Empowering Consumers: Evolution of Software in the Future

Updated on June 6, 2023
19 min read

In this captivating GOTO Unscripted conversation, Linda Stougaard Nielsen, an accomplished data scientist and agile practitioner at AVA Women, joins forces with Derek Collison, the visionary Founder & CEO at Synadia Communications. They embark on a thrilling exploration of cutting-edge topics, including the path to personal data sovereignty beyond GDPR, the quest for innovative AI systems that break traditional boundaries, unlocking the utility paradigm in cloud computing's exciting future, the shift from developers to empowered consumers shaping the future of software, and their fascinating adventures in programming diversity. Prepare to be inspired by their insightful and thought-provoking dialogue!

In this captivating GOTO Unscripted conversation, Linda Stougaard Nielsen, an accomplished data scientist and agile practitioner at AVA Women, joins forces with Derek Collison, the visionary Founder & CEO at Synadia Communications. They embark on a thrilling exploration of cutting-edge topics, including the path to personal data sovereignty beyond GDPR, the quest for innovative AI systems that break traditional boundaries, unlocking the utility paradigm in cloud computing's exciting future, the shift from developers to empowered consumers shaping the future of software, and their fascinating adventures in programming diversity. Prepare to be inspired by their insightful and thought-provoking dialogue!

Introduction  

Linda Stougaard Nielsen: Welcome to another episode of "GOTO Unscripted." We are here in Copenhagen at the GOTO conference. My name is Linda. I'm sitting here together with Derek. I just had a very interesting experience. I went to a talk with Derek, and I would really like to know more about what you were talking about. First of all, you talked about Connectivity 3.0. What is the first two versions of it?

Derek Collison: I think the first two versions were kind of the early versions of the internet, the IP protocols. And I think version two was, everything was HTTP. All right. And what we inherited from that were some of the things that I feel got us to where we are, which I think is a great thing. But also potentially if we don't pay attention, can hold us back going forward. Things like, everything's a one-to-one conversation. Everything's asking a question and getting an answer-back. And everything is location dependent.

I think, when we talked about the fundamental pillars of how we want to reverse course on those, those are the ones that I see potentially, holding us back as we try to move into a world, in which, none of us know what it's really gonna look like. All I can tell you is from my 30-plus years in the industry, I can tell it's going faster than we've ever seen before. And it has a level of uncertainty that we've never seen before. Meaning that in the next decade, where we end up I think will astound everyone, to be honest with you.

Linda Stougaard Nielsen: Yeah, I agree.

GDPR and Beyond: Charting the Path to Personal Data Sovereignty

Derek Collison: And I know that, Linda, you're in data sciences, and, we talked a little bit in the talk about this notion of collecting and analyzing and drawing, insights from more and more data to get better and better at answering the questions that a lot of these systems are doing as they interact. And I'm interested in your view on how fast do you think the world is progressing? And again, it can sometimes feel like a frog in a boiling pot of water. As you're in it day to day, it feels slow, but then as you take just a little bit of a step back and look at like where we were just a year ago or two years ago, at least from my perspective, it's pretty mind-boggling.

Linda Stougaard Nielsen: I mean, if you're talking about algorithms, then of course, that's happening something every single day, and we get more and more data. And the possibilities get larger and larger. And of course, as with big possibilities, with big data comes also big responsibilities. It's also...what I've noticed this happening in the last couple of years is actually people getting more and more aware of data privacy. And that is also something that you touched on a little bit in your talk about security because that was something that I was a bit interested in how that actually happens, because you said that it sort of takes a backseat, the security because you put the security somewhere else in the system, not so much in the protocol, but rather somewhere else.

Derek Collison: I didn't mean to infer that it took a backseat. I actually tried to say almost the opposite, meaning it's secure by default. You don't have to do any unnatural acts to secure the system. Now, what you're talking about is, who owns the data and where's the data allowed to move, right? GDPR within Europe, there are some provisions in California, I think are gonna continue down that path, for sure. But no, what I meant was that things are secured by default, meaning that, it just works that way. Now, where you have to potentially lean in is around zero-trust constructs. And I think that comes into play as the desire for larger amounts of data and diverse data to train these systems is gonna start needing to become more dominant from multiple providers, not single providers.

Even the behemoths of the technology world like the Googles and Apples that have such a tremendous amount of data, I believe that for them to continue to become more intelligent on some of these systems, they're gonna start needing to partner with folks. And during the talk, we even talked a little bit, I made fun of myself about Netflix probably knows more about my sleep schedule than anyone. And right now, all of these companies that we interact with, these technology companies, they use that data. And I don't think it's for nefarious purposes, at least by default. But there's an assumption there that they have the data and they can use that data to make a better service, to kind of do that flywheel effect with customers. And I think that's gonna come to an abrupt end.

And what I try to think about, and you might have better ideas than I do on this, is how does my mother who's, getting into her late 70s, understand and rationalize that the data she's generating as she interfaces with Netflix or whatever streaming platform or any technology is hers? And how can she be solicited to share that in a way that's meaningful to her? For example, Netflix might say, "50% off if we can see generic data about where you are, kind of the genres that you like," and things like that. It's not so much the younger generation, I think who will understand that. But everyone needs to kind of understand, hey, every time I interact with a system, I'm generating data. And I believe regulatory within the next probably less than a decade, it will be mine, it will be yours. And then it's your prerogative of whether or not you wanna share it to get benefits or incentives or whatever like that.

Linda Stougaard Nielsen: I think also we are moving in that direction, and we have seen it with the GDPR and also systems where you actually have the data stored on, for example, on your mobile phone instead of it going to a server and being stored there are becoming more and more popular by users. So, I think it is moving in that direction.

Derek Collison: It is. And I believe GDPR is a little bit of a blunt instrument, but I don't think we're gonna go backward.

Linda Stougaard Nielsen: No.

Breaking the Frames: The Quest for Innovative AI Systems

Derek Collison: Right. We're gonna keep going forward. How do you feel about, I guess, the field of, today I would say it's termed more narrow AI, but like this race to however you define AGI? From my perspective is more of an average type class of worker, an intelligence could actually take on that type of role. Which camp do you sit in? Do you feel that it's within the next couple decades or you still think it's 50 to 100 years out, or...?

Linda Stougaard Nielsen: What are you talking about this concept of a sort of citizen data scientist or..?

Derek Collison: Yeah, artificial general intelligence. Where there is an intelligence that can be trained kind of similarly to our brains and can mold itself to different types of inputs.

Linda Stougaard Nielsen: I think it's still far away, to be honest. What usually...I mean, we can still not create anything that's smarter than us. I mean, it always depends on the data that you put in, you get exactly the... If you put good data in, you get good results out. If you put a really bad data model in, if you don't cover your whole...

Derek Collison: Yeah, it's interesting. I've vacillated back and forth, and I've always had a thought about it, I just wanna hold on long enough to be able to see it. And I'm getting up there. But over the last 10 years, especially watching DeepMind, I think we're gonna see something that resembles it within two decades. I really do. I could be wrong. I'm wrong about a lot of things. But my sense is that just even watching the notion of, AlphaGo and learning the program of Go, what I think people might not realize is that, one, you can't brute force the game. You have to have some level of strategic intelligence to play, and the players, have built-up careers. It's very hard to be, a good human player.

But that's not the interesting part. The interesting part to me is that generally accepted rumors were that it was about a room full of specialized hardware to beat Lee Sedol. And less than two years later, it was down to a small cabinet size. And all of a sudden it wasn't only beating Lee Sedol, it was beating, everybody and all the chess players and stuff. And then all of a sudden, we had another massive order of magnitude in reduction of the hardware cost all of a sudden beat every chess game played, every Go game ever played. I think it played another, video game, which apparently is very hard. But what happened was, is that my understanding is, that the hardware to run that was all of a sudden fitting in the palm of a hand.

People don't see that part of it. They see the, oh, it's getting better at stuff. But when you look at something like that, that boggles my mind.

Linda Stougaard Nielsen: But still, I mean, even if it is getting smaller, the technology, I mean, I still think when we are training artificial or neural network or any model, we sort of have the frames. And yes, the more data we have, we can train it within the frames. But getting as a neural network or artificial intelligence to go outside of the frame and do something different to be innovative, I think that is not something that we can achieve very easily. And I still think that there's a way to go there. Because if you're talking about playing a game, you can show it a lot of different strategies and then it will learn to adapt within the frames that you give it. But going out of the frames and doing something that we can't understand or that we haven't trained it to do, I don't think we are there yet.

Derek Collison: Well, I would agree we're not there yet, but we're further than people think. The last Alpha system out of DeepMind was Tabula Rasa, meaning it had no predefined rules. It's kind of like the laws of physics, but they are baked in the laws of whatever. And they also came out with a model just a little while ago that the same model can play chess, play Go, and also drive a physical robot to do menial tasks. And it was never trained to kind of do any of those things. And I also think, you being in the space, you probably have...you see it closer than I do. Still, I do believe there's going to be a couple core algorithmic breakthroughs and then the multimodal algorithmic breakthroughs that I think it's gonna be like click, click, click, and all of a sudden we're like, "Uh oh. You know, it's here." We'll see.

It could be that, I'm totally wrong, but my sense is, is that, as long as I don't get hit by a bus, I will be able to see it, in my lifetime, which will be interesting to see.

Linda Stougaard Nielsen: Yeah. Maybe. Maybe.

Derek Collison: The other one that's interesting, and again, you probably have more insight than I do is, we're also starting to see a lot of folks switch their energy towards sparse networks. The guy that created the Palm Pilot, wrote a book in the mid-90s, I think, called "On Intelligence," Chef Hawkins. And it was really awesome. And now he's been pushing on sparse networks, less data, because people look at the human brain, they're like, "Well, maybe it does have all this data, but it seems like toddlers use a lot less data to form, a mental model and what looks like a level of sophistication around something."

Linda Stougaard Nielsen: But that is something that you're actually using to train neural networks is to remove some of the data and you actually often get better results, which is quite interesting.

Derek Collison: Because it doesn't fall into the local minimum and things like that.

The Exciting Future of Cloud Computing: Unlocking the Utility Paradigm

Linda Stougaard Nielsen: Well, I think actually I want to return to something you talked about today because that was something that I really got a little bit excited about. You talked about that we should look at... For example, I've been working with different cloud providers. I've been working with Google Cloud, I've been working with AWS, and they have all their different ways of working, and you have to spend a lot of time if you're moving from one to the other. And then if you have a system using both, I mean, that's almost impossible. But you talked today about using it as sort of, you said utility, right? I think that is a very exciting idea. But, I mean, how far are we away from something like that?

Derek Collison: That's a great question. I think what's happening is that, and I just actually attended a talk right before this meeting where the person was talking about the way boomers and Gen X, I guess I'm Gen X, approach problems and software, architecture, engineering, whatever you wanna call it, versus what he was calling kind of the Zoomers, the ones that are even still in school. And the way they approach things or his opinion is kind of aligned with where I think things are going, which is either through regulatory concerns, or financial pressures, meaning you want to get a better price on running some analytics over some data. People want to have a choice across multiple clouds. But the cloud providers, they've invested a tremendous amount of money and time, and they want to make customers happy, like, everyone does, but they also are incentivized to make their offering sticky, so it's hard to leave, kind of like the Hotel California, you can check in but you can't check out.

However, again, because of, either it's choice pressures, financial pressures, or regulatory, like, fin services in a lot of Europe now are being forced to be multi-cloud by default. They can't just be pinned to one. It's starting to have people look at certain services that could be very helpful to them but would pin them into a certain club or writer and they're starting to either not select them or trying to move away from them.

Linda Stougaard Nielsen: I mean, you also see these services that just operate on top of a cloud, so it makes everything easier.

Derek Collison: And you're seeing a lot of, at least in my opinion, a lot of software startups are trying to take advantage of that. See the patterns that have worked in the past, abstract those patterns out to be, either multi-cloud or what I call edge wear, in the talk. I mentioned a prediction I made about four or five years ago that Edge would dwarf Cloud by two orders of magnitude within 10 years, and we have four years left. And with any exponential technology, 50% of the way through, you look like a failure. You look like you've totally missed the ball, because it starts out dead flat, and it kind of goes like this. But I really do believe that most people even today are interfacing not in terms of maybe building software and working with cloud providers, but consuming technology. They're interfacing with whatever you wanna define edge, not necessarily, origin servers that are running in cloud.

The Future of Software: From Developers to Empowered Consumers

Linda Stougaard Nielsen: But that's more the consumers, right? We're talking about software developers here instead. But you think of that also being abstracted away so that software developers also become consumers in the way that they're just using a service and you don't even know what's behind it.

Derek Collison: I do. And, to kind of illustrate that point, what I'm trying to describe is, and again, I've been around for a while. I remember that, you would go into a bank and you would meet with a bank professional or a teller, and they were driving the system and it was, windows machine or NEC box and X, Y, Z, and I can't remember last time I've even stepped into a bank. And when you transition from someone who works at the bank and could be trained on a specific set of software that they might go to training for, and it has its quirks, but it is what it is to, millions upon, hundreds of millions of consumers with an iPhone or an Android phone in their pocket, you see the shift.

Even though the software, backend software developers for the bank might be doing things that feel very similar today as they were doing maybe 15 years ago, the system as it's being consumed has radically changed. Even the notion of, I call it the spinning wheel, and then in the talk, we talked about the human psychology of our brain switches at about 160 to 200 milliseconds, we flip. And that's when we realize something is slow when our brain starts concentrating on something else. And  even the notion of, doing a deposit on your phone, if that thing spins for over a couple of hundred milliseconds, your brain says, "Wow, that's slow." You know what I mean? And there's this voracious appetite to, for lack of a better word, not to context switch the end recipient's context in their brain.

And I think that's driving a very different way of thinking about it. And again, if let's say the ability to process that transaction, scan the picture, validate it, all other stuff, let's say all of the backend stuff, ran in let's say five milliseconds. It was perfectly designed. Linda designed it, she's one of the best software architects we've ever seen in the world. But if that software only runs in Zurich and I'm stuck in Los Angeles, California, in the United States, my experience is, that could have taken like 280 milliseconds, depending on all network traffic, how to get over there, how to get back. And you're starting to see this notion of, we can design it and it can work very well, but we now need to replicate it and spread it out.

And a lot of these systems can't operate in a vacuum. They're not like math, where it's like two plus two. They need a data set to actually interface with, to collect data, and massage that data to return the response. And that's during the talk I was trying to say, they're kind of coupled at the hip and they need to be able to move and flow in kind of a relationship pattern across clouds and edge and whatever you wanna define it as. It's very fascinating to see how software development is still evolving at such a rapid rate. Even things like CoPilot, which, is open AI's attempt to scour all of GitHub source code and essentially you can kind of say, "Oh, I want this type of data structure, I wanna function that does this," and it's gonna try to write it.

And when it first came out,  a lot of people were skeptical and a lot of people were laughing a little bit at it, even like the other day at AI Day for Tesla, they were laughing at the AI robot, and I tweeted, "Look, be careful because what you just saw is the worst it will ever be, and it's gonna continue to get better tomorrow and the next day." And when you laugh, what you're really saying unconsciously is you're betting that the team behind it can't make progress that's on par with the general industry. But if you think they can, and they've proven that they can in the past, instead of laughing, you should lean in. And it's a tough concept to wrap your head around. But yeah, every technology we see for the first time is the worst that really will ever be.

And it's a different way of thinking about it. But if you kind of go, "Wow, what do I think they'll pull off in a year from now? You know, what do I think is gonna be in five years from now?" And of course, with the pandemic, at least in the states, we call it COVID time. And it'd be, "Oh yeah, that was last year." And it's like, "No, that was three years ago." Because everything kind of was getting waffled. Even a year from now is very fast. So, Linda, in the world of data science, what were the first tools or programming languages you used, and going forward, what are you looking forward to? What do you think's gonna revolutionize that space for you?

Language Quest: Linda and Derek's Adventures in Programming Diversity

Linda Stougaard Nielsen: Well, if we start with what did I start with, I actually started with something simple as Pascal, and then I moved to R, but I quickly realized that R was way too slow, so I had to start programming for real. So, I moved over to C++.

Derek Collison: Oh wow.

Linda Stougaard Nielsen: But since then, of course, we have gotten Python and I'm actually very happy with Python. And when I first saw Python, I was not a fan at all. I thought it was too simple. And you can make a big mess out of it, and yeah, you still can. But it's really, really powerful, also because you can reach a lot of people who otherwise would not be programming.

Derek Collison: Do you think something will replace it? I mean, because I know the basis underneath of it is like NumPy and Pandas that are usually optimized over a very low-level fast C level or C++, libraries.

Linda Stougaard Nielsen: I don't know if anything will replace it soon because I think the power in it lies that you can use it in two different ways. You can use it both as a research tool. Where a lot of research still uses R for statistics, for data science, but a lot of them have moved over to using Python in the same way in a notebook format. But you can also use Python, exactly the same language you can use to actually do a real program, something that you put into production. I think that's a big strength of a language that you can actually do both things. And we well know that researchers don't move that fast. If they hold on to Python, I think the rest of the data science, data engineering community will have to do that as well.

Derek Collison:. Last question, at least from me, TensorFlow or PyTorch?

Linda Stougaard Nielsen: I like TensorFlow. Well, using Keras.

Derek Collison: Yeah. Yeah. Exactly.

Linda Stougaard Nielsen: Not TensorFlow role.

Derek Collison: Right.

Linda Stougaard Nielsen: What about you? Have you done any machine learning neural networks?

Derek Collison: I have. I was part of the first coming of the AI winner in the mid to late 80s and university. But I got my start when I was 11 and basic, but did Pascal, Fortran, C, and Assembly. I think the last check, I'm at 40 in some languages. I try to learn a new one every year. Last year was Rust and this year... Actually, the creator of this language is giving a talk tomorrow, which I'm gonna attend. And he and I are both going to South Africa in February for a conference, but it's called Zig. And it's one that's really got my attention, so I'm interested to hear him talk. And I've already started on a client for the systems that I work on in Zig. Looking forward to that. Thank you, Linda, for the time. I appreciate it. Looking forward to your talk tomorrow.

Linda Stougaard Nielsen: Thank you. And thank you for your time and it was very nice talking to you.Derek Collison: Thank you.

Related

CONTENT

Unraveling Insights About Places with Deep Learning Computer Vision
Unraveling Insights About Places with Deep Learning Computer Vision
YOW! Melbourne 2023
Web3 Beyond Blockchains
Web3 Beyond Blockchains
GOTO Chicago 2023
Abzu’s Pioneering Technology, the QLattice®, Introduces a New Standard of Interpretability to Artificial Intelligence
Abzu’s Pioneering Technology, the QLattice®, Introduces a New Standard of Interpretability to Artificial Intelligence
GOTO Copenhagen 2021
How we Built Google Tulip by using  Serverless Technology and Machine Learning
How we Built Google Tulip by using Serverless Technology and Machine Learning
GOTO Amsterdam 2019