 Thank you and Welcome welcome everyone. We are absolutely Delighted to be hosting all of our AI HN Researchers here for AI Research Week. You're probably aware. We've expanded You know how we do this in addition to the colloquium that we normally do in our Yorktown lab We'll also be doing the poster session, which is an enlarged format. You'll see 70 probably 80 posters there We have the open house tonight at IBM Research in Cambridge just a couple of walks away We'll walk over there after the poster session and then we have the entire week of you know Technical workshops and meetings and so on to help us to to hopefully advance our research agenda as we're going But it was especially important for us to start this entire week With a gathering of our AI HN University collaborators So our closest collaborators here having sort of a more of an internal close-knit Meeting to talk about the research and then we'll go out and as I said tomorrow We open it up for the colloquium for a greater part of the public and and so on so so thank you And and you're very very welcome. So with that we'll we'll get started Since we do have some new members this year, I just wanted to maybe reintroduce a little bit about IBM research As you're sure you're aware IBM research takes its responsibility its research focus very seriously Our research is is recognized worldwide. We have 3,000 researchers in six continents in 13 labs About a thousand one-third of those are focused on AI so core AI algorithms And driving AI into industries in transformative ways this this quote here in terms of 9,000 patents in 2017 Roughly 1500 of those alone were in AI. So AI is obviously Core to to what we do in in IBM research And that's because we see AI as something that will transform All of our products from our cloud platform to the services we offer to the solutions for industries We see that as as transforming all of that work So it's critical for us and as part of that we are expanding our age our horizons Network and I'll talk about that in a moment So this is a you know just to show you right so we've actually Just this year alone. We've expanded. There are two new members that are here with us today The UMass at Amherst is a new program that we've started with Andrew McCullum focused on information extraction and representation for Essentially being able to create, you know learned representations for more advanced forms of reasoning and Andrew will talk about his research In a few moments We've also have a new member as we start to expand globally So our first are a second actually we also have the University of Montreal in Canada but our first addition to that is IIT Bombay and here again what we're focusing it is on knowledge representation information extraction reasoning Better training for AI systems in terms of dialogue and conversation and and so on And if we look at you know kind of overall this you know this this landscape of research that we have What you'll see is that it's Fantastic in terms of both being able to advance the core AI algorithms as well as how we Transform you know industries with with leading-edge applications of AI to those so with UMBC We're focused on AI for cyber security with MIT. We're focused on core AI algorithms as well as two Very important industries healthcare life sciences and cyber security and now this year We've just started to expand into into finance as well With Michigan we focus on Conversational systems dialogue especially goal-oriented dialogue or AI systems for goal-oriented dialogue The University of Montreal. It's all about deep learning With RPI we focus on a cognitive Immersive environment that can learn and adapt and interact with humans in a very fluid way As well as our heels program, which is about health care effectiveness and improving that through knowledge representation and reasoning At UCSD you're going to hear Rob talk a little bit later But we have two main thrust there one is the the microbiome and in addition to that We're doing work in terms of how AI can help with healthy aging Enabling people to be able to use a system so that they can age in place, you know in their homes you know in their environment and and Hopefully lead very continue to lead very productive lives as they do that and then we also have the University of Illinois So with University of Illinois UI you see our work is focused on high-performance systems optimization of those systems for AI and machine learning as well as some work in computational Creativity and an interaction systems as well Okay, so I also wanted to make sure that we celebrate some of your work The team AI Chen has had really a fantastic year this year. We're up to about 130 students about 90 faculty 120 publications this year Over the over the course there And what we're especially excited about is that what we're seeing more and more is that the publications are collaborative between IBM researchers and between the researchers in the in the University, so The the UI you see team has been especially busy this year So one example is their DNN builder tool, which is a which is a poster that you can see at the poster session one the best the best paper award at the automation design conference in The front-end category really focused on you know AI accelerators for for systems for people who are building and Accelerating systems just really quickly as anyone here from the DNN builder team. Just take a look here. There we go Thanks also In CVPR there was a competition in terms of look into person so being able to actually understand somatic parts of humans and doing parsing of that and The UI you see team Placed first in all three of those human parsing tasks. So a fantastic job for the UI you see team Go ahead if you're if you're here Raise your hand as well Excellent excellent job. So within that they also won the the the system design competition for the you know Automated design for automated cognitive design as well Next so we will at some point move on from from from UI you see but Additionally, they've won the the best paper They are a finalist for the best paper award for micro architecture 52 for their work in terms of you know near-memory Processing architecture for for the memory network channel. So well done. Here you here There we go. Okay. Thank you And they find their results in October. So just a couple of weeks. You'll get you'll get the final results Okay, additionally so in terms of a new, you know our new collaboration collaboration with UMass is already yielding best paper award This is a very interesting piece of work in that there was a data set that was released to especially try to target Advances in question answering that would require going beyond Information retrieval and PMI types of techniques and this team You know received a best paper award for that work. So is UMass team here? Let's recognize. I know I saw Andrew earlier He's gearing for his talk. Okay, let's let's give a hand to Okay Additionally the moments in time. So this is a data set Essentially a one million video data set I think they're up to around two million labels for that video set So the most labels of any video data set. It's really focused on trying to go Beyond object recognition into action recognition from videos So what the team has done is they've used their AI machine learning machine vision capabilities to you know amass this huge data set to do it in a way that the labeling much of the labeling could be Automated it could be augmented by humans more doing, you know Decisions based off of that as opposed to having to do the full brunt of of annotation for that So not only did they release the data set in 2017 as part of NIPS But they've also gone back and they've issued their first challenge at CVPR The CVPR challenge was the most participated challenge at CVPR I think it was a hundred and twenty different hundred twenty three different teams a hundred and fifty Submissions so really a great success in terms of this. So do we have our team here? Let's give them Let's give them a hand. I think oh it is also Think oh it is also in the in the back preparing for her talk Okay Now in terms of Michigan, so what the work that's been doing there is basically to say that we need Better data and well-defined tasks that will help us to advance in goal oriented dialogue So the combined team has created a data set of over 80,000 Conversations they released that along with tasks for a DSTC seven Which is a program with multiple different tasks? And as part of that people are already starting to to work on it as part of that They propose a workshop at triple AI that's been accepted And then they'll be able to show the results and have interactions on those as well So so well done to the to the Michigan team or raise your hand or you've got some here Let's give them a hand Okay, so now what I do like to do is shift a little bit and start looking ahead, right? One of the things that we really want to do this year Now that we've you know We've really got a very strong partnership a very strong set of universities that we're working with Collaborations that are starting to really you know both yield fruit and new ones that are that are emerging and coming up We want to make the program better for you for everyone So we think that in terms of doing that that some of the things that we can do to make The collaborations better to help with your research One is to to have tighter collaborations with IBM researchers, right? So as an example one program that we've been working on crafting is an in-residence program So while we all do you know internships over the summer and so we believe that you know The research is happening obviously, you know during the semester as well So we're looking at creating programs so that the so that you You know researchers our researchers with their their faculty can come to our research lab Spend some real time, you know working So we think it's great in terms of when researchers come across and and do seminars and talks and so on but we think expanding upon that so that People can spend time actually working side by side, you know fleshing out their plans, you know drawing on the whiteboards You know continuing to work together is will be huge in terms of helping us to advance So we're looking at some in-residence programs And we'll be talking more about those soon And then finally we'll start a regular webcast program So roughly, you know every four to six weeks We'll have a webcast that will invite all of our, you know researchers who are part of any of our eight AIHN projects to participate and we'll do a combination of things on these webcasts, right in some cases It may be, you know tutorials Maybe on some of the tools that one of the labs are creating so that other people can use them And in other cases research seminars just to to disseminate results from the the projects and get others involved Our first we thought it might be useful as our first one to to give a tutorial on our AI cloud platform So our cloud platform if you're from in case you're not familiar with it We've released our deep learning as a service technology that allows one to Create, you know their their their networks their architecture And then be able to push that on to our cloud platform get access sufficient access to the GPUs have that load balanced and And then results coming back and and make it easier to to manage those results with what we refer to as An experiment assistant So as part of again trying to you know help with the research What we'll do is to provide a tutorial The tutorial is not to teach people how to to do machine learning or deep learning It's tutorial really to just so that you can understand how to get on to our cloud How to use the resources there how to take advantage of deep learning as a service and so on so we'll do our first one there The teacher will be here actually as part of the open house. So our herman is is out in there There's herman. He's out in the audience. He'll be at the open house. He'll be around the poster session And then we'll so you're welcome to to ask him any questions during that time Otherwise we will uh, you know, we'll we'll start with this this tutorial. Let me know. Okay, so now One of the things that we're doing also that's that's new and different is that we uh, you guys are kind of like Packed in here And so being able to ask questions by having people get up and walk across the room Maybe a little bit difficult and we thought we could actually use some technology To do that. So we've set up an app Slido That hopefully you've already seen as part of the registration and let me see here. That's So there we go. Okay, great. Um, so what happens is uh, you you simply go to this to this url You can join there. Um, you feel a very short survey. It's pretty easy to do if you haven't already do it I'd welcome you to do that because we're going to be using this throughout the whole week Okay, thanks Okay, so we'll be using this throughout the week. So if you haven't already take a few minutes and Please sign in and we'll use it for two things, right? So one of the things that we'll use is that if we want to take a poll of the audience and I have a couple of questions Of specifically on how we can improve the program that are out there That'll ask you in a moment And then I also, uh, we'll also use it for questions, right? So If you'd like the the way that it works is that you can ask a question Others in the audience can kind of upvote it and then we'll take the top questions and we'll use that for the for the panelists So we have a couple of panels additional talks here So I definitely encourage you to to go ahead and get set up if you haven't already done that So let's go to the to the first survey first poll question Okay, so great. It looks like people are already getting started on this. So the first thing is To better understand, uh, you know research areas and interests that you are most interested in and I would really encourage you to Both, uh, you know provide kind of the the higher level Areas so, you know NLP or reinforcement learning or these things but also to be you know more specific In terms of research areas that that interest you as as well So I can see that this is this is growing up and not surprising there. We're seeing a lot Around learning so that's that's fantastic I think while we while we let that Continue to accumulate, please, you know, please keep, uh, you know putting in in your answers We can come back to it in a moment. I'll go ahead and seed with the additional question That I wanted to to ask people so if we can Kind of go back that okay Fabulous okay, so The thing that we want to do is we think that you know We've learned actually a little bit from this The fact that we've established this MIT IBM lab here, you know in close proximity with MIT Researchers going back and forth across the labs working in each other's environments and that has really Helped a lot and so we're looking at expanding the way we do Residencies, you know in the IBM research lab. So we've always done, you know, uh, you know internships over the summer But we think that during the semester actually having researchers come in Like I said, not just for a for a day, but potentially for a week or two weeks three weeks Or even an academic, you know a full academic semester would be very useful obviously not All of the you know all of the years of once PhD but there are key periods where we think that that You know much deeper interaction the ability to really work together You know side by side shoulder to shoulder on the research could be valuable And it's certainly valuable for us as well The difference obviously here is that during a summer internship usually people come in and they they work on projects that are assigned by the By the IBM research lab in terms of priorities for projects in this residency program We really be looking at for the research that you're already doing as part of your ai hn program and project It's more about just continuing to that research But doing it, you know, you know side by side with the researchers that you are working with So your your pis and your leads and your you know your co-researchers on the projects So Okay, so let's just take we can take a quick look at the the results here because they're starting to come in Okay, so not not surprising here in terms of you know, learning deep learning Quite a bit on healthcare and nlp microbiome reinforcement and so on So this is this is excellent I think that you know we can you know, we'll keep this open and going for a while And so if you have other ideas, please continue to do this We'll also you know find other ways to to get information on on what you're interested in in terms of research areas So maybe we go to the to the next one as well Oh, I'm sorry the the next question. I was just going to take a quick look and how it came out in terms of the the timing there, so Okay, so it's roughly 61 so for this full semester. That's fantastic. Actually, this is this fits very well with what we were thinking would be would be valuable For people. So so thanks very much for doing that Okay Now any any questions, can we just take a look? Are there any questions here mark that have come up as part of this or not? Ah, so okay. This is good All right. Is the residency program just for students are also IBMers and how about faculty? Yes, so we think that Definitely in addition to students having faculty come in as well for residency programs Would be very good. It turns out I was talking to Andrew just before we got started And he mentioned the importance of our researchers also spending time in in university So we'll we'll be looking at how we do that as as well. So very timely comment from from Andrew as well More international expansion this year. Yes We are working on those are actually, you know in work as we speak There are a couple more that we've been working on. So what we've started in addition to, you know, the regular programs We are starting to You know have call for proposals for additional programs will get you more information on that for now You know, we are focused on expanding internationally. We think that's a a very important part of our Our agenda here because as you're aware, you know, we have 13 labs We're you know, we're on six different continents and this, you know, collaborative research You know for the purposes of advancing AI is critical to us So making sure that we have additional international and global labs is a key part of the the program Okay, great So we'll you'll have other chances to to ask questions as we progress But I think we're fine for this so we'll switch back To the to the program here and I'd like to introduce our first keynote speaker So if you can go ahead Oh, actually one. Yeah, can we just advance? There we go Okay, all right, so For the rest of of this session, we have two keynote talks AI and the human micro microbiome is the first one And then we'll go to AI for for representation and reasoning And then what we wanted to do is because we feel that, you know, the emphasis, you know, the the the strategy here is really about driving up the collaboration There are a few projects that have created assets that we think will be fantastic For helping collaboration not only between the researchers and university But also collaborations across the different universities So we've selected a couple of those projects As as collaboration opportunities that we would like to make everyone aware of And then hopefully that will that will help in terms of seeding those collaborations These are short talks. They're meant to sort of You know seed the discussion and what we're planning Is that you know as part of that that, you know, the seminars the regular simulators that we'll be holding We'll ask the speakers if they can come back and do a more expanded version To help people actually really get started with with some of these tools and technologies and so on And then last what we have is an inside peek at IBM's AI strategy. So this is hopefully, you know, somewhat of a unique You know session for you and I wanted to tell you a little bit about it So that you can be thinking about the kinds of questions that you want to ask We thought it would be, you know, especially interesting if you were able to ask You know our leads for our core programs, you know, AI science AI tech AI engineering if you were to ask them Anything Right. So literally think of it as as an ask me anything And we will be as open as we possibly can about, you know, what we're doing today Why we're doing it, you know, what we see in the future how you might be able to help with this So take it literally We they the finalists have signed up for and asked me anything Session so so please please be ready and we'll use this same system that we were already using in terms of asking the other questions Okay So for our first talk, we have rob knight who is the the founding director for the center for Micro biome innovation. He's also the professor for pediatrics and computer science at uc san diego I think that Rob also wins the award for cool project names. So in addition to the earth microbiome, he is also the lead for the american gut project Which I which I thought was just fabulous in terms of putting it Is right out there and it's also fabulous in terms of the the project, right? So if you if you look at this, right So what's the connection between the human biome and ai? Why was it so important for us to have this collaboration? Well, rob's team is focused on, you know, the software the tools and the technologies and the the approaches that will enable Very high volume You know, you know analysis of all this data to give you to give you an indication In the earth microbiome project, there's already roughly 20,000 or 27,000 Samples that they've collected and posted about 100,000 that they've collected in addition to that Well over two billion DNA sequences from that So the tools that they need from ai and machine learning in order to be able to Identify and match and try to find their way across that space is is critical