Welcome to NVIDIA's AI podcast recording from the floor of the 2017 GPU Technology Conference, a gathering of the AI faithful here in Silicon Valley.
Here's the question. Are AI and VR the peanut butter and chocolate of computing?
Help us answer that question. We have Michael Ludden, Senior Project Manager at IBM Watson.
That's him cackling there in the background.
And he's eminently suited to answer the question.
Michael heads up IBM Watson's AR and VR labs.
You can see the connection right there. and is one of those technologists who gets to let his mind travel far into the future, not just in the realm of VR and AR, but things like conversational interfaces, IoT, and because he is part of the Watson team, pushing the emerging field of AI as a service.
Michael, thanks for joining us. Now look, AI and VR are two of the most creative, interesting, and fast-moving fields of computer science.
That said... Do they have a match made in heaven?
Absolutely. They're star-crossed lovers.
They're destined for one another, to change the metaphor a little bit.
So, yeah, I think that it's inevitable, right?
I mean, it's very easy to be, quote unquote, among the AI faithful at the moment, right?
Because it's sort of eating. If software is eating the world, then AI is eating software, right?
It's the hot buzzword right now. Machine learning algorithms are starting to pop up in every product from your photo app, from Google to, you know, That's something you're doing on Evernote or whatever.
It's just very useful. I think that, of course, it's inevitable that VR, being a brand new computing platform... is going to leverage AI natively for certain things.
And what we're trying to do at IBM, and especially as part of the ARVR Labs initiative that I have on my team, We're trying to provide tools for developers to address those sort of like green field, blue ocean, whatever the metaphor is, opportunities.
And the first one that we launched is around interactive speech interfaces in VR, but we plan to do more around data analytics and visual recognition, stuff like that.
Now, VR, we know, and hopefully we know, most of us have experienced it.
I mean, it's got a sort of high compute requirement and it's doing a lot, right? right?
It's got an engine that you need behind it to make it work and to make us all enjoy it or be scared by it or whatever.
But how does AI then help that problem? How does it do kind of what it does best within the VR world?
Yeah, it's a great question. So I think that's actually one of the really fun, interesting things and why I started this initiative is because as a cloud platform, Watson and other AI platforms that provide access to machine learning algorithms via REST APIs actually reduce the computational load that a developer has to build into a locally based application.
Because at the end of the day, you're just making an API call to a cloud server and and all the heavy lifting of parsing, for example, in the case of interactive speech, what a person means, what they're saying, what their accent is, all that kind of thing. is not handled by the application.
So it actually doesn't include a load. And then, oh, by the way, it's scalable because it's a cloud platform.
So you can build it once and then you could potentially do a build to a different VR platform and it will work just the same.
And it sounds like then the issue of like, I need a special rig, I need lots of equipment that I think has made VR a tougher sell in a lot of places and in a lot of situations that goes away to some extent.
It does, I mean this sort of is another way of looking at that.
I think there are places where it makes sense to train your own machine learning algorithm and have it run locally.
I think there are places where it makes sense to leverage a cloud platform to get right to advanced functionality.
I think because of the computational load, because you're already probably... you know, SLI-ing two 1080 Ti's to get your adequate 90 frames per second in VR.
It's smart for developers to take advantage of cloud platforms that offer that functionality without having to increase the CPU load.
So where does this take us? What does AI plus VR and AR allow us to do?
And where does it go in places maybe that we haven't seen it yet?
Boy, that's such an interesting question.
I can't wait to see. What I'm trying to do is have tools built for developers so that they can more easily plug in some stuff. and hopefully potentially be creative and inventive and combine it with some of the things they're already doing.
Like for example, with the voice interaction, A system that we built, and by the way, is freely available on GitHub.
Actually, developers can take that and build their own version.
In the demo, it's essentially a whimsical god mode demo.
So you can point somewhere and say, create a big black dinosaur or a small blue car, and you can modify things just to get a feel for how a natural language command interface might be.
But I'd love to see people build an advanced emotional story with a virtual partner in a game. that's story-based or something with a productivity tool where maybe you're using Autodesk or Tilt Brush and you say,
Give me the brush I have, three brushes a grow, make it two points larger and black and have that be a little easier and break immersion less than rifling through a bunch of menus. if you're a serious sculptor.
So really, the sky's the limit. I mean, I can rifle through for hours on end, as I'm sure you can.
Instances I've seen so far, things that are interesting...
But really, yeah, I mean, AI is going to be used left and right in VR.
It already is. And we're just hoping to play a part in building the underlying technology that makes it easier for developers to do that.
And if it's easy for developers to do it, does it make it easier for it to show up in healthcare in kind of non-entertainment venues?
Yeah, so specifically we're looking towards productivity, education, therapy, training.
These are the use cases that I think are really interesting from an IBM perspective.
In terms of VR, I mean, everybody knows the power of immersive storytelling and of gameplay and whatnot in VR.
And so I'm excited about that too, but these are the opportunities that people haven't taken advantage of yet.
And yeah. certainly if we make it easier for people to implement advanced solutions, like being able to talk to a system and have it respond, that's just going to sort of accelerate the development of those solutions you're talking about.
Take us through a scenario in education and also one in, I don't know, Manufacturing.
I'll take you through one. This is sort of education training.
How about that? So there's a company that does dangerous job training.
So there's a simulator where you as a site inspector, person working in construction, put on a hard hat and go into a high rise in virtual reality and you have to identify all the dangerous objects that shouldn't be there.
There's an open band saw, there's a light with a a short in the wire or something like that.
Right now, what you can do is you can call it out with your voice.
You can ask for help. You can say, what am I supposed to do?
Or give me a hint, things like that. So that's sort of a light touch implementation.
Another one that's interesting is there's a surgery training simulator that previously had a menu selector that you would select your scalpel from.
Well, what if you could just say to a nurse, and again, I don't know if this is actually a real scalpel.
Whatever you could say to a nurse, hand me the N5 scalpel.
That's actually a skill you'll need in life.
In terms of training, the sky's the limit.
In terms of therapy, there's some really interesting work being done on patients who are quadriplegic, who are regaining some vestigial motion in their extremities, especially people who suffered a stroke.
There are some shortcuts that you can make the mind take by using virtual reality.
And I would love to see how machine learning algorithms, artificial intelligence can play a role in sort of expanding on that concept.
Yeah, I mean, I could keep going forever.
Really, like I'm still looking for the next killer use case.
And so if anybody who's listening to this podcast is interested in sharing what they're working on.
In terms of AI and VR, I'd love to collect that and bring that into my talks going forward.
I was going to say, I know you could go on forever, but I know you have to go speak to a bunch of people.
So give us the... the call out to developers, what should they do?
What do you want from them and where can they go?
Cool. Yeah. So first of all, if you want to follow me on Twitter or DM me or something, if you have an interesting idea, um, I'm Michael underscore Ludden is my last name.
I'm sure it's in the show notes. But I would also say go to our GitHub.
Just do a search for VR Speech Sandbox IBM and you'll get our Git repo.
It's a fully open source thing. You can take the sample code.
There's a comprehensive how-to guide. And if you guys have questions, feel free to reach out.
Great. Michael Ludden, thank you so much.
Thanks a lot. PostScript here. If you are listening to this podcast on iTunes, Google Play Music, or whatever your favorite podcast player is, do us a favor and leave us a review.
Thanks for listening.
