How to get plugged into the AI community

Welcome to another Fully Connected episode of Practical AI. In these episodes Chris my co-host and I keep you fully connected with everything that’s happening in the AI community. We’ll take some time to discuss the latest AI news, and we’ll dig into some learning resources to help you level up your machine learning learning game.

I’m Daniel Whitenack, I’m a data scientist with SIL International, and I’m joined by my co-host, Chris Benson, who’s a chief strategist for AI and high-performance computing at Lockheed Martin. How are you doing, Chris?

Doing great. How’s it going today, Daniel?

It’s going good, no complaints. I was dealing with some CUDA version issues for a couple days, and that is in the past, so all is looking up. What about you?

Well, doing similar stuff… I moved into a different position at Lockheed Martin. My boss became the chief data officer of the company, so we’re doing all sorts of cool stuff at work, and that’s taken my time.

Awesome! Sounds exciting. Well, we’re moving into summer, and I would kind of consider summer and into the fall the really heavy part of conference season. I know I’ve been trying to figure out what I wanna put on my schedule where I wanna be, what I’m involved with… And I know you’re pretty involved in the conference circuit as well, right?

I am. When we decided this is what we’re gonna do, I went back and looked and I think I gave about 20 talks last year.

Yeah, that’s a lot… I’ve scaled back probably from that level; I’m probably involved in a handful of conferences each year now, but…

I’ve scaled back this year as well. It was getting to be a bit of a burden. A little bit too much there.

Yeah, it can be too much… But since we’re going into conference season, what I was thinking was – you know, we’ve talked about a lot of technical subjects on the subject, whether that be certain types of neural nets, we’ve also talked about overviews of infrastructure and other things, but we haven’t really talked about what types of events happen in the AI community, and by events I mean conferences, meetups and things like that…

[04:14] If you’re in the AI community, how you choose which conference to go to. There’s a lot of hyped conferences out there - how do you know what’s gonna be valuable for you, and how do you get involved in these things in terms of speaking and other sorts of involvement… So I thought it’d be good to just go around and see what your thoughts are on those subjects, Chris, and maybe we can parse through some of this world of AI events. Does that sound okay?

Sounds good to me.

Awesome. Well, I’d be interested - and I’m sure listeners would, too - in how you went from maybe… I’m assuming at some point in your life you were not giving 20 talks per year, or being involved in 20 conferences, so what does your background look like on that front and how did you ramp up into that? Just generally.

For me, I’ve done talks on and off over the years, but actually as I turned toward this, I’m gonna point right back at you and note - this is something obviously you know, but I turned to you as I was thinking about doing some talks here, and got some great advice from you, and you kind of steered me into how to do it. That was very welcome, I’ve used that over and over again, and then I kind of dived right into it. I did some stuff, and kept building upon it, and did a little more, and building upon it, and after you get a little bit of experience, even if you’re starting from nothing, you quickly learn how to do it. We’ll talk about a lot of that today, and dive in.

I started with meetups, and moved up to doing conference talks, and then eventually keynotes, and they all have a distinct flavor all themselves.

Yeah, definitely. For me, I came originally from Academia, and the expectation there is probably that you’re gonna be doing some teaching, and research talks, and that sort of thing… So I did a little bit of that; that sort of world is very different from a lot of events in tech, maybe not so much like some of the AI events that we’ll talk about here in a bit.

So I got familiar with that, I actually really enjoyed teaching when I did that, when I was a grad student. I think I always had a desire to be involved in community and teaching in some way. I do remember when I made my first step to giving a talk at a tech conference, which was a very new thing to me, outside of Academia… I was pretty terrified, because I really had no idea what it would be like, what sort of thing I should present, and all of that… So yeah, it can be kind of a rocky road if you’re getting into it, so hopefully we clarify some of that road today, and also even if people aren’t wanting to get involved in a speaking context, at least try to figure out what are the sorts of events that are going on in the AI world, and what might bring value to them in other ways.

On that front, let’s talk through the general categories of AI-related events and community things that are out there for people to get involved with. So you mentioned – I think the first one is meetups; you mentioned you got started getting involved in meetups, so could you describe maybe for people that haven’t ever been to a local meetup, what that is, what you mean by that?

[08:02] Sure. It kind of originated – there’s a site called Meetup.com, that is probably the most prominent organization site, but there are others as well… But it kind of became a thing onto itself, as a term. It’s essentially just a local gathering of people who care about a particular topic; it’s super casual, and there’s different things you can do. You may have a speaker or a group of speakers in a row, each doing short-form talks, or you may just have group conversations.

I’ve done a little bit of all that, including things having nothing to do with technology. Just as an incidental side thing, I happen to be a vegan and I go to vegan meetups, and we explore different food, and stuff. So it’s not a tech thing strictly, but… In this context, I had been to a bunch of different meetups in the Atlanta area, which is where I’m at, and I decided that there wasn’t really a great meetup out there addressing AI, and in particular deep learning, which was what I was most interested in… So I had been to some, and then I just took a wild step in the dark and I set up a meetup for deep learning called “The Atlanta Deep Learning Meetup.”

For the first one I just made myself the speaker, with an intro to deep learning content… I put it out there, and I had no idea if anybody would show up. I was pretty amazed that even coming into it cold and having no experience, a whole bunch of people showed up, and it’s gone ever since then.

I’m always telling people – people are afraid of speaking, or getting involved, and I say “Just go to your local meetup and just offer to do something.” If you have an interest in a topic, but you don’t actually know it, a great way to learn it is to promise in a month or two that you’ll do a short talk on it… And even if you know nothing today, it puts the pressure on you to go learn it in that time. Then show up and casually give a few minutes’ talk on that topic. There just really isn’t a better way to learn, in my view… Because if you have to force yourself to learn well enough to teach, then you’re gonna at least understand the topic at a basic level.

I think you drew out some important points there… So a meetup, as opposed to a lot of other conferences, is generally like a grassroots sort of thing. You saw that there wasn’t a deep learning meetup in Atlanta, so you created one… There’s just people creating these, and so you will find a good bit of variability in these. Some are really small, some are a lot larger; some are very structured, some are very unstructured. Some have free food, some don’t.

I think it’s one of those things you have to be a little bit persistent with. If you’re not starting your own meetup, then if you go to one and you don’t like it so much, be persistent because some – I would say about maybe half in my experience of meetups, it’s kind of like you go and everybody seems like they’re just there for free pizza, and they’re not really interested in the topic… But then about half of the people are really interested and really engaged, and great discussions, great talks, great content… So don’t give up after that first one. If you don’t like it, find another one. There’s a lot of different subjects around which meetups are organized. Some of those are around certain technologies, like TensorFlow or PyTorch; some of those are around certain topics, like deep learning, or maybe geospatial data, or other types of data; distributed computing… Lots of other things.

So go to meetup.com, start typing in what you’re interested in, and I’m guessing if you’re in a fairly populated area, there’ll be something near you that you might want to attend. But I think this is, like you said, a great first step. If you’re just getting involved in community sorts of things, go ahead and – even if you’re not prepared to really take a leadership role, just get involved in the community, go there, meet some people, talk to people, and start making some connections.

[12:19] And volunteer. Even for small things. It makes a difference, and it opens the door to meeting new people, and that’s probably as important as any of the topics you’re gonna hear - meeting other people that are interested in that and forming those relationships.

For sure. In addition to meetups - these sort of grassroots things - in the AI community there are a series, or actually a lot not, a lot of large conferences. A first group of these larger conferences I would think of as large industry conferences. Here in a second we’ll talk about large research conferences, but there are large industry conferences, and think you’re probably familiar with this distinction as well, Chris. Could you give your thoughts on that?

Yeah, and there’s a level of structure that you’re gonna find at a large conference, compared to a smaller one. We can talk a little bit about both, but… At a large conference they’re gonna have a very organized way and deadlines to submit your talks, and they’re gonna have a panel typically of people that are going through, if you’re speaking or setting up various tracks… So they will attract hundreds or thousands of people to them. Something large like – I know well the O’Reilly conferences, like O’Reilly AI and Strata and such… They’ll have multiple tracks, so you’re not gonna see everything. You’re gonna see a very small percentage, but it’s very structured. We’ll talk a little bit about strategy on getting the most out of that as we go. If you contrast that with smaller ones, you may be invited to speak, or invited to a keynote, that kind of thing. So it’s the amount of structure and the amount of formality around it, the larger it gets.

Yup. And I think that these conferences - like you mentioned, O’Reilly AI, Strata, QCon, Spark Summit, MLConf, Applied Machine Learning Days… There’s a whole bunch, and we won’t attempt to list out all of them on this podcast… But you’ll kind of get a sense of these conferences being places where people from larger companies, and also people from startups as well, but mostly people from industry are really presenting about their applications of AI, and their unique infrastructure around AI, how they used AI to solve certain problems… So I would say it’s more tech and application-focused in the industry context… Which is different from some other things that we’ll talk about here in a second.

Yeah. Another thing - it might potentially be a downside in some of the larger ones, as you’ll also see a lot more commercialization these days.

Yeah, some of them are expensive, too.

Yeah, they can be very expensive, and you really have to pick and choose what you want, because there are talks that they will – there may be a technical topic, but it’s really about what a particular company is doing very specifically, and there’s a little bit of branding around it… I’ve gotten very particular in how I choose to spend my time at the larger ones.

Yeah… In some cases, you’ll see sales-pitchy talks, but I would say there are good, large industry conferences where the content is really good. I think, just like you said, you have to be aware that that could be the situation…

[16:03] That it’s a possibility…

Yeah. I think that one trend that I’ve been seeing, which is kind of another category of industry conferences or events is kind of like smaller, topical events, or events focused on particular toolsets. I’ve seen a lot of these pop up lately. There have already been the TensorFlow DevSummit and PyTorch Developers Conference… But I’ve seen things recently pop up around certain languages, of course, like our conferences… But there’s also a spaCy conference called IRL (In Real Life). I recently saw announced an Allen NLP Summit… There’s a lot of different events that are smaller. It might be like 50 or 100 people there, but it’s the 50 or 100 people that are really interested and in the weeds with spaCy.

So if you’re that type of person that is really into a particular topic or toolset, and that’s kind of where you live and do most of your work, that could be the most valuable thing for you. Instead of spending $2,000 or $3,000 to go to a large industry conference where there’s very broad presentation of content, it might be better to just go to a place where you’re gonna be able to focus and hear a lot of content about the specific topic or toolset that you’re focused on.

I totally agree. And there’s another niche that I’ve seen small conferences do - the space between where a meetup is and where a full large conference is, and I’ve seen it kind of as an intermediary thing. You might have a meetup or user group that is meeting on a monthly basis in a meetup context, but then once a year they will try to put on a larger event that still caters to their community, and bring in external people, and such… I’ve gone to several of those. DataSciCon in Atlanta is one, Applied Artificial Intelligence Conference in London a few months was another one… It’s not the massive scope.

I even saw – there was a really good one that I went to that was actually put on by a group of college students at Vanderbilt called eMerge. It particularly impressed me, because these were young adults that were not really out into industry yet; they were still taking classes, but they managed to come together and do it… And it turned out to be a fantastic conference.

So I’ve really gotten to where I like those in-between levels. They are large enough to get more than you’re gonna get out of a meetup, but yet they’re still small enough to where you actually meet most of the people there and have a great chance to network locally.

Yeah, definitely. The last category that I think I would put on my categories of events when I’m thinking about events to get involved with, or research, is actual AI research conferences. There’s a set of events that are really geared towards original AI research… This is a lot of people from Academia, from universities, but also people from industry that are from R&D departments and other places, and places like OpenAI or maybe Hugging Face, and other places that are really pushing the boundaries of what people have done with AI before. It might be people that have a new large-scale language model, or people that are doing very interesting, unsupervised video methods, or something like that, that people really haven’t done in this way before…

[20:05] So a lot of times those things are presented at conferences like NeurIPS, or EMNLP, or ICLR, CVPR… These are all very big, mainstream AI research conferences. A lot of times they sell out all of their tickets; you might not even get a place there. It’s also kind of a high barrier to get something submitted to those. At the same time, the things that are presented there are just mind-blowing. So it’s not worth totally writing those things off, and if you want to go down that research path, there’s definitely chances to get involved there.

Yeah, the topics are very advanced, and they’re probably not for a general audience. You really have to already by at some level of expertise in the field to get meaningful stuff out of that. But that’s also where you’ll find all the big researchers around the world in the various organizations, whether it be OpenAI, Google, Microsoft, you name it - they’re there. So it’s a totally different type of thing than one of the standard, large industry like O’Reilly’s and such.

Alright, Chris, now that we have our general categories of conferences… We’ve got research conferences, and smaller topical or focused conferences, we’ve got larger industry conferences, we’ve got meetups… In your opinion, maybe where we should go next is talking about our opinions on why we participate in these sorts of events and maybe why we would recommend other people participating in these. What are your thoughts there?

It’s interesting, something I noticed both about myself and other people that I’ve talked to - when people are first starting to go to conferences, they tend to say “I’m gonna go and I’m gonna listen to all the talks.” They wanna get their money’s worth and sit through… But after you’ve been to some, you start - like I alluded to earlier - kind of picking and choosing, and figuring out where you’re gonna get the biggest bang for the time that you’re spending there…

I know for me there are a couple of different areas, and one obviously is that during that conference you have a concentration of expertise about the topics that the conference covers. So if you are doing work in a particular area or are interested in a particular area, it’s a great way to try to accelerate your learning process. And the way that I do that at least is I differentiate what can I get done on my own when I’m not at the conference, versus what do I do with all these experts around that I have access to.

So I will try to pick and choose the issues, and I will try to connect with people so that if I’m having trouble finding answers, or how to get started on something in a certain area, or maybe I’ve run into a difficult problem that’s not well-documented, that’s a fantastic kind of thing where you can go up to people who have been there and done that, and quickly get through those hurdles. So that’s one of the parameters that I’ll use when I’m trying to figure out what talks I’m gonna go to and what people I wanna meet.

Yeah, and I think that particularly the larger industry conferences - you can kind of go into them and say… You know, instead of saying “I’m gonna just try to consume as much as material as I can”, you can be (like you said) very targeted.

I remember going to a couple conferences… It was probably a couple years ago now, when I was thinking about the best way to run TensorFlow distributed in Docker containers. Is that possible? Are people doing that? What are best practices for utilizing GPUs when you’re using Docker containers? That might sound kind of focused, but at a larger industry conference there were a ton of people there that were first of all using TensorFlow, second of all doing some sort of deploys involving Docker… There were a bunch of people there involved in GPU stuff, and maybe people from NVIDIA… So what I could do at that time is just say “Okay, who are all those sorts of people? Let me go to their talk, which is probably interesting anyway… But let me go up to them afterwards and kind of hear more about their talk, but also pick their brain on some of these things, discuss some of these topics, how they’re doing certain things, what recommendations they would have…” So it’s really a shortcut to very concentrated information, like you were saying.

Yeah, totally. I use it to accelerate things… And it kind of leads to the next thing that I wanted to mention, and that is a big part of conferences is networking. It’s getting to know people. It’s part of this. You may think “I wanna solve the problem that’s been bugging me, that I’ve spent the last two weeks trying to figure out how to deal with, and I’m struggling with it…” You start with that, but then you get to know and you may keep up with people thereafter and start building relationships… And over time, in a lot of ways, that becomes even more dominant than access to the expertise itself - having those relationships; it kind of enriches what you’re doing, your professional life. You’re developing friendships, and stuff.

A good example - you and I both know Mat Ryer originally from the Go programming language community, but we got to know him at conferences. We met him, he’s a funny guy, super-smart… He’s now on Go Time as one of the co-hosts. He was also our first guest on this podcast. He was our very first guest, one of a couple.

[28:08] Those kinds of friendships – you will meet people and see them over and over again, and I have quite a few good friendships that I’ve started at conferences, of people that I’ve met, and that we just kept up with, and then periodically we’ll see each other at these events… And I can’t tell you how much value and enrichment I’ve gotten out of that.

Yeah, and I think that that networking is really great for building those relationships… And also just in general, making your life a little bit more productive. I know now I’ve been working on distributed teams or as a remote worker for – well, I’d have to count up now; it’s probably at least 3-4 years… I forget. But a lot of times the places where I meet with my distributed team is at conferences, so if there’s a conference we’re interested in, why don’t we all meet there, so we can both attend talks, but also have some in-person time? So there is that useful bit. But also, if you’re involved – I mean, the AI community now is really driven by open source… Whether it be TensorFlow, PyTorch, the Onix project, things related to NVIDIA, things Microsoft is putting out, spaCy - all these different sets of tools that are really driven by open source… And if you’re working in AI very long, you’re going to at the very minimum start using open source projects very heavily… But ideally, you’ll also start contributing and being a part of those communities.

Maybe you’re a part of special interest groups for an open source community, or you’re opening issues… Hopefully, you’re submitting PRs, and that sort of thing – and I know there’ve been so many instances when I’ve been involved in projects where I’ve submitted PRs, I’ve talked to people on Slack, online, or in forums… And then I see them at conferences, and it’s both great to put a face to that person, and have empathy for them and interaction with them, but also discuss those things like “Hey, this is really what I was getting at with this proposal, or what I was thinking would be awesome to have in this module”, or whatever it is. They can also really help you and accelerate that side of your life as well.

Absolutely. I know when I was at NVIDIA GTC a couple months ago – I work for Lockheed Martin, and we probably spent as much time with a bunch of us that had come in for the conference to doing Lockheed-specific stuff off to the side, and getting face time, instead of it just being conference calls, and such… So that was great. And really I think when you get to that point and you have the expertise available, you have the networking side of things, and then you have this capability of getting other things done, you really start to plan your time out in terms of how you’re gonna get the biggest effect for the time spent at the conference. I find personally that doing that I get a lot more out of every conference than just going and just planning to sit through just talks, or something. At this point I’ll get to the end of a conference and it’s always about “Did I get the most out of every time spent, considering that any given hour of the day there might be multiple options of things I would otherwise like to do?”

[31:45] Yeah, definitely. So now we’ve kind of got our categories of AI events, we’ve got some of our opinions about why we are interested in being involved in community events, whether that be conferences or meetups… Let’s talk a little bit now about – maybe there’s listeners out there that are wanting to get involved in events in one way or another… What are some ways that they can jump in and get involved?

I think depending on your personality and your particular interests, there’s a lot of different ways to get involved. There’s of course opportunities to give talks at events, and at meetups… And those talks could just be anything from a lightning talk, which if you’re not familiar with what a lightning talk is, generally that’s like 5 to 10 minutes… All the way up to track talks and keynote talks that are maybe a little bit more high-profile. Of course, there’s ways to submit original research, like we talked about… There’s also ways to mentor and help communities organizer, so you can volunteer to help at events, providing certain services, volunteering, making the conference safe and accessible.

There’s also ways that you can help contribute by teaching. Maybe it’s a workshop; you have expertise in a certain area that you’d like to get out to the wider community… So there’s also opportunities for teaching and workshops. Depending on what of those things you’re interested in, you might have to jump through certain hoops… But when you’re thinking about these different routes to participation, what do you think are some of the best ways to get started along that route, Chris?

Well, it kind of depends on where you’re at. If you’re just getting started in the field and you wanna be able to go to conferences… Maybe you’re a college student and don’t have a budget, you mentioned volunteering, and that’s huge. There are so many volunteer opportunities that people can do and start working their way in… And then you can get a pass because you’re part of the staff there, you might manage a room… There’s lots of different things that you can offer up that the conference is gonna need. And it’s a good way of not only getting involved when you may not have the budget for it, but it’s a good way of getting access to people and working your way in, and becoming part of the scene.

That’s certainly something that I’ve found - after I started getting involved in conferences, it tends to build on itself and you actually will develop a little bit of a reputation; I don’t mean in terms of being famous, but in terms of within the conference community, people say “Oh, I know Chris, or I know Daniel can help us on that. They’ve done this in the past, they might be interested”, and they’ll reach out to you and ask you if you wanna do things, and stuff. So that’s a good way of getting involved initially.

How about yourself, what have you seen there?

In my mind, a good general scaffolding or roadmap to think about if you’re new into the community, wanting to get involved – we’ve already talked about get involved locally first, at a local meetup, ask to give a talk, ask to help volunteer, and as you do that, you’ll find out where you’re interested in being involved… Whether that’s speaking, or on a particular topic, or you have specific expertise in a certain place; maybe it’s transfer learning, or maybe it’s computer vision, or whatever it is… And then that will kind of help you decide how you want to contribute.

Once you’ve decided that, then go out and do some research in these different areas in the industry conferences, in the research conferences, in the smaller conferences… See what’s coming up maybe later on down the line, like “If I’m at this point now going into summer, I might be even looking at spring of next year, or into summer of next year, and what’s further down the line that I can plan for in advance?”

[36:03] Try to think of some ideas for talks and workshops or papers, find out what that event that you have targeted out ahead of time, what opportunities there are… Are there talk opportunities, are there lightning talk opportunities, are there workshop opportunities? Document those things, make sure the prices and the submission process are clear to you… And one of the biggest things that I would recommend is get feedback. You’re already gonna be involved in your local meetup, and maybe you’re involved online, in forums or open source projects or other things - ask some of your contacts in industry to review your abstract, or to listen to your talk, or to review your submission. That’s a huge help as you’re preparing to submit.

The other last thing that I’d love to mention on this front is, you know, for a lot of people there’s financial or other barriers that may stifle or prevent you from participating in certain events… But I’d really encourage you if you feel like you’re in that place, don’t give up; be encouraged. We’d love for you to reach out to us on our Slack team or LinkedIn page or something. You can find us at Changelog.com/community… And there are conferences out there that are willing to give scholarships for minorities, and for other people coming from other communities. Maybe you’re a student, maybe you’re coming from a non-profit… There’s a lot of different routes into this, so if you feel like you’re having trouble finding something that fits, reach out and we would love to do our best to help connect you to those programs and those ideas. There’s something out there, and I’d encourage you to be persistent and find that, and hopefully you’ll find a welcoming community that’ll really help build momentum in your career.

Let’s turn to actually what it takes to do a talk at this point. That’s obviously only one of the ways to get a lot out of a conference or a meetup, but it’s a big thing, and it’s not as hard as a lot of people expect it to be. A lot of people have fear of getting in, but honestly that’s kind of everybody that’s there is in the same boat… I’ve seen people really blossom by deciding to start off maybe at a meetup and give a talk, and they do that and it works out better than they thought, so they decide to “Hey, I’m gonna try a small conference” and submit. There’s a lot that goes into those different things, and there’s different types of talks.

[39:51] We’ve talked a little bit about starting off with a small, very casual talk on a topic at a meetup, but then after you feel pretty good about that, if there’s a particular topic and you feel a little bit more confident about it and you’ve spent a little bit of time working in it, and you think that you have developed enough knowledge in that to share with people, then I really encourage people to go ahead and submit their first talk to a small conference. It’s a great learning experience; a good change that you may not get it at first, and hopefully with some of the things we’ll talk about over the next few minutes we can talk about how you’re gonna increase your chances of being accepted.

But you kind of work your way into that. It might be a lightning talk as well, which is very short, and in some ways it’s very different. It can be very time-sensitive in terms of what you’re doing… Then if you get good at it down the road, you may find that people are inviting you to do keynotes. It’s essentially working your way gradually up, a natural progression up a ladder, in terms of where you wanna go and the types of talks that you wanna give.

I’d love to dive into what you’ve just mentioned, Chris… What to think about as you’re preparing content. This may be for a track talk, or all the way up to a keynote, but what should you be thinking about, even if it’s for your local meetup - what should you be thinking about as you’re preparing to present some type of content?

There’s a lot of ways to go about this. One of the things that I would suggest is – you know, I’ve heard people talk about “How do you get started writing a book, or something like that?” First of all, you have to read books and be familiar with the format and all of that, and what other people have done. I think it’s similar here. To get familiar with what an abstract is, or the format of a lot of talks and that sort of thing, do some research.

A lot of times now for many of these conferences previous year’s talks are all online, including the abstracts and all of those things. One of the best ways to get started is listen to some of the talks from previous year’s conferences, look at some of the abstracts, and look at what topics were covered, look at the format of the abstracts, look at the format of the talks, the pace that people are going at, and kind of get a flavor for the styles of talks that are done at a conference. I think that can kind of – not that you want to just outright copy someone’s talk, because that probably wouldn’t get accepted anyway, but you can understand “What are some good ways to format abstracts? What are some good ways to create titles for talks? What are some good ways to format your talk?” and learn some of that from people that have been doing it for quite a while.

Yeah, I have a little tie into this - as you’re trying to figure out what it is that you want to talk about and how you’re going to organize your abstract so that it captures their attention – I hope our listeners know last week we had O’Reilly’s chief data scientist, Ben Lorica, who was interviewed; that was when I was at O’Reilly AI… And off, when we were not recording the interview, separately, I asked him – because he’s the chair for the program, and I asked him, I said “Ben, I’ve submitted talks and been accepted by O’Reilly, I’ve also had talks rejected… What advice would you give me?” And it was just the two of us talking at this point… And he said “It’s amazing how many people in their abstract will talk about what it is they want to cover, but they don’t talk about how they’re solving it in the abstract.”

[44:01] He said it becomes a little bit too loose… And I’m paraphrasing Ben, I’m not quoting him directly, but he said “Really focus not only on the problem statement, but on how you’re approaching the solution very specifically.” Until I just said that, that was my secret weapon, but now everybody has heard that; that came straight from Ben… And it was a great point. Share with your potential audience what problem you’re trying to solve, but then talk about where you’re gonna go. Because you could actually have multiple talks about more or less the same thing, with different types of solutions, and so those might get accepted together, because they’re all well thought out approaches to a given problem. I know that for myself, I’m kind of going back through how I’m doing abstracts, and being very solution-focused in what I put out there.

Yeah, I think you have to be specific. If I’m thinking about submitting an abstract - let’s say I’m working in machine translation for low-resource languages. If I’m thinking about that, I wanna tell a story, but with specifics. I wanna grab people’s attention, but not in a very general way. So the way you should not start out is statements like “As we know, data is the new oil”, or something like that. Everybody that’s been at these conferences and reviewed abstracts, they’re tired of hearing that. What they want to hear is “90% of the world’s languages have no automatic machine translation available through Google Translate or other means. We at company X have developed blah-blah-blah with specifics about solving this issue, and really helping these target people…”, or whatever it is. You wanna grab people’s attention, but be specific.

Absolutely. You wanna be unique, and show that you have a unique way of assessing the problem and providing a solution to it, so that people understand that when they come and see your talk, they’re gonna get something they’re not gonna get anywhere else. If you’re a conference organizer, the reason you’re looking for that is a conference is an event. It may be non-profit oriented or commercial, but it’s still something they want people to want to come to and to get a great benefit out. And if you’re able to be that speaker who’s going to provide the unique bit of value there, then that’s going to truly grab the attention of both conference organizers and a potential audience, so that people actually come to your speech.

Yup. I guess maybe we should mention too what you’ve already mentioned - that there’s different levels of talks. A keynote talk might be higher profile; everybody that’s at the conference is in the same room to hear this keynote talk maybe… And then there’s track talks. Maybe at a larger conference everybody together might be more than a thousand people in the room to hear a single person talk, whereas a track talk at the very same conference - it might be like 30-50 people in a room to hear a person present.

Now, most of the time the track talk is addressing a particular – there might be a track on computer vision; or it may be even more specific than that - it might be a track on OCR, depending on the conference context… But these might be very specific tracks, and so the way that you frame your story when you’re submitting to a particular track might be even more specific than it would be for a keynote talk, where you’re really trying to maybe inspire and motivate, and that sort of thing.

I know you’ve done both, Chris… What’s your perspective on that?

[48:03] Starting with the track talk, I would say that the larger the conference, the more specific you need to be. If you’re at a very small conference, and your audience are people who might just be getting into AI and machine learning, then you could potentially do a little bit less specific and talk about maybe machine vision with convolutional neural networks, and provide some use cases… Because there might only be a dozen talks in the whole conference, and that’s unique enough to provide that.

If you go to NVIDIA GTC, their GPU technology conference, which is a pretty big conference, you’re gonna be really, really specific there. It’s gonna have to be something where you’re using CNNs in a specific use case, and trying to get a specific objective. So keep that in mind - the bigger the conference, the more specific you’re gonna wanna be.

Keynotes are a different animal. It’s a larger audience, and you kind of have a job for the conference of framing the conference. There are both opening keynotes and there are closing keynotes, and sometimes given tracks will have a keynote… So if you’re gonna keynote, then you really have to look at the various talks that are in a track, or within the conference, and you’re providing a topic that can extend out to lend themselves to those individual track speakers. So you’re framing how you’re starting, and then a closing keynote - you’re kind of framing the value they got out of it; you’re hitting a topic, but you’re also trying to pull all those little track talks together and tie a bow on the end of it.

Yup. And the content that you present – I would say regardless of whether it’s a track talk or a keynote, there’s some guidelines that I think are just generally helpful for all of these talks. The ones that I usually keep in mind are to not assume that the audience knows very much about the subject that you’re dealing with. This is dependent on whether you’re doing a track talk or a keynote, but if you’re doing a keynote at an AI conference, my general level of thought is that the audience knows maybe what AI is, but not much else, and that’s where I start… Because you’ll have a huge variance in people that are AI-curious, all the way to people that are really into the weeds; but I’ve always found it better to err on the side of being too accessible.

I’ve had a lot of times where I’ve been too detailed and I know that I’ve lost people, and I’ve had almost not times where I feel like I’ve been too high-level or too accessible for people when I’ve created a talk. So I think it’s better to try to err on the side of accessibility… And part of that I think is using pictures. Using pictures and trying to keep yourself from using too much jargon. You’re gonna have to use some jargon, but if you’re using a whole bunch of acronyms more than you really have to, or other things… You know, sometimes it’s better to use pictures and to not make too many assumptions. And of course, for a track talk maybe that’s adjusted a little bit; if it’s a track on machine translation, maybe you assume “Okay, these people may have a baseline understanding of machine translation”, but that’s as far as I would assume. I wouldn’t assume that they’re even familiar with the type of machine translation that I’m gonna be talking about… So I think it’s better to err on that side.

[51:57] I completely agree with you. And if you are gonna use jargon, define it along the way… Because especially these days with so many conferences having the talks on the internet after the conference is over, you may have a large number of people – the majority of your viewers may happen after the conference is over, when they’re viewing it on YouTube, or some other platform… So be very conservative in what you’re assuming about your audience. And there may be people who just know everything there is to know that are there, but there may be people, as you said, who are just barely aware of the topic and they’re interested in learning more, and if you don’t help them through that by either not using jargon or defining it, then you’re gonna lose them along the way, and they’re not gonna get the value that you’re hoping they’re gonna get out of it.

And pictures absolutely. This is a medium that lends itself – you’re doing a talk, but you definitely… And a little bit of this is probably some of the standard stuff that we talk about; the same thing that you’re saying, you don’t want that on the screen. You certainly don’t wanna read off your slides. Have a picture that emphasizes what you’re trying to say and is really focused on getting that idea across, so that you’re complementing - you have the verbal and you have the picture, and together they help communicate what you’re trying to to the audience.

Yeah. And maybe we can start to wrap up and discuss one more thing, which I think is relevant even beyond the conference setting… Which is the way that you present content, even if it’s in your company. I think this is relevant if you’re giving an AI talk at a conference, or if you’re presenting your new AI-driven application to stakeholders within your company. A lot of this stuff carries through, like we’ve already talked about assuming knowledge and using pictures, but also in the way that you prepare. You’ve already hinted at this, Chris, but in the way that you prepare, you can prevent a lot of pitfalls.

I think one of the biggest, biggest things that people are probably making the wrong assumption about is that these keynote speakers that give really awesome and dynamic keynotes at big conferences are just naturally good at doing that. It could be that they’re used to it, and it could be that they have developed some good methods, but these people practice… And the only way to present yourself as confident and not very rambling, and not very wordy but concise and effective is to practice, and know exactly what you’re gonna say, and practice it over and over. Practice your slide transitions, practice, practice, practice, and not only in front of yourself and not only silently. Practice out loud, practice standing up, and find someone, even if it’s via Zoom or some remote way that you work, have someone listen to your talk and give feedback; take note of when you are rambling or maybe going off subject, and really dial those things in.

People think that some people are just really good at this and others aren’t, and there are people that it comes more naturally to, but for the most part I think the people that do a good job at this put in the time to practice.

I completely agree with that. One of the advantages of practicing a lot is as you go forward, it gets easier across multiple talks, because you will develop a style as a speaker, and you’ll come to be able to rely on your style to help understand how you’re gonna transition. It’s not gonna be your first, second or third speech, but eventually you kind of know how you approach a certain thing, and you know how long you need, and that will help. You’re still gonna be practicing, but it’s like anything that you do a lot - you kind of have that muscle memory effect, and that will happen with giving talks as well.

[56:08] A couple of big gotchas that I wouldn’t wanna close out without mentioning are a lot of times speakers will put way too much on a given slide. All sorts of data and charts and stuff, and it is compact, and the audience will be trying to figure out what each of the components on that particular visual are, and it detracts from what the speaker is saying.

I often tell people, there are two types of visuals - there are visuals for when you’re speaking, and there are visuals that you study when you’re not speaking, that you would sit down in your office and read and try to understand and analyze, and those are very different. So be aware of what kind of visual this is. In this case for speaking, in my view it’s much better to separate out the points, so that your audience is not confused, and then as you talk, to get to each of those slides in turn, so that you can relay that and it’s a lot less clear.

The other thing that goes with that is time discipline and self-awareness as a speaker. A lot of people that would otherwise be really good speaker, whether it be some level of natural, or maybe they’ve just practiced enough to get good at it, don’t start rambling. Don’t get into a topic where you have side things, and you let yourself be drawn into that, because it chews up time, and it is not fair to your audience. So really understand where you are, stay to your message, and always know when you’re supposed to be in your presentation, and make sure that your sequence of slides and the things that you’re saying for those lines up with where you should be time-wise. Brevity is your friend.

Yup. Some of the best advice I’ve got over time – and again, I think this really carries through, especially as we present AI information to stakeholders within our businesses… We need to be effective communicators about that. Some of the best advice I’ve got in terms of the things that have influenced my style of content is that – you know, someone told me once that for every slide, I should look at it and I should take out until I’m uncomfortable taking out words; I should take out as many words as I can and as much text as I can, until I absolutely feel like I can’t take out any more text. And then I should take out more text. That really needs to be a priority - making that effective use of each slide.

And then also I think a good thing that I do is when I’m preparing my slides, I zoom out from the slide until the slide is pretty far away on my screen; like, if I was sitting way back at the end of the conference place, or way back at the end of the conference room, or whatever it is… I make it super-small, and then see if it’s immediately clear what’s on the slide. If not, I need to make things bigger, I need to remove text, all that sort of stuff.

So this is challenging with the AI sort of things, where maybe you want to show this ginormous model tree, or something like that. Well, that is a challenge, but you should really take this sort of challenge seriously and try to remove the urge to create a slide with this big chunk of code, or this big graph that was output of TensorFlow, or something. Resist the urge to put that on the slide, because no one’s gonna be able to parse it anyway, so it might as well just not be on the slide. I think that’s some of my personal style.

We really hope that this conversation has been useful to our listeners. It’s a different sort of conversation than we’ve had on the podcast before, so we’d love to hear your feedback if this sort of information is interesting to you. Again, you can find out how to get in contact with us on Slack, or on LinkedIn at Changelog.com/community. Join our Slack channel, tell us about the conferences that you’re interested in, about your personal style of presenting, and how that differs from ours. We’re always trying to learn, and we’d love to learn from our listeners as well.

I hope this has helped everyone get a sense of how to approach speaking and conferences and meetups in general. If we’ve given you some tips on how best to get value out of each of these types of meetups and conferences, then we’ll have done what we set out to do here… So good luck, be bold, and just get out there and do your thing.