI was the first mainstream business customer. We get the $7,000,000 check, very big day for a start up company to get a $7 million dollar check. So as you get hurt at work or working around the clock or writing code six months or whatever goes by and we finally get all the code done and I'm like OK, thank god it's done, test is ready to go OK, send it out, get it over to them, and that's great and, I get to work the next day and we get the angriest phone call I think I've ever received because we actually did not have a process for shipping software and so the deployment engineer burned the software onto a CD-ROM and put the CD-ROM in a Ziploc baggie and shipped it to MCI.
Literally, we paid $7,000,000 and you sent us a CD-ROM in a baggie. QUENTIN HARDY: It was a new baggie. [INAUDIBLE] MARC ANDREESSEN: It was not his sandwich bag. The good news is that it was freshly fetched from the kitchen. VINT CERF: At least it wasn't in a bag of Wheaties like that other AOL guy was doing. MARC ANDREESSEN: Right. Exactly, exactly, exactly. QUENTIN HARDY: OK so, I want to move to the challenge round off of Twitter. First off, if software is eating the world and software will live in the cloud, what forms of software regulation do you think will take place? VINT CERF: I'm sorry, I didn't hear the question. QUENTIN HARDY: Software regulation in the cloud. VINT CERF: Oh oh OK. This is not a popular opinion to express. I did it when I was the president of the Association for Competing Machinery. For all you guys out there that are writing software, it probably all has bugs at some point. You're not going to get away with it anymore. You're going to have to take responsibility for your bugs.
We need better tools to avoid making mistakes that turn into exploitable bugs, but I think that's going to be the biggest challenge lying ahead. Software is surrounding everything. Now it's in everything and the mistakes will become more and more visible and potentially more harmful and so, I think that there will be pressure for some sort of regulatory oversight. Maybe you'll have to get a license in order to do major projects like a civil engineer has to do, I don't know how that's all going to evolve. But I don't think that we can get away with anymore I used to make a living writing software so I'm very empathetic about this but it's a serious issue now. QUENTIN HARDY: OK. Marc will cloud and its suppliers become totally commoditized or is there some value in the cloud chain down the road? MARC ANDREESSEN: There is a theory at foot and there are people on Wall Street who certainly believe the cloud by definition has to be a commodity because it's a bunch of servers in Iraq and anybody can do it.
My experience was that, I think from 2,000 when– AWS launched in what, '05 from '06 to I think '13 or maybe '14, 2014, every industry expert I spoke to who was associated with any kind of systems vendor box vendor, they swore up, down, and sideways emphatically angrily that it would be impossible for the cloud vendors to ever make any money. Just flat impossible because literally they're just putting boxes in Iraq and what can that be worth. And specifically, AWS there's no way like AWS has to be bleeding cash and then Amazon was either forced or chose to finally break out AWS and it turns out, of course, AWS is Amazon's most profitable business. QUENTIN HARDY: By far the best margins. MARC ANDREESSEN: And every time you buy a book on Amazon, it's being actually cross subsidized by AWS not the other way around. That caused a, I would say, rapid reevaluation inside these companies, most of whom it's now too late to react to what's happening. I think existence proof is that Amazon is plenty profitable on AWS and then, I just think this whole conference is an example of this, I think cloud is just getting started.
I think the layers of functionality that are going to come in on top is all the stuff we've been talking about, about AI and all these things. These things are just going to get more and more useful. As they get more and more useful, they're going to be able to command higher and higher prices for real value received and fantastic margins. I think they'll end up being very high margin businesses and with very happy customers. QUENTIN HARDY: OK VINT CERF: Marc has just used a term which I'd like to underscore the layers in the architecture because what we're really involved with now is the sciences of the artificial. We're creating an artificial world that we can manipulate and change and experiment with and that's just going to continue so that means more opportunities for new businesses QUENTIN HARDY: OK. Two edge core questions. Utility at core and power at the edge. How does that dynamic play out you think? And likewise, we talked about death of cloud you don't see that coming at all.
But, are we moving towards an increasingly powerful edge or does the core become a stronger and stronger repository? Do they kind of chase each other, the way they always have? VINT CERF: The pendulum swings back and forth. We've seen it happen many times. Now we're in the same place where we were in the 1960s. People would show pictures of giant buildings called the computing foundry with smoke coming out of the top and then everybody went to the private machines and departmental machines and things like that and so it keeps swinging back and forth. When you start thinking about the very low power things that we use for mobiles and other related IOT devices, those may end up in the computing environment of the computing centers as well simply because they are not power hungry. I don't think the clouds are going to go away at all I think what will happen is the edge is going to keep getting more and more and proliferated but the clouds are going to have to be there in order to make things more coherent QUENTIN HARDY: OK.
What kind of structures do you think could become fundamental technologies, like DNS did? VINT CERF: I'm sorry say– QUENTIN HARDY: What kind of fundamental structures do you think might become important in the future? VINT CERF: Probably the most fundamental one has to do with the semantic modeling of what's going on. The thing which makes AI powerful is its ability to build models that are analytic. Right now we don't do that very well. We have neural networks whose function we don't understand. If there were a powerful addition to the computing environment that we could point out, it would be the ability to extract a model of the real world. for example, or a model of an artificial world like a company and then try to understand how to reason about it. We don't have a vocabulary for that yet, let alone grammar, so, if there is a breakthrough somewhere, that's the one that I think would find the most interesting. QUENTIN HARDY: OK. Last question for the both of you.
Biggest regrets and proudest moments in your career? VINT CERF: I'm sorry, the– QUENTIN HARDY: Biggest regrets and proudest moments in your– VINT CERF: Wow. I wish that I had used IPv6 instead of IPv4 to begin with because– QUENTIN HARDY: I meant to bring that up. VINT CERF: Who knew. And we probably could have done a better job of using public key crypto except that the timing wasn't very good. The paper that was published in '76 didn't include an algorithm. Then RSA comes along a little bit later just as we freeze the internet design and I want to get it implemented so we can figure out if it'll work, and so I put off the public key stuff for a while. If it had been at the right time, we would have included it except for one thing, the internet was built by graduate students and you were all one of those once upon a time. They're the most undisciplined people in the world and getting them to use crypto properly would have been impossible anyway. It's too bad but that's an alternate history I think.
MARC ANDREESSEN: No regrets. QUENTIN HARDY: You're shocked right? MARC ANDREESSEN: I'm from the Luke Hayes school, always forward and never backwards. Proudest accomplishment I got to say, the people I've been able to work with and certainly Vint is one of these but just the people I've been able to work with. Especially now, younger people coming up in the industry and the ability for people to be able to realize their dreams and create new things and be able to learn. I think we're all in the process of learning and growing all the time. And I think our ability to help each other and teach each other in the end, is the most important thing and the most rewarding. QUENTIN HARDY: That's a really great note to end on. I had some more questions but we've run out of time. This was fascinating, high value and we ended up over $3,000 so, thank you so much. MARC ANDREESSEN: Thank you. Thank you, Quentin.
Join two Internet legends for a discussion moderated by Quentin Hardy (formerly of the New York Times) on the evolution of enterprise computing. Starting with stories from the early days of TCP/IP and the browser, the panel will then go on address key strategic questions about computing: What lessons did we learn about business adoption and computing trends from the ’90s that are relevant today? When thinking about today’s shift to cloud computing, is there a “right” or “wrong” way for a company to embrace the cloud? What trends might surprise us in the near future?
Missed the conference? Watch all the talks here: https://goo.gl/c1Vs3h