 So hi, we here at the narrow connect here 2018 in Vancouver. So who are you? I'm Chris Benson I did the initial keynote to introduce neural network and AI technologies so a Lot of people are talking about AI. So a lot of things are happening right now. That's true So what's going on? Well, it's a it's a fantastic set of technologies that is maturing very very rapidly It has has been noted in several of the speeches. We're still in the very early days but the the It is just transforming what the world of computing is is being able to do at an exponential rate and so literally month to month to month We're having these fantastic new changes coming out and there's a lot of new things to learn all the time So who's best at doing AI is it Google Amazon or Tesla or Nvidia I don't I know it's not me as being the best at doing that. I think there are a Lot of great companies doing a lot of great work, and I really I don't really think of it as who's best I really think of it is probably who is serving their customers best with these technologies and in terms of impacting their customers lives Because for example Google has a slogan. They used to say mobile first. No, they say AI first. They are that's true I actually talked about that in my keynote as they kind of because from a popular standpoint they kind of led the way and You know we're noticeable in the popular media for that initiative a couple years ago And so I think you know where they lead many other companies have followed it's not just them taking on New buzzword and and using it at like to satisfy their investors, right or something It's actually really using it a lot. Oh, they're doing amazing work I have a podcast called practical AI and I interview people. I've interviewed Google Brain people and And from other teams and stuff and the work that they have been doing is just amazing one of the just Is a very small project. They did it was kind of an on-the-side thing almost They were contributing to astronomy classification of new planets In the universe that they were able to detect and because they were able to process the information Much much better than humans had before that or other machine learning algorithms So there are just so many use cases that Google and many other companies out there are utilizing this technology They're doing something good with cancer research Yeah, their deep mind has has been doing cancer diagnosis over the last couple of years. I believe But there are That's that's one of many many of the applications in healthcare that are being applied Neural networks are going to be so embedded in healthcare going forward to where you will you will not have healthcare without the technologies in the years ahead So so it's really working. It's not just some little what's it called a project on the side. Oh, no. Oh, no I mean, there are Many many billions of dollars being invested in these technologies something that I didn't mention in my own keynote But the US Department of Defense just allocated two billion dollars for AI research And that's just one budget. That's one budget of many And so and if you look at the you know some of these top companies such as Google Apple Amazon Microsoft that are working in this space they have gigantic budgets for for AI initiatives and they expect results They're not going to allocate such such amounts of money unless they're expecting to get a return on their investment on that And that's and you know, we've only talked about American companies or at least companies that originated in America They're the Chinese government is investing many billions of dollars into AI research The Russians are there is I've talked to Israeli AI companies lots of companies in Europe It's it's a global it's a global movement at this point in terms of driving value to customers through AI What is Microsoft doing? Well, I you probably have to ask them for detail because I don't work for them. I have They're doing some pretty amazing things with their framework And there are certain use cases that each of the frameworks are stronger in but I'm gonna defer the question to somebody who would know better than me and Amazon is doing all their They're Depend on a lot of the internet sure That's a great point. So all of it. So whether you're talking about Amazon AWS or whether you're talking about glue Google cloud platform or Azure by Microsoft all the cloud platforms are Providing different services around AI technologies now at different levels to where anywhere from Renting GPUs just by themselves and you have to do all the work yourself to to specific services That give you a full offering as a data scientist or engineer to be able to go do your work entirely on their platform with all the tooling so you can kind of buy in to that at whatever level that you want and Did you on your podcast did you talk about the Elon Musk house care use of AI well We do actually taught and we actually I have mentioned Elon Musk specifically We do talk quite a bit about different perspectives. I think AI is is a technology that's a little bit unlike others and that in the years and decades and even centuries ahead We have some careful steering to do on where we want to go and you have people like Elon Musk being probably the most notable That is very worried about the downsides And you also have a lot of advocates talking about the upsides I tend to to stay in the middle and give both sides each time Is it possible? I mean my little kind of like understanding of the Elon Musk angle is that maybe he doesn't really like Google and He's warning people not to let them have too much power in this field. So or so that so You know full disclosure. I don't know Elon Musk. I've never talked to him I've just read and you know get his tweets and such like everybody else I believe that he's very genuine in what he's trying to say and I think that there is Merit in being careful with AI going forward Because as with any new technology that is powerful enough there can be great upsides and terrible downsides and My position is probably we will do amazing things with AI in the years and decades ahead And we will probably make some awfully big mistakes also And have to deal with that just as we have with other technologies So my advice in that realm is you don't necessarily need to take sides of it's good or it's bad I would just say we all need to be thoughtful and careful on behalf of our children and grandchildren that we make good decisions today Usually humans are pretty good at just being careful a little bit, you know, and they're doing amazing things We're still being careful, right? Well, I hope so. It's we're dealing with the technology that is In today as we as we record this in 2018 it can be very good at solving very Specific complex problems, but they're fairly narrow in scope in the years ahead as new Technological advancements that we don't know what they are evolve It may be that it doesn't it's not so narrow or we may be able to combine many many narrow models together to do more generalized problem solving and and therefore We need to be cognizant of how that will evolve over time And be smart about how we approach it and another thing Elon Musk is doing is the open AI. Is that a good thing? Is that working out? I think so so and I've interviewed some with well actually my podcast co-host has interviewed an open AI person I think that that is one of a number of organizations where that conversation across organizations is fantastic And we need more of that We need both the engineering the data science side and as we've been talking about here a bit the ethical side It needs to all be part of that same conversation and we need to have all those voices part of it I think that the engineering The business and the ethical all need to roll into one to make a good decision going forward and actually a Little bit more about Tesla Elon Musk is you know They ship all these cars with the promise that eventually they'll be able to activate self-driving mode So hopefully they'll figure it out, right? This is another kind of AI that's kind of Has to run locally in the car and kind of work out just based on the cameras sure and in Tesla is one Nvidia also is doing the same stuff as Google As is a number of others at this point. It's a it's a fairly It's a space that's expanding rapidly my personal belief is that we will have some challenges Going ahead, but I also believe that those companies will get that right I think it is highly probable in the years ahead that Autonomous driving will be orders of magnitude safer than having humans that are not communicating among themselves Making their own decisions as they go forward. So I think I think though we will have some bumps in the road to get there I think autonomous driving will make transportation Many orders of magnitude safer than it currently is it sounds like Google is saying that they already are safer than humans, right? so I think statistically they are and I think I think that all of the All of the companies that are in this space have established that but you know all it takes is one terrible accident where one person dies and that is a Deep tragedy so I want to I want to note that there's the humanity of recognizing that when you make a mistake It can cost lives and that's that is horrendous If you're looking at it as a mathematician though from a statistics standpoint then then yes Then if you look at the number of people who die on a given day Like here in the United States and compare that to the number of deaths from autonomous vehicles statistically It's already a much safer thing and I think that's only going to increase over time So for example uber had one accident. They can't cancel the whole program. Yeah, and I think that's that's one of the challenges I the the lady who died in that You know that she was jaywalking, right? I'm not yeah I'm not familiar enough with those circumstances, but that I mean that is truly a tragedy as is every death And and and you know independently actually when I'm not talking about AI I'm talking about you know rescue and animals and so I'm very conscious about trying to save lives That's a big part of my life. And so we really need to recognize the tragedies when they occur Work in every way possible to try to prevent this from happening But I actually think that this technology will contribute to life-saving in a much greater form than it would Cause death so I actually I'm a big fan of autonomous vehicles But I think we will have a tragedy here and there on the way But probably far fewer than if you just go out on the street of a city on any given day and the number of traffic Assets we have maybe the lesson that could be learned from that is you know This whole lawsuit that was gone between Google and uber about their self-driving technology. Yeah Maybe it's because Google they want to share their awesome Maybe higher quality self-driving technology to uber That's maybe the reason that this actually happened because possibly the Google AI would not have had this issue Maybe so I'm not gonna get in and litigate that I'm not familiar with that particular issue there I'm gonna recognize that it's unlikely for any particular for-profit corporation to share proprietary technology Across but having said that we do have a common ethical concern across Organizations to make sure that these technologies don't cause harm much like doctors start with that do no harm Saying I think that we really need to do that with autonomous vehicles, but I would be very very surprised If proprietary technologies were shared across companies, but this is what happened in the keynote after your keynote, right? So arm is talking about they want to share a whole bunch of software sure code for their new network system Yeah, I think and I think there are places. This is very typical when you are doing standards work and that is that The communication between different functions and how that workflow works There is an incentive So that you make it less painful on your customers who are using your platform to be able to have common way So they can only they're only required to learn one way of Making those workflows work But having said that the workflows are working between components and I would fully expect those components to have Proprietary vendor specific stuff in it, but they're basically abstracting that away through these standards bodies So I do think standards in what Lenaro has I and I can speak myself as a practitioner in the field What Lenaro has announced today? I think is fantastic I really wish they had announced this You know it had been able to two years ago because it would have saved me and my team so much pain What kind of ways because as you as you want to move out onto the edge you can find some It's not as hard to standardize if you are in a centralized cloud environment or something like that You can apply whatever framework you want in whatever platform But it but historically up until now as you've moved out onto the edge And you have different vendors with different devices Different arm processors that have different Where acceleration is being implemented in different ways and it has different support for different frameworks You notice I keep saying the word different and that means that you get out onto the edge And there's so many choices and for a business that's trying to Get these technologies out onto the edge They quickly get pinned in with a particular vendor's approach And if they're going to expand across multiple channels Then they are gonna have to learn a particular deployment approach for every one of those and that's the pain I've experienced myself I think what Lenaro is saying with this announcement going forward is why don't we all agree as an industry to push outward in a unified way And and and have that support be unified and there's still plenty of room for proprietary competitiveness within that without making the customers life difficult and that's what I Definitely applaud with what they've announced today because maybe a little bit what the company is like Google for example I'm thinking is that if they have their best AI they might be controlling the world the future Like if this becomes a big deal everybody needs to use it Then suddenly is this the key to kind of like become the next hundred trillion dollar company or something potentially I don't see this is one party or even just a few because they're so I think what's different now in the AI industry Versus ten years ago now is ten years ago There were only a handful of players in the space And that was Google and Amazon in earlier days as they were kind of moving into this and that Ten years might even be too far to say that but what's happening now is there's so much Business value that's being sought from these technologies and so much investment to achieve that from different organizations That you don't just have you don't just have a handful or even dozens or even hundreds You have thousands of companies with substantial investment in AI at this point and that is only going to grow So I think what you're going to find is different layers of capability And different amounts of scope if you're if you're a Google or Amazon or Microsoft You're able to handle many many different use cases If you're a smaller company you may be just tackling one or two or three use cases But there's many more of you that are out there doing that So it's a it's an industry that is growing exponentially right now So do they have to standardize and collaborate on the algorithms or on the protocols or some well? They already they already are there's there's the concept of transfer learning where you have what are essentially libraries of Algorithms or models that are available out there and model architectures that are available out there that you can utilize and So you are you can already go out with TensorFlow And there there are a number of different TensorFlow examples for different use cases And so rather than me having to sit down and start from scratch I can take somebody else's Architecture and then say and then try it on my model I may not get results that I want and say well How can I how can I move this? Architecturally toward what I need and that's where probably most of the work will be in the years ahead Not everybody is a research scientist in AI that that's a fairly select group I think 99% plus of the people that are working in this industry are going to be the engineers or what you might call a data developer who is taking going you picking a framework and Taking the different types of examples Available within that framework and through the through the wonder of transfer learning being able to pull that in and then tweak it Into what they need make the adjustments they need to get productive quickly So so what's next with your podcast or with the what are you doing? What what are you planning to do? There's just so much that's gonna come you don't need to plan it out To be honest, I'm not entirely sure the world is moving so fast right now a big part of my life is is coming to Organizations like Lenaro and this Lenaro connect Conference and talking to people about these technologies just like what we're doing right now. I and through the podcast We'd like to bring in luminaries in the AI space Whether they be on the business side or the technical side as well as just practitioners that are out there actually doing cool work But that nobody may know of them and and we like to highlight those and give people the taste of how you make AI Practical and accessible for anybody out there. So how many episodes have you had so far in the podcast? We're fairly new we've had I think there are 13 that have been released and we always are working on several episodes In the funnel so what kind of examples do you have in those 13 episodes? pretty much everything we are we have a series of Of what I'll just refer to a senior executives at some of these key AI companies That are gonna be coming on in the days ahead We've actually had some In the weeds Technologists that are implementing some really really cool stuff come on and explain how they did it So if you are for instance a small business owner and you say hey I'm willing to invest in on some computing power in my cloud of choice And I'm gonna pick one of the frameworks, you know, maybe TensorFlow pie torch Keras Whatever you want to do and I'm gonna get in do it our podcast tries to make it a Real engineering thing instead of a pie in the sky Initiative and so if you were that person who wants to get in and understand what the realities are both The good and the bad the constraints and everything that's the audience that our podcast is trying to address