MLOps Zoomcamp: Free MLOps course. Register here!

DataTalks.Club

Home Articles Slack Events Podcast Books

The Essentials of Public Speaking for Career in Data Science

Season 2, episode 10 of the DataTalks.Club podcast with Ben Taylor

We talked about:

  • Ben’s background
  • AI evangelism
  • Ben’s first experiences speaking in public
  • Becoming a great speaker
  • Key takeaways and call to action
  • Making a good introduction
  • Being remembered
  • Writing a talk proposal for conferences
  • Landing a keynote
  • Good topics to start talks on
  • Pitching a talk to meetup organizers
  • Top public speaking skill to acquire
  • Book recommendations
Did you like this episode? Check other episodes of the podcast, and register for new events.

Transcript

Alexey: This week, we'll talk about public speaking. We have a special guest today, Ben Taylor. Ben is the chief AI evangelist at DataRobot. If you use LinkedIn, you must have seen Ben in your feed quite often. If you don't, you most likely saw him on one of the conferences or webinars, or somewhere on the internet. Ben does a lot of talks – at least one per week, if I'm correct. Maybe even more often. (1:32)

Ben: Yeah, it's a lot. I feel like it’s every week. (2:02)

Alexey: It was very difficult for me to think of anyone better for this talk than you. As we can see, you are right now sitting in a car. You’re ready to give a talk – to do public speaking – to be on a podcast in any situation – not only in your studio. This is also a very special moment, because typically people see you in the studio. But today is different. We caught Ben driving. Right? (2:09)

Ben: Stopped right now. So I'm being safe. (2:55)

Alexey: That’s good. Welcome Ben. Thanks for finding the time in your tight schedule to talk to us. Welcome. (2:59)

Ben: I'm excited to be here. I love this topic. I think it's very important for anyone that's listening. Anyone can become a great speaker. I’ve definitely been a bad speaker before. Anything you want to become better at, you can get better through practice. Practice is important. There's a few rules that I've learned along the way that are useful, which will hopefully short circuit people's learning. (3:08)

Alexey: We will definitely get back to these rules. But before we go to that – maybe you can just briefly tell us about your background. I know a bit about you because I follow you on LinkedIn. Maybe you can just tell us a bit about your career journey so far. (3:35)

Ben’s background

Ben: I studied chemical engineering in college. I planned on going to medical school – I'm really glad that didn't happen. I don't think I would have been very happy being a doctor. I went and worked for Intel and Micron for five years in semiconductors. This is semiconductor manufacturing. It was NAND Flash memory. I worked in photolithography, process control, fault detection, yield analysis. I got a really good sampling of applied data in semiconductors. Then I went to work at a hedge fund as a quant, building stock models, and then I went and joined a Sequoia Capital company called HireVue and I was their chief data scientist. I built out their data science team, I helped them launch their AI product. Then four years ago, I got the itch to go do a startup that I co-founded with David Gonzalez. That was a deep learning auto ML startup called zef.ai. We were acquired by DataRobot a year ago, and I've been working for Data Robot ever since. (3:54)

Alexey: I actually didn't know that it was acquired. I didn't know that part of the story. Congratulations. A bit late, maybe – but better late than never. (4:58)

Ben: Yeah, it's so funny going to do a startup. It's 100 times harder than you thought it would be. I feel like after going through that process, if someone's a founder, I feel like they're a brother or sister to me automatically. Even if they're a competitor. There's just a very special thing – a special place in my heart for people that understand the pains of payroll and delivering on contracts. (5:07)

AI evangelism

Alexey: What is evangelism? What does an AI chief evangelist do? (5:40)

Ben: Evangelism – that's something that a lot of tech companies have. Someone who's an evangelist, they could be a professional speaker, or be a frequent speaker where they're out speaking. For my role, it's not just evangelizing the Data Robot products, but figuring out how to evangelize AI. So it’s “How do we get companies excited about using AI, even if they don't want to use our particular product? How do we get people…” You know, thinking about the art of the possible. “How can we stir the creative juices and get them thinking about ‘What could they do this year with AI?’” For some people, I joke and say, I'm the “AI missionary”. If you didn't think your company needed AI, there's a good chance I can convince you otherwise – with the right format. (6:04)

Ben: To be a little bit more specific. I see evangelism sitting in between product and engineering and marketing. Marketing for a tech company is very complicated, because a lot of the marketing can be technical. You never want to have marketing that feels misleading, or it's incorrect or embarrassing. Because for people that are traditionally marketing, they might not have the technical background to get the messaging right, so a lot of times I'm reviewing blog content, and some of the messaging stuff. For me that's a lot of fun, because it feels like it's a tough challenge. If you and I are competing on writing a blog article about AI (or anyone listening) if we're competing together, how can you – and this all goes back to speaking – how can you create content that changes an audience or moves them emotionally? The criticism I'll throw out to end this rant is – the common criticism, is that people fall into the rational mindset and that's the features and speeds. So it’s “I want you to like my product, because it has one more feature than the competition,” or “I want you to buy my hardware because it's a little faster than NVidia or Intel.” That's a very weak argument. Because humans like to make emotional decisions and then they like to confirm their emotional decisions with rational insights. The reverse doesn't really work. To just sell you on rational arguments. Yeah, I love studying that part of it. (6:56)

Alexey: So, your role is to speak in public – to promote AI and to promote your company as well. Can you also say that you’re kind of an editor in a blog? So you also review posts? (8:34)

Ben: Yeah. We're also starting up a program this year where we're going to be doing active research. I'm still programming, I’m still doing AI applications. But normally, when you do R&D, it answers to products and it answers to customer needs, where this will be very different. I'm very excited about it this year, where we're doing R&D that answers to the attention needs of the market. So it’s “Can we do an AI application that is inspiring?” We need an AI application that's relatable. Can we do an AI application that furthers this cause of “everyone needs AI?” That's a very exciting thing for me to be working on this year. (8:51)

Ben: The last thing I'll say is, really all of this is about understanding of process. It really isn't about me speaking, it's about “Can I understand the recipe, or the process, for developing an excellent keynote?” or training someone else to go give it instead? That's where things get really exciting because you always want to… You hear people say “nail it and scale it with product” but that's also true with individuals – it's true with yourself, and it's also true with marketing. If you can nail a process, then you can critique it, you can measure it, and you can scale it. So my obsession is around scaling. (9:37)

Alexey: A way of scaling it… I don’t know if it's a bit ambitious, but being on this podcast, and then talking about public speaking is kind of a way of scaling and educating other people how to do public speaking? (10:18)

Ben: Yeah. So for anyone listening, if they can go give a better talk about AI that motivates an audience and changes them, that honestly helps me. (10:38)

Alexey: Do you know if there's any difference between evangelism and developer advocacy? Are they synonymous? Maybe they’re a bit different? (10:50)

Ben: Maybe there's some overlap there. There could be some overlap. Some of these titles are a little funny. Because what do you actually do? I think you can look at different companies that have evangelism roles and what they actually do could be very different between the two of them, (11:01)

Alexey: It's like data science, where you don't really know what a company means by “data science” until you actually see a job description and talk to them and figure out what they actually need. (11:20)

Ben: Exactly. I do like thinking of evangelism as fitting in between marketing and engineering. They have to know both worlds. For me, I really liked being stuck in the middle. At HireVue, I was stuck in the middle, between the data science team and the IO psychologists. The IO psychologists, they're the ones that develop hiring assessments – you're gonna go fill out this assessment to go work at the bank or something and they want to flesh out your personality traits and map that to potential performance in the future. I love being in the middle. When you're in the middle, you can understand both sides. You suddenly have a very useful perspective. If you're just in marketing, you're not going to appreciate engineering, and you're not going to even understand how to communicate it. If you're just in engineering, you're going to fall prey to the rational mindset and a lot of factoids that don't matter emotionally – they don't move the needle. By living in the middle… maybe, if that's the theme for me, I just want to spend my whole career living in the middle of whatever that is, whatever the next thing is, I just want to be right in the middle. Because you get a very unique point of view. (11:34)

Ben’s first experiences speaking in public

Alexey: Now you give one talk a week, right? Even more often, maybe. But I can imagine that this wasn't the case for you all the time. Do you remember your first experience speaking in public? (12:54)

Ben: Yeah. I think it actually started when I was at Intel and Micron. I remember – I was so excited. If you want to go back in time and find the worst talks that Ben Taylor has ever given on planet Earth, they were inside Intel and Micron. I remember I was so obsessed…. I had a math minor and at one point in my life, I thought I wanted to be a mathematician. I was so obsessed about some of this math regarding controllers. This math is pretty intense – it's state space, linear algebra. I remember giving a talk about some breakthrough on a certain type of controller for state tracking – and the math was intense. Imagine me sitting you down to walk you through an hour of math. You and I are technical, but even today, it'd be like, “I'll pass. No, thank you.” I remember giving that talk and I think I wrote the whole talk and it was in Beamer or low tech. It was just awful. You should never give a presentation in Beamer. (13:09)

Ben: So I'm giving this talk and I remember looking over and there's a senior engineer at the end of the table. And these rooms are dark – we have the lights off – and he's asleep with his head cocked back and his mouth is wide open – just sleeping during my talk. At the time, I think I took offence. I just thought, “Oh my gosh, I can't believe this individual is missing out on this good content I'm sharing.” But today I would say “Yeah, I gave a terrible talk. I didn't understand my audience and it was awful.” Today I care a lot about the audience feedback, because it's a point of reference. “Did you give a good talk?” “Could you have given a better talk?” You need to consult with the audience after the talk, to find out. You're not going to make everyone happy. You're always gonna have some people — no matter how good of a speaker you are — you're always gonna have some people that didn't like the talk or they've got some criticism. The challenge that I fall into is – I've been criticized for giving talks that are too technical. And I've been criticized for giving talks that weren't technical enough – and I've had those criticisms on the same talk. That's an example of like, “You're not going to make everyone happy, but there's always room for improvement.” I've learned a lot since then. That was eight years ago – making an engineer fall asleep. I gave plenty of terrible talks since then. But my terrible talks are becoming less frequent now. (14:08)

Alexey: So that was the time when you would give one talk a year? I guess it happened to you in seven, eight years? (15:44)

Ben: I started to speak at HireVue – they helped my speaking career. When I started at HireVue, they needed someone out there talking about AI and data science in the HR space. I started going to HR conferences and I think I could have made a name for myself right out of the gate. It's interesting, because right out of the gate I made a name for myself, maybe not because I was a great speaker, but because I was controversial. I was rough around the edges. I remember presenting a PsyOp – I think I was one of the only data science speakers there, and I was saying a lot of things that were pretty… I don't want to say disrespectful… But I remember I gave a talk to an audience and I'm essentially telling the entire – I'm insulting the entire audience. (15:57)

Ben: That's a rule of storytelling. In speaking – don't insult the audience. But I remember insulting the entire audience. I said, “You guys don't understand statistics.” Essentially, I'm insulting the entire industrial occupational psychology field. I'm saying “You guys don't understand statistics, so I'm going to teach you how to do proper cross-validation.” This talk was seven years ago in Philadelphia and I think I'm showing them how to do K-folding. Because I found out they weren't doing that and I was upset about that. I remember… and I'm not thinking I'm being a dick, I'm not thinking I'm being mean – I just give this talk like, “Hey, you guys need to know this. You're welcome.” And I remember seeing… I ran into someone six months later who had been at the talk up in Park City, Utah, and they were saying, “Hey, remember that talk you gave in Philadelphia?” I'm like, “Oh, yeah.” I'm thinking a compliment’s coming, I go, “Yeah.” He said, “Yeah, you were a real ass.” I was like, “Oh! Okay, tell me why.” He essentially said – it was interesting – because he agreed with everything I said… He said “You were right about what you said,” but he said, “We didn't like hearing it from you.” And part of that – I don't want to gossip or go down another path – but it's essentially saying, “I'm not an IO psychologist, I'm an outsider. I don't have a PhD in IO, and I'm essentially raining on their parade and insulting their career.” (16:47)

Ben: I'm not a controversial speaker anymore. There are tricks in storytelling where you actually can offend the audience, but you have to be really careful how you do it, and you actually don't ever want to leave a talk where the audience still feels offended. To finish the thought – at HireVue, I started doing a lot of speaking. I was speaking all over the US, I got to speak in Sydney. Then when I went and did our startup – I started speaking even more. I spoke in South Africa, Madrid, Dublin, all over the US. (18:12)

Ben: I got to the point where I started to get invited to go speak to some impressive companies, like, Red Bull, Goldman Sachs, Procter and Gamble, Amazon – different companies. Like, a rocket company says, “Hey, will you please come present to us?” And those invites are coming because some of those employees had seen my other talks. For me, that has been some of the most fun I've had – giving private talks to these companies. You meet some really interesting people. It's important for the audience to know – there's nothing about me, there's a process here. It's not like I'm better looking or my voice sounds better than anyone else. There's nothing about me specifically. There's a process here for anyone listening to become a better speaker and to become a great speaker. (18:52)

Getting attention

Alexey: Let's jump into this process. So what is the process? (19:48)

Ben: You have this concept of attention. You walk out on the stage. You being the speaker, you're given attention for free. But you can quickly lose it. You can imagine if you or myself, were walking out on stage – COVID’s over – we're gonna walk out in front of an audience of 1000 people, or even 100 people. You all have their attention, but they are going to quickly decide if they're going to pull out their phones, their laptops, or disengage. There's lots of sub topics. One of the sub topics is around the first impression. How do you introduce yourself – what's the best way to introduce yourself? But getting into the meat of the talk, I think it's important to have a talk that leverages storytelling. You jump in with vivid details, you have a story that's engaging. Think of your audience as being like cold taffy – you have to warm them up. You can warm them up with humor, you can warm them up with storytelling. If you can warm them up, they're going to be much more receptive to the key points you want to share. If I had to simplify the talk, the beginning of the talk is the warm up. And that could be storytelling, it could be something that's more creative – maybe something that's silly, nonsensical, surprising. There's a lot of emotions you can play with. (19:54)

Ben: When I would design my talks, I used to write the emotion on every slide that I wanted the audience to have. The emotions could be concern, relief, anxiety, happiness, sadness – they can be all the emotions you can imagine – and by writing it down on each slide, it would allow me to amplify it. What is the goal for the felt audience to experience? I think of it as being a monotonic speaker – no one wants to listen to a monotonic speaker. In the same vein, I'm not going to warm up the audience if I just have one emotion. If I have this whipsaw, and warm them up – and humor can be a very powerful tool for that. (21:11)

Key takeaways and call to action

Ben: It's also really important to have key takeaways. If I just have a talk that is storytelling, warming them up – then the end, you might say, “Well, what was the point of this talk?” And that'd be a fair criticism. “What is the point of this talk? Is this a storytelling festival? It's supposed to be a tech conference.” You need both. It's really important to decide what is the gift that you will give to the audience? That probably sounds arrogant to say. You need to give the audience something for free – some learned insight, some key points. If they never see you again, they need to be able to leave and say, “Thank you. That is useful for me.” You need to identify what that is. Identify the warm up, identify what the key takeaways that they can have. It's hard because there can't be a lot of them. I can't have 10 key takeaways – people don't remember. The other thing for talk is — they need to remember it. If I give a talk, and everyone says, “Oh, that was a great talk.” And they wake up the next morning, “Oh, I forgot what the talk was about.” It wasn't a great talk – that was a terrible talk. Decide what the key takeaways are. There's other tricks you can have in the talk where you can have a call to action, you can have something at the end where you try to engage people to go to a certain place, go watch something or reach out to you specifically for further questions. Sometimes people forget the call to action. (21:55)

Alexey: You mentioned warm up. When you get to the meat of the presentation, you need to use storytelling. Then think which kind of emotions you want to appeal to. Then you really need to think of the key takeaway messages, there should be very few of them, 1-3. At the end, there is a call to action – what do you want the audience to do? What do you want them to do after this talk? These are the main elements of a talk. (23:28)

Introducing ourselves

Alexey: Imagine I go give a talk and at the beginning, I start saying, “Hey, I am Alexey. I'm a lead data scientist at OLX, blah, blah, blah.” Is that a good way to introduce myself? Is there a better way? (24:01)

Ben: Yeah, there is a better way. There's different tiers of expertise on how to introduce yourself. (24:17)

Ben: You can introduce yourself as just your name. I could say “My name is Ben,” or “My name is Alexey, I’m going to talk to you today.” You haven't given the audience a reason… “What are your credentials? Why should I listen to you? Are you some random homeless person on the street? Who are you? Why should I listen to you?” The reason it's important to give them a reason to listen to you is – you can think of it as pulling the band in storytelling – they're gonna lean in. You get on stage – should I have my phone out or put it away? Should I listen to you? If you introduce yourselves with some credibility, then I want to listen to you. If you say, “I have this much experience.” Most people do a resume overview, like, “I worked here, this is my title. This is my experience, I'm gonna give you this talk.” And that's not terrible. Most people do that. I've done that for a lot of my talks. (24:46)

Ben: A better approach, which is very difficult, is – you would immediately jump into a story where the audience concludes that you are the hero of that story. It's very interesting, because if you force conclusions for the audience that weakens the point. Imagine if I jumped into a hero story – right there, in the action, I'm talking about an AI problem I was solving or some impossible problem that I'm up against. And then if at the end of that, I say, “And therefore I'm a great data scientist,” – what just happened to your confidence? Now, I'm actually crossing the line into arrogance, but if I tell a hero story, where there's chaos, there's an opportunity to fail and then there's success. That can actually… what I've done is I've taken you on an emotional journey. If I just do a CV overview, that's a rational argument. The funny thing is, if you win on the emotional journey – if I tell you anything else, it just confirms your first impression. (25:44)

Ben: If I tell you, “I was the chief data scientist at this company,” or if I tell you, that “I did a startup,” it just confirms the first impression. But if I try to start with a rational approach, I'll only be able to feel the line of “If I'm lucky, I'll get too smart.” The other approach – you can actually exceed smart. If you exceed smart, you land in hero territory. If you're in hero territory, then the perceptions are very different. Then you're perceived as being in the top 1%. Which is interesting, because you're not being dishonest. You're not lying. You're not trying to pull the wool over people's eyes, but you're entertaining them with a story. (26:57)

Ben: The last point I'll add is – the hero's journey is the story that's been told for 1000s of years through all of society. If you look at these Disney cartoons, or Pixar cartoons, a lot of them are templated after the hero's journey. It's almost like there's something in our DNA, where humans will celebrate any hero that has some element of risk. It usually comes back to them achieving some value, not just for them, but for society. We benefit from a hero – if you're a hero, I benefit, and vice versa. (27:37)

Alexey: That's a pretty high bar. I'm wondering – maybe you have a 30-second example, just to understand how that might look like? (28:13)

Ben: It is a high bar, and it's something that I don't expect people to do. There's an important alternative that you can do, and that is having someone else introduce you. A lot of times you'll have a chair at a conference that will introduce you. But I would say they don't do it well. They're reading your bio. Imagine me – I look at your bio, and I'm “Blah, blah, blah.” Or you're doing it for me, just reading the bio – that doesn't feel authentic, there's no emotion to it. But if I read your bio through a storytelling lens, and I say, “I'm so excited about the next speaker!” You probably heard this before at a conference, but it really isn't something that should be laughed at. If I say, “I am so excited about this next speaker, Alexey. He is one of the best,” and go down the list. It changes everything. I can actually get you into hero territory, just with my introduction of you. If you introduce yourself, or if I introduce myself, we can't get that high. To go back to your question – to tell the hero's journey, I might pick an AI project in the last 6 months or 12 months, where I've nearly failed. I might just dive into the story. In reality, it's about maximizing attention. For most of the people in our industry, they're very technical. A lot of them tend to be more introverted. They're not going to dance on stage or scene. Mastering the storytelling techniques is not something that most of them feel comfortable with. But I want to kind of plead with the audience that if you do – if you master these techniques, it's game changing. It can be life changing. (28:23)

Ben: You can make an audience cry talking about AI – which is the craziest thing. If I told you like, “Hey, I’ll give you $100, if you can go find someone in the front row that dropped…” and I'm not saying I've succeeded at doing this, but I'm saying it is possible. You could go and give a keynote, and find someone who you've been able to push past that emotional barrier in a way that feels authentic. A lot of times we talk about the technical gift – like the factoids – but I would also offer with a really good talk, there's an emotional gift. You've given them a perspective, or a point of view, or an experience that they're grateful for, and they'll remember how they felt during that experience. (30:08)

Storytelling, crafting a story

Alexey: This sounds almost impossible for somebody like me, who is a technical person. I'm a data scientist. I'm used to just using facts, and then I think, “Okay, I am a data person, and I like data. I will just show a lot of data to others and they will like this data, and they will get convinced.” (30:57)

Alexey: How can somebody technical like me, or a developer, a data scientist, or some other engineer – a technical person – how can they go from this data-based presentation to something story-based? How do we craft the story? (31:22)

Ben: I would say that leaning too much on the data can handicap you. The standard data science presentations that I walk through, will be: They’re talking about one project they did. They're gonna have 10 or 15 slides that go through their approach. The approach is not useful if I'm an executive. If I'm a data scientist – I'm going to critique you, fine – I can go through the approach. I like to tease data scientists that do this style of presentation by saying they're fishing for partial credit. Or they're trying to remind you that they're smart. (31:45)

Ben: I'm not criticizing – I'm actually criticizing myself. If I go back six or seven years, I remember giving some presentations to HireVue executives, who were during the presentation – the CEO would interrupt me and say, “Ben, we know you're smart. That's why we hired you. You don't have to keep reminding us.” That's not a compliment. What that's telling you is – I'm not communicating. Executives, they got so much bandwidth and I'm not communicating. Sometimes going through the data can actually hurt your ability to communicate. But you still want to have some data and a rational conclusion. (32:23)

Ben: For the people listening – storytelling is not just about giving a presentation, storytelling can be used at a strategic dinner – you're with your potential customers, with customers, with prospects. You're at a conference networking with people… your ability to tell stories can impact you on the personal level. It can impact you as a speaker. It's a very important skill set. If you're a parent and as a parent, our kids want us to read stories at night. You can read stories at night, but I would encourage you to tell stories at night. You and I, we've been on this Earth for a while – we've made some mistakes, we've been dumb kids before. I noticed that me telling my kids stories at night – they love my stories, way more than they love me reading a book. Some book about a hen or some chicken or something that does something. One of the things I'd encourage people to do is tell your kids stories at night about you, or about things you've done and try and practice some of these storytelling tricks and see if you can captivate their attention. (32:58)

Ben: If you tell a story at night and if your kids are not asking you to tell another one – you've got work to do. But if you begin to become a better storyteller, where every night your kids are begging you for the second story, or “Tell another story, please!” You're making progress. That might sound very silly – that you practicing telling your kids stories at night is going to make you a better speaker. But it's true. You can also read books – there's stories that stick, there's “The Hero with 10,000 faces”. (34:12)

Ben: One last point. I love that this feels very childish. It feels like storytelling – hopefully people in data science are rolling their eyes, like, “I'm grown up. I don't have time for storytelling,” BUT Pixar – how much money do they make with these movies? Pixar makes hundreds of millions of dollars with these movies. This isn't a guess. It's not a game. This is a science. When Pixar comes out with a new movie, they're going to invest hundreds of millions of dollars, sometimes, in this production – this movie better work. It better get that type of return. If you look at it that way – if you're a data scientist working for Pixar or Disney, these movies are gonna make money and storytelling works. It's a science. It's not some shaman or some goofy person saying, “You know, get on stage and tell a story.” There's real business attached to this – significant business. It's an opportunity for anyone to learn from it, leverage it, use it. In the end, you're trying to maximize attention of your audience. The saddest thing is – what if you get to your key point – we talked about two to three points you want to share with the audience – what if you get to your key point at the end, to share, and half of them are on their phones? Half of them are checked out. Even the ones that think they're listening – they're not actually listening. They're daydreaming – they're giving you like 30% of their focus. That's too bad. That's sad. You actually had a key insight to share, but the catalyst to transfer that is dull or dead. (34:45)

Being remembered

Ben: You go to the data conference, how many talks are forgettable? (36:24)

Alexey: Most of them? (36:29)

Ben: Most of them. And that’s not an insult. It's just true. Most of them. Most of the talks are forgettable. How many of those people were smart? All of them. They're very smart, they're very accomplished. Unfortunately, most of the talks were forgettable. The challenge – make sure your talk… I think you mentioned impossible a few times and Ilike that, because I like having impossible goals. When I go to give a talk, there's a lot of little talks that you just have to give – you don't have a lot of prep to do. But for the big talks, big keynotes – the most important talks that you're looking forward to – my goal with those talks is when I give the talk, I want it to be the best talk I've ever given in my life. And that's not always possible. That's an impossible goal. But sometimes it happens. I also want the talk to be remembered for five years. What I mean by that is – if you attend the talk, if I never see you again. But if I run into you somewhere traveling five years later, you can say “I remember your talk at that conference.” (36:31)

Ben: The third part would be – if there is audience feedback, I want to be the number one speaker for the entire conference. From audience feedback – not from speaker feedback, not from conference organizers feedback – from the audience feedback, I want to be the number one best speaker. I've accomplished all three of them individually at different times. I've had conferences where I've been the best speaker. I've had talks where people remembered it for five years or longer. But to have the goal to accomplish all of them… I obviously have one talk somewhere that was my best talk ever for someone. But for me. I have the best Ben Taylor talk. I don't know what that talk is. But to have that as a goal, as you're getting ready to go into these keynotes. It's an impossible goal, but I think it's a goal that everyone can have. (37:36)

Ben: You're gonna go give an important talk in the next six months. But for people starting out, that could just be a local meetup. They've never given a talk before, present at a local meetup – obsess about it. Try to make this the best talk you've ever given. Try to make sure you can do something so this talk will be remembered for five years. If there is audience feedback, and there's other speakers competing against you, try to be the best talk. I'm not the best talk all the time. I'm sure I've been the worst, like bottom tier on some conferences. You can learn from it. (38:24)

Alexey: Going back to the question I asked. You told us a story that some executive at some point told you that, “We know you're smart. But here you need to make it digestible for us.” How did you go from Ben Taylor back then, who spoke in front of executives and was not able to communicate to them, to Ben Taylor now – who gives talks that they remember for five years? Is it giving your kids stories? Or is there something else like people can do to come from that state to the state where you are now. (38:58)

Talking to executives

Ben: That's tough because a lot of it just comes down to experience. Even that example I gave – when he said that, I was confused. I don't think I had a follow-up action. I probably didn't learn from it for a couple years. It wasn't until I went and did my own startup. Because one of the things I was doing — I would try to spoon-feed, or teach, the executive statistical facts, like, “Hey, I want you to understand why an AUC chart is so nice,” or “I want you to understand what I had to do, because your data was so dirty. I want you to appreciate…” and my thought is – if I give you enough of these little factoids, you're gonna fall in love with it, just because I'm in love with the stuff – you're gonna fall in love with it and we’ll all be better off. That was a very naive perspective. What I didn't know is the CEO is living on a plane, trying to save these accounts – very large million dollar accounts that could be churned or lost and he's constantly up against these quarters and arguing with board members, and firing and hiring. When it comes to mental bandwidth, he has no mental bandwidth. I didn't know that. There is an experience that you just have to make. (39:55)

Ben: Hopefully, you can learn from other people's failures. Try to follow people and learn from people. For anyone that's put machine learning in production, they've screwed up – at some point – they've had something go wrong, the model didn't generalize, there was feature drag, version control, they went to go retrain the model, and the original training set was gone, it was deleted. They made mistakes. You can learn from other people. Try to follow people that have made these mistakes. I wish I could have learned these lessons sooner. But part of it is just being hit over the head enough. (41:11)

Ben: The last thing is seeing the world through other people's shoes, or from their perspective, is useful. In the moment, if I had actually really understood the horrors and the stress that a CEO deals with, my talks would have been very different. Or if I knew how much anxiety my Amazon bill was causing them… Because from my perspective is like, “Oh, it's the cost of research. Grow up. It's gonna cost you 1000s of dollars per month, and Amazon burn, if you don't like it, you're apparently not ready for AI”. It's way more emotional. For them, it's very upsetting. These are all the things the CEO was thinking about you but they didn't say. And hopefully, if people realize this, it will scare them into doing the right thing. As a CEO – I'm not a CEO, but now I'm turning into a CEO for a second — Ben Taylor's is a CEO. You sit down and you're going to present for 30 minutes or an hour. You schedule an hour with me. The more you make me understand the data and the steps you took to get to success, the more I want to fire you. (41:52)

Ben: Why? Because I'm confused, you're wasting my time, I don't get the point of this. What's the recommendation? If we could have a 30 minute meeting, where you say, “These are the final results. This is the team that did it. These are potential concerns. These are potential pros. This is my recommendation to you.” A lot of times we, the data science team – we show them all of this work. They're trying to understand it. They're confused by it. We see it as validation that we're smart – validation that we're working hard for them. But at the end, a decision has to be made. Is this model ready to go in production? What are the next steps? A lot of times we rely on the CEO to make that decision. But they hired you to make that decision. It's better for you to give a recommendation and for them to disagree with it. Or for them to challenge it. (43:10)

Ben: Imagine me saying “I recommend that this model is ready to go on production and we should sell it.” or “We should sell it to the customer.” Now there's an opportunity to say, “I don't trust you.” Great, let's talk about it. The saddest part is — that meeting we just talked about 60 minutes was wasted going through data. Instead, we should have spent 60 minutes talking about “I don't trust you.” You don't? Why? Let's talk about it. Why don't you trust me? That's a much more productive conversation. Because in the first scenario – CEO is confused. No real actions came out of the meeting. They're wondering when you're going to be productive. They're wondering if they should fire you. They're wondering if they made a mistake. They don't know if this is a good idea. They know they need to invest in AI because they're supposed to. There's just a lot of terrible things they think after that other meeting. This meeting is radical candor. Right to the point. Your recommendation – no confusion, no jargon. I've been this data scientist, I've been the first one – I've been the bad one. Most data scientists fall into the bad category – very technical. They can communicate with each other, they can't communicate with an executive. Where now, in my career, I love talking to executives, because they're so focused on value – focused on growth. You can talk to an executive and not use a single word of data science jargon. How do you learn it? You just go through the meat grinder. (44:02)

Alexey: I’ve heard about this “pyramid principle” – when you start from the recommendation, and if need be, you go down deeper in the technical implementation and whatnot. But you start with the results. And then, if needed, go down. (45:41)

Ben: You should have an appendix. If a CEO says, “Did you guys try anything else?” you say “Yes, absolutely. Jump to the appendix.” The thing that is very top of mind for me now is this idea of time efficiency. If I could scream at the top of my lungs – “time efficiency,” For an executive, time efficiency is everything – get to the point. Tell them what they need to know. The thing I'm realizing now – if you are having a conversation and if the CEO of the company is not mission-critical — we're talking minutes — within minutes, if they're not critical for this conversation, they should leave the meeting. That's true and I know Elon does this – if he realizes he's not needed for this meeting, he just leaves. He does that all the time and I think most CEOs should do that. Most CEOs should be rude. Let me rephrase that. Most CEOs shouldn't be worried about your feelings – they should do what's prudent. They should do what's useful. They should make the best use of their time. Even if you're a very senior person in the company – if I'm the CEO and I walk into the meeting, and within a few minutes, I realize that you don't need me or if this is not productive, I might be kind enough to say, like, “Hey, I don't get the sense I'm needed, I'm going to leave this meeting.” Which is really like, “Hey, you screwed up. You have an opportunity to save it, or tell me why I should be here. And if you don't have a good argument then I'm leaving.” Most CEOs should act like that. Because their time is so valuable. (45:58)

Getting to speak on conferences

Alexey: We talked a bit about talking to executives – how about giving talks at a conference? You said you should set the bar high, give big talks, try to be remembered for five years. I have a few questions here. The first is – how do you get to a conference? To get into the conference, you need to send a talk proposal. How do you structure this proposal? How do you write it in a way that gets you into the conference – so you get to speak there? Because sometimes the committee rejects your talk proposals. At some conferences it’s hard to speak. You need to write a good proposal. Do you have any suggestions? How should I approach that? How should I write my talk proposal? (47:38)

Ben: For people that are starting out their speaking career, it's easier for you to go land a local meetup, because you might be able to actually meet the meetup organizer. They're always looking for interesting talks. If you can work with a local meetup organizer, get a talk there and actually get a recording – something on YouTube – that'll give you something you can share. Then you have a talk of yourself. You can even imagine a scenario where you could give a talk to no one like, “Hey, everyone listening” You'd organize a talk, you could go record a talk, post it to YouTube. You could share that – trying to pitch yourself into a meetup or pitching yourself into a conference. That way they know that you can give a talk and it's not gonna be terrible. I do like to understand the conference organizers’ priorities. They have different priorities. You doing a bag of words, NLP, they might be like, “Oh my gosh. No, thank you. We've already had that before.” But if you’re doing a talk around reinforced learning, or BERT or something… Definitely understand the conference organizers’ needs. Sometimes I'll tease out a few topics. I might give them two or three topics. It's also important to understand who the conference organizers are, because they're really the decision makers. The conference is their baby and the conference has to be good. In conferences it's almost like your resume. In a work resume, you say “I worked at this company, I did this project.” And you have a speaker resume “I went to this conference. Here's the talk. I went to this other conference.” You can kind of level up from being a speaker, to a featured speaker, to a keynote speaker in different conferences, and that's super useful. (48:40)

Ben: You need to fight sameness. You need to run away from sameness, because humans are novelty-seeking creatures. We remember things that are unique. If we walk through the forest, we're going to forget everything except the bright red mushroom or the new animal that we saw – we've never seen a rabbit before, we saw the rabbit, we're gonna remember that rabbit forever. I think humans are so funny that way. When you submit your idea, try to be creative. (50:20)

Ben: The other thing I'll tell your audience – you can actually scare yourself with a proposal. What I mean by that – I've talked about reckless commitment. Imagine you're a conference organizer and I'm saying, “I'm going to give a talk on this.” And this is some ambitious topic. This is something that I'm not qualified to give a talk on. But if you're the conference organizer and if you bite at this, then “Oh, shit, now I've got three months to bust myself to make that happen.” That's something I would encourage people to do. The joke is, if you land yourself on the schedule to give a talk on any topic, that's months away – you will be an expert by the time that talk comes. You might cry at night, you might lose sleep over it, you might read a lot of books. But by the time you give that talk… I think there's also this concept of “imposter syndrome.” I think that can shut people down. I remember giving a few talks – I gave a talk in San Francisco once and there was someone in the front row that was asking some question about different norms on my cost function or something. I think sometimes, as a speaker, that can shut you down. (50:53)

There’s always someone smarter than you

Ben: For people listening, I like just saying – guaranteed, there's someone in the audience smarter than you. Don't try to be the smartest person in the room, because that can backfire. During Q&A, if someone asks you a question – if you don't know the answer, or if you think the question is embarrassing – the temptation is to try to answer it. That's really bad. Don't try to answer a question you don't know the answer to. If I ask you a technical question, there's a good chance that the textbook question is so technical, and you don't know the answer – it's actually not a lot of value to the audience. Sometimes you get people in an audience where I want to prove that I'm the smartest person in the room. You give the talk. I'm in the front row and I raise my hand – I'm asking you a very technical question. You don't know the answer to it. It's not a valuable question. No one in the audience cares about this question. It's just me being a jerk. Because I was invited to speak, I want people to know I'm important. So that's something that's interesting to get over – you're never gonna be the smartest person in the room. Just give up. If you ask a smart question, if I don't know the answer to it, I can say, “does anyone else in the audience know the answer?” or “let's talk about this after. I'd love to hear your perspective.” And then go on to the next question where you actually can answer it. Start speaking, get past the parts where you feel intimidated – everyone's gonna be nervous, at the beginning. Q&A can be terrifying, but not if you decide that there are smarter people in the audience and you decide that you won't pretend to answer a question you can't. And it's okay. (52:13)

Giving a keynote speech

Alexey: You mentioned that you shouldn't aim straight away to a keynote at Strata or something high profile. That it's okay to start with local meetups and then gradually build your portfolio, build your “speaker CV”. This is how you get on conferences — by speaking first at local meetups. This is how we get on Strata – by first speaking at local conferences. You get on keynote, by speaking at conferences. Is it correct? (53:48)

Ben: That is right. I'd also say – nobody could land keynotes at these bigger conferences, because you need to have a track record. You need to have a personal brand. Or maybe an employer's. a lot of times for those keynotes – the keynote has to be a draw for the conference and so the keynote might be a specific brand, or an individual who's built up a name for themselves or a company. They really want some director from this particular company. You don't have to work at Google or Facebook or these bigger companies to land a keynote. You can land keynotes by working in a very, very small company if you've built up a reputation, because they want the talk to be good. If you've proven that you can be a good speaker and if you're getting momentum, then you can begin landing feature talks and keynotes. Just asking to be a keynote speaker, the default answer will be “No,” until someone builds up their brand – until they get this track record or until they have this role, where they can do that. You've probably seen keynote speakers too, where they're awful – they're bad. That's because they're being pulled in because of a company brand. Like “so and so at Google” or “at this company is going to be slotted in.” If they’re not a good keynote speaker, then the decision was made because of the company brand. Or it could be a sponsored talk – sometimes those keynote talks are actually paid for by the company that's giving the talk. So… roll the dice on if it's good speaker. (54:25)

Topics for talks for new professionals

Alexey: Eric is asking: “As a new data science professional, what topics and groups would be good to start with? I'm not a deep learning expert, but I understand business and data.” (56:37)

Ben: Everything in life falls into a Levy distribution. It's like the normal distribution with a long tail. Most of data science calls this normal distribution “usefulness”. Think of Pandas data frames, Bayesian methods, building models, like Sklearn data frames, just the Python foundation, Docker – is all useful. All of this is great. On the long tail, you've got deep learning, reinforced learning and these things are very sexy – very exciting. A lot of people getting into data science, they say, “I want to have deep learning and do video games, I want to do this.” But a lot of the jobs, and a lot of the value in the work to be done, it's not that. You will find use cases, but they're quite rare. A lot of the times AI moves the wrong direction – it's AI looking for a problem to solve and that's dangerous. Because sometimes you find the wrong problem. Whereas it's much more powerful if you have a problem that's being backed into the right tool set. For people starting out, whether you're currently employed – try to find a business problem that's big, it's important, it matters. If you can automate it, or augment it or get some insight here, it will impact the company, and then try to back into the appropriate solutions. Then there's a good chance deep learning is not the appropriate solution for customer churn or click-through rate or something related to the business that matters. It could be something that's handled with just structure data – and you've got options for that that are very easy. (56:55)

Ben: One of things I like to bring up is – everything used to be hard. 10 years ago, everything was hard. Today, a lot of this stuff is very easy. If you have a structured data set, you want to build something, this isn't a research project that should take months. This is googling, or using pre-built software. These solutions should be very easy to do. Start there – find the problems first, back them into the right solutions, leverage the data science community to help you. Go to data science meetups, you'll be very impressed with… You have two different worlds between academia. Coming right out of school, and then the applied meetup space. There's a lot of behaviors and skills over here that don't do not exist over here. Definitely try to become injected into the applied meetup space and you'll find peers that can help you. That was a much longer answer to a short question, but it was a great question. (58:34)

Pitching a talk to meetup organizers

Alexey: Eric is also asking – I think this is a follow up question to the previous one. “How would you pitch to speak at an event like a meetup?” I think we briefly covered that, but let’s use the same scenario. You take something relatively straightforward, a business problem, you use a tool like Scikit-Learn to solve this problem. You have the solution, how do you pitch to meetup organizers to talk about this solution? (59:32)

Ben: It comes back to first impressions and introductions. One of my favorite stories – Rico went to Data Science Go in 2017. He loved the conference. This is a young kid, he was 20. At the end of the conference, he walks up to the conference organizer and says, “I'm going to present here next year.” The conference organizer in his head is thinking, “No you’re not. You're some beginner, new Junior data scientist, and you think you're going to present here?” But the kid was able to. The kid was unable to in that short conversation, but the kid ended up doing a lot. The kid went out there, started a meetup, did all these things where the organizer kept track on this kid and thought, “Holy crap! This kid is out of control, in a good way.” The kid ended up presenting the next year and gave a great talk. First impression matters. “Who are you? What company do you work at?” You don't need to have that to back your reputation, you can also get endorsements. Imagine if you and I know someone in the community, and we endorse them. We say like, “Oh, this person is really good, they're up and coming, you should definitely give a…” Actually, I had a neighbor of mine, a young kid, in college. He was just doing research in college and he presented a local meetup. I recommended that he should. I pitched his idea. You can leverage other people. Networking is so important. They don't teach you that in college. The people you know can really help get you that first talk, especially if they trust. It's funny because sometimes we get people asking for recommendations where we don't actually know them. That doesn't really work. But if we know someone, if you meet someone, you're more likely to take some risk and say, “I think this person would be good, you should have them present.” That can go a long way. (1:00:06)

Top public speaking skill to acquire

Alexey: Matthew is asking “In your opinion, what is the top public speaking skill that people should reach for?” (1:02:18)

Ben: Storytelling is a big umbrella, so I hate to just say “storytelling,” but the biggest thing they should reach for is storytelling. To slant that a different way – you should have a goal to maximize attention. In data science, we're used to this – like, “optimizers, convex problems / non convex”. We like optimizing. Treat this like a data science problem – you should maximize the attention of your audience. Not just one person – the average attention of the audience. You should maximize it. That's the number one skill you should have and there's a good chance that's going to back into storytelling. It's hard to do, but it's not an impossible problem. It's a predictable problem. If you gave me two recorded talks, this person, this person – and those talks are gonna be streamed – they're live. I'm pretty sure I could tell you ahead of time, which talk will do better. Maximize the attention of your audience for the entire talk. (1:02:31)

Starting in AI evangelism

Alexey: One more question. Again, from Mathew. I know you can give a long answer to this question. But maybe you can keep it short? (1:03:43)

Alexey: “How does someone get started in AI evangelism?” If it's, of course, possible to give a short answer to that question. (1:03:58)

Ben: Having a speaking track record can be very useful. We actually just hired a new evangelist role, literally this week. For some of the candidates that were looking at, they had been speaking. So get out there, start speaking, build up your LinkedIn. That's a great way. Honestly, and I don't want to make this answer longer. The other thing too – if someone could actually convince me that they're a good speaker, I'd be very interested in hiring them. (1:04:08)

Alexey: How would they do that? (1:04:40)

Ben: If they presented at a local conference and they were number one at that conference. They had some audience feedback and said, “I presented here. 30 speakers. I got feedback, I'm the number one best speaker. I have the recorded talk.” I'd say “Great. I'll watch it.” I'll watch your recorded talk. That could be an opportunity to. Speaking – that's the main way to build up that speaking experience. (1:04:46)

Book recommendations

Alexey: For people who are still here on our stream, do you have any recommendations? A book or a course or something that people who are interested in building their public speaking skills – what is the number one thing they should try after this chat? (1:05:14)

Ben: Books – I like “Stories that Stick”. I forget the author's name, so I'm just gonna read titles. There's “The Hero with 10,000 Faces.” There's a book called “Play Bigger”, it talks about category creation, which is very useful. I love “The Lean Startup” by Eric Reese. It's not to make you a better speaker, but I love the mindset of failing before you start. Can you fail on a talk before you start? If I have a talk, before I invest time and energy into doing it… or research, if I've got a research idea for a talk, can I just fail today? Before I even start, can I fail? I love the idea of failing today and saving all that time rather than committing to something… and something you can start today. I do love the idea, if you've got kids – start telling them stories today. These aren't stories you're reading, you're winging it. Tell them a story from your childhood, something stupid you did, something funny, exciting, sad. Tell them a story and practice that. You become a better storyteller. The evidence of success is if your kids begin to ask for another story and another story after the second one. You can really level up. And practice maintaining their attention. Kids have short attention spans – this is a good practice. You know if you have their attention, so practice maintaining their attention. If you get better at this, you will be a better speaker. (1:05:40)

Alexey: And for those who don't have kids? (1:07:08)

Ben: There are some local meetups. I think Toastmasters. I've never done Toastmasters. But I know they will go and practice presenting to each other. It'd be the same thing. A big part of being a good speaker is having good emotional intelligence and being able to read your audience. If your audience is checking out, if the energy is going down, you have to fix it – as a speaker. As a speaker, you're not scripted. You can do all sorts of things to fix that attention. Sometimes it's hard, it takes a lot of energy and sometimes you screw it up. It’s just like a comedian – some comedians, they're really funny and sometimes they have off nights and it just doesn't work. It’s terrible. You need to practice. I still have terrible talks on the horizon. Hopefully fewer than I had before. (1:07:12)

Alexey: That talk where somebody fell asleep – was it recorded? (1:08:07)

Ben: No. (1:08:14)

Alexey: A pity. (1:08:15)

Ben: I wish it was. (1:08:15)

Ben: Pat Wright. He teased me that he has some recordings on YouTube of my early talks in the Utah tech scene that he said were terrible. So maybe I should reach out to him to see if I can fish out one of the worst talks ever recorded that I've given. It'd be fun now to actually critique it – to break it up and say why it's bad. (1:08:18)

One last story

Alexey: I suggest we should be wrapping up because I think your next meeting starts in 20 minutes and you still need to reach home. I don't think you want to have it from your car, like this one. Thanks a lot for coming, for joining us today and sharing all these stories with us and demonstrating your storytelling skills. I think… I didn't count the number of stories you told, but I think it was at least five. (1:08:44)

Ben: You reminded me one last thing. I've had people say that I have an unfair advantage. And they've said I have an unfair advantage because some of my stories are so ridiculous. One of the things I didn't mention. When I went to college, I lived in the woods. I lived in the snow and it made national news in the US and I had two sponsors. While students are living indoors going to college, I'm living in the snow, every day. I have lots of these types of stories. I used to hitchhike across the Nevada desert and I did all these things as a kid, so I have a lot of stories I can pull from. But I would argue that everyone's everyone has stories, but mine are kind of over the top. (1:09:20)

Alexey: Not everyone can pull this… have this kind of background. I heard about that. I think it was the Ravit show? When you talked about this – your story about living in the woods. For those who are interested, check that out. (1:10:06)

Ben: I got a news link so I can message you. It’s on YouTube, so they can actually watch the news clip when I'm 20 or 19 being interviewed on the news. (1:10:28)

Alexey: I will definitely put that in the description in the YouTube video. Thanks a lot. If you have any more stories to share, maybe you can also send a few links? It was great chatting with you. (1:10:39)

Ben: Yeah, thank you. Sorry about the car setup. (1:10:56)

Alexey: It was fun. I can even say “I had Ben in his car, not in his usual studio.” Not many podcast hosts can actually say that, I feel special. (1:10:59)

Ben: Yeah. Great. Great to see you again. Alexey. Thanks for doing this. (1:11:14)

Alexey: Have a nice day. (1:11:18)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.

DataTalks.Club. Hosted on GitHub Pages. We use cookies.