AGP Ep 3: Anthony Lake—On AI, machine learning, algorithms and influencing people, for data science!

by | Oct 24, 2016

Anthony Lake is the Principal Data Scientist for Complexica. Anthony has experience in leadership and management, but his current role is very hands-on. But he uses his leadership skills to influence clients and stakeholders within the business to get the best outcomes.

Listen

What did you think?

Thumbs up
Thumbs down
Close podcast feedback section

Spread the love!

Close show some love section

If you enjoyed the episode, you can help us by doing one or more of the following...

Subscribe via email

Give us feedback

Close feedback form section

Thank you for your honesty. How can we improve?

6 + 7 =

“You can use a stick for good or evil. AI is a tool and we need to understand the best way to use it” —ِAnthony Lake
"If your attitude is bad, it will never work, and if your attitude is good, skills are learnable" —Anthony Lake
"To have more impact, I needed to use the help of others to multiply what I was trying to do" —Anthony Lake

Show notes

 

Credits

Topics/Transcript

Andrew Ramsden: Hi Anthony. Welcome to the show.

 

Anthony Lake: Thanks very much Andrew.

 

Andrew Ramsden: I feel like I should be whispering though, because we’re sitting here in the State library in South Australia and they’ve kindly lend us a room, but I still feel like I should be respectful of the people studying around.

 

Anthony Lake: Yeah. That very much looks like a quiet space out there.

 

Andrew Ramsden: Just respectful of the architecture and the books. They need t o be left in peace.

 

Anthony Lake: Nice.

 

[00:00:30]

Andrew Ramsden:

 

They’ve got that sort of stoic, quiet attitude to them. As you would know, I have a soft spot for geekdom in all it’s forms and I think you have a really fantastically geeky role as the principal data scientist for Complexica.

 

Anthony Lake: Yes.

 

Andrew Ramsden: Tell us a little bit about that.

 

Anthony Lake:

 

[00:01:00]

Okay. Much of my role is spent looking at data in different ways. Playing a range of different algorithms to it, with the purpose of coming up with answers that are actionable by business people. Giving them answers that they can do something about or do something with to either improve their profits or the revenue reduce their costs, that sort of thing.

 

Andrew Ramsden: That’s the real trick to it, isn’t it? There’s so much data these days.

 

Anthony Lake:

 

 

[00:01:30]

Oh yeah, and you can ask a whole heap of questions, but it’s very easy to get off track and lose sight of why you’re asking questions of the data, because yeah it’s a lot of fun to play with but you have to be goal orientated. You can’t just throw a whole bucket of data at the world and see what interesting bits stick.

 

Andrew Ramsden: Nice, there’s not a scientific method for that. The hurl it against the wall technique?

 

Anthony Lake: No, you can come up with interesting insights, but not necessarily stuff that you can act on.

 

Andrew Ramsden: That’s excellent to be able to take actionable insights, because you know there must be nuggets in there somewhere.

 

Anthony Lake: Oh absolutely.

 

Andrew Ramsden: How do you think about that? How do you go about finding the nuggets? Is there a methodology?

 

[00:02:00]

Anthony Lake:

 

 

 

 

[00:02:30]

 

 

 

 

[00:03:00]

 

Yeah, there’s a loose methodology. We always start with a business problem, rather than a bucket of data. Then from the business problem we do look at the copious of data an organization holds. We try to identify the data bodies that we believe will help to derive an answer for the question, do some analysis around those. A little bit of back and forth with the client just to clarify the scope and the scale of the questions and their data holding and maybe to identify certain abbreviation and definitions and other bits and pieces. Then there’s bucket loads of statistical analysis, hypothesis generation. We come up with an idea that we believe will deliver value, then we test that hypothesis through a whole range of different techniques, such as machine learning, AI based concept, regression modelling, more statistical analysis, until we come up with what we believe is an answer to the fundamental question that the client is asking.

 

Andrew Ramsden: I guess there must be some clues early on as to what the best statistical role or other analysis method will be to use to answer that particular problem.

 

Anthony Lake:

[00:03:30]

 

 

 

 

[00:04:00]

Yeah, there are up front, what interestingly happens is, the upfront stuff will usually give you some easy answers, but once you gain more familiarity with the business of the client and what they’re actually doing. You start to broaden the range of techniques that you can use. For example, we had a corpus of data from a pharmaceutical company with sales information and sales rep information, GP information and that was an interesting body of data to start with, but it was only after spending the time of going out and riding for the day with the sales rep and understanding what they actually do, that we’re able to identify whole other big bins of incredibly useful data from different techniques to it. There’s a lot of exploration.

 

Andrew Ramsden:

 

 

[00:04:30]

Data that’s potentially missed, to begin with, because of the external to the company or it’s external to what they … Because I know that Michael Markovitz who is the CEO of Complexica, when I’ve seen him present about it, he talks about the power of the external data sets and the connections that can be then made when you can sync those up with your internal data.

 

Anthony Lake: Absolutely. We enhance or supplement the data that we get from clients with a whole range of external data. Whether that be bureau meteorology data, ABS data.

 

Andrew Ramsden: The weather, for example, can have such a huge impact on sales or use of taxis or Ubers.

 

[00:05:00]

Anthony Lake:

 

Yeah, absolutely. Slushies is a great example, you want to watch the weather movement versus the sales of slushies. It’s just perfectly in line with that.

 

Andrew Ramsden: There you go. They need a hot beverage to stick on the side of those machines so that they can capitalize on both ends.

 

Anthony Lake: For the winter market.

 

Andrew Ramsden:

 

[00:05:30]

That’s really exciting, it’s also really interesting to see some of the, almost the emulation or the simulations, the modelling work that you do and maybe that’s what you already talked about in terms of testing a hypothesis, but I know you them propose some different ways that the business can adjust their pricing structures, for example, or change the timing or change what they do and then you run that through simulations to see what are the best variations given a range of other inputs.

 

Anthony Lake:

[00:06:00]

There’s a three step process for that. First we come up with out hypothesis and the simulation allows us a way to test that hypothesis. Once we can simulate, the next step is to optimize. To work out the absolute best solution from a whole pool of solutions and that’s where algorithmically, at least, all the heavy lifting comes in.

 

Andrew Ramsden:

[00:06:30]

Nice. That must be, I imagine when you come up with the hypothesis, that must be the fun part where you get to sort of say, hey, wouldn’t it be amazing if this actually added value, this improve the situation.

 

Anthony Lake: Yeah, exactly. At that point, it’s real blue sky thinking. You’ve got a problem and just think laterally about anything that you can pull in, that might give you some insight into how to solve that problem.

 

Andrew Ramsden: It is nice.

 

Anthony Lake: Yeah, it is nice.

 

Andrew Ramsden:

[00:07:00]

How do you go with that bit? Because I imagine there’d be a deep understanding of the business or market forces or other aspects outside of data science and outside of information management, which is part of your background and correct me if I’m wrong. This whole insight to how business and the market work, that I think you then need to apply to make those hypothesis.

 

Anthony Lake:

 

[00:07:30]

Absolutely, you can’t do it in isolation, you must, fundamentally understand how this works and what it’s trying to achieve. Obviously different verticals have hugely different goals and different strategies to achieve those goals.

 

Andrew Ramsden: Do you see it as your job to get across that?

 

Anthony Lake: Absolutely. Yeah, absolutely. If you don’t understand what the business is trying to do, it’s very difficult to help them.

 

Andrew Ramsden: Yeah, absolutely.

 

Anthony Lake: You really do have to understand the business domain as well as the data domain, to be able to do a good job.

 

Andrew Ramsden: What’s the approach that you take to getting inside the heads or getting inside the business?

 

[00:08:00]

Anthony Lake:

 

 

 

 

[00:08:30]

 

A couple of different things. There’s always literature to read, trying to understand how business A is in their market, compared with all the rest of the vertical, all the rest in the same vertical. What their day to day operations look like, what their short, medium, long term strategic goals are. A lot of conversation with different people within the business, to try and get as many different viewpoints as you possibly can. It’s like a anthropology in a way, just trying to understand the character and the nature of the company that you’re working with.

 

Andrew Ramsden: That’s incredible, it’s a powerful combination because you get to play in this fantastic, science driven space, but it’s highly commercial and it’s highly focused on adding value.

 

[00:09:00]

Anthony Lake:

 

Absolutely. I also like the balance between the rigorous data driven science component, and also the human engagement component. The being able to go out and have face to face discussions about what is in some person’s day, how they actually operate and how the business operates. It’s really cool. It’s fascinating.

 

Andrew Ramsden:

 

[00:09:30]

Yeah and you do it really well. I saw you at the digital strategy conference this year, talking to a range of different people about their business. I that something you’ve always been good at?

 

Anthony Lake:

 

 

 

[00:10:00]

I’ve always loved to listen to people. I’m always curious about what drives people, what makes people tick, why people do the things that they do. I’m fascinated by a field called behavioural economics, which looks at decisions hearing. Pure economics would say that you’re always as choose the solution with the most utility, but behavioural economics casts a light on that and talks about why people really do what they do.

 

Andrew Ramsden: Technologies that were not really rational after all.

 

Anthony Lake: Exactly right. Why are we irrational and why are we irrational in certain ways? That side of science is always fascinating.

 

Andrew Ramsden:

[00:10:30]

I’d like to think that I’m rational. I’d love to think that and I guess the more I look into those sorts of biases, the more I realize that I’m as susceptible as everybody else.

 

Anthony Lake: What we are really really good at is post hoc nationalization.

 

Andrew Ramsden: Yes.

 

Anthony Lake: That must be why I did that. Of course.

 

Andrew Ramsden:

 

 

[00:11:00]

Good friend of my, Rory Daily and I, we’ve been talking about, we must really go away and write a book called, Man Math, because as men we seem to be able to justify, certainly after the fact, in more detail, in just about any purchase, any toy, any technology that we want to go out and buy.

 

Anthony Lake: Yeah, gadgets.

 

Andrew Ramsden: That’s it.

 

Anthony Lake: Yeah, absolutely. I do the same.

 

Andrew Ramsden: I’m sure there’s women math as well. That’s possibly a separate book again.

 

Anthony Lake: It’s probably beyond my ability to help within a data science perspective.

 

Andrew Ramsden: The man math is a scary thing. It can lead to some really interesting decisions.

 

Anthony Lake: Yes, and some buyer regret at times too.

 

[00:11:30]

Andrew Ramsden:

 

Yeah, absolutely. You talk about the use of machine learning and artificial intelligence and I guess, just for the sake of our listener who may not be as familiar with those. Is there a good way to describe what that means and the difference between them or is there a difference?

 

Anthony Lake:

 

[00:12:00]

 

 

[00:12:30]

Artificial intelligence is a little bit of a hard to define term. It’s been kicking around for many many years now. We through an AI winter over the last 20 years or so, where we thought that it reached its limits we can get a computer to Beta chess play, but our computer will never be able to insert sentence here and they can, it’s just a matter of defining the goal clearly and starting with the right environment. What is called artificial intelligence, that solution space is becoming much much smaller every single day, because people are finding ways to solve these problems. There’s the classic cheering test, we’re getting close to doing that. We’ve done go, pick the best human player on the go, which is …

 

Andrew Ramsden: It was terrifying when I read that story. I thought that’s amazing.

 

Anthony Lake: Insanely complex.

 

Andrew Ramsden: Insanely complex.

 

[00:13:00]

Anthony Lake:

 

Yeah, and as of last month we have a four Q bit quantum computer. Everything that we’ve done with cryptography is just about to go out the window. The world is changing very quickly.

 

Andrew Ramsden: That does my head in, quantum physics and how quantum computing works. It’s incredible.

 

Anthony Lake: Very cool.

 

Andrew Ramsden: I think it does everybody’s head in.

 

Anthony Lake: Yeah, it does.

 

Andrew Ramsden: That’s the conclusion I’ve come to, after all my reading is, no one really understands what’s going on at that level. We have some theories but …

 

[00:13:30]

Anthony Lake:

 

 

 

 

[00:14:00]

 

 

 

[00:14:30]

 

It’s quite cool. It’s machine manning I guess. It’s starting off with some information. There’s two types, there’s trained and untrained basically. Classification or analysis type exercises with a trained type exercise you give the computer a range of inputs and values associated with those inputs. It could be a sentence and then the value you give that sentence is positive so you start with test data and you train the computer as to what is positive, what is negative, what is neutral. Then you throw more examples at it and it will allow it to use that prior training. Example, that would be based on info. When you talk about untrained, that’s when you throw a whole bunch of data at the computer and do things like, okay, tell me what the most important variable is here, or given you’ve seen this set of data, if I give you an entirely hypothetical situation you’ve never seen before, what’s the answer, what’s the output?

 

Andrew Ramsden: If they can apply learning from one to the other.

 

Anthony Lake: Yes, exactly, have interaction.

 

Andrew Ramsden: Is that how genetic algorithms work? Is that term familiar?

 

[00:15:00]

Anthony Lake:

 

Yeah, genetic algorithms, that term’s quite familiar. Our chief scientist, speaking of Michael, is one of the world’s leading experts on genetic algorithms and evolutionary algorithms.

 

Andrew Ramsden: The system has to know what success look like. Does that make it a trained algorithm?

 

Anthony Lake:

 

[00:15:30]

It’s not trained, it’s not trained as such but yes. The system does, to an extend have to know what success looks like. Genetic algorithms and evolutionary algorithms are sort of similar, conception-ally, in the you start out with something simple, a very simple set of rules, and iteratively refine every time you run an iteration of that algorithm, the algorithm changes.

 

Andrew Ramsden: It rewrites its own code, doesn’t it?

 

Anthony Lake: Yes. It becomes better and better and better. The more stuff you give it, the better it gets.

 

Andrew Ramsden: It’s amazing.

 

Anthony Lake: Yeah, it’s really cool.

 

Andrew Ramsden:

[00:16:00]

I’m going to butcher this story, but I heard, and this is from a long time ago, I heard about a genetic algorithm that would print circuits onto a circuit board and it was trying to come up with the most efficient circuit design that met the requirements for that circuit board, and it kept coming up with these designs that on paper were completely impossible and the engineers that were looking at it would just throw them away and start again. There’s a bug in the system, we’ll run the algorithm again.

 

 

[00:16:30]

It kept coming up with this design that was really elegant and simple, but it wouldn’t work, it shouldn’t work. It did work and what they’ve found out was there was a magnetic field next to some of the equipment that was printing the circuit boards, so when it was tested it would actually factor that magnetic field in and you’d never do that on paper because you wouldn’t even know about the magnetic field and you wouldn’t factor it in. I just thought that was such a great example of the power of the genetic algorithm to find something in reality that we wouldn’t even think to look for.

 

Anthony Lake: Yeah, exactly. It’s very cool.

 

[00:17:00]

Andrew Ramsden:

 

Does that worry you at all? I guess this is the peripheral question with artificial intelligence, isn’t it? If the algorithm can rewrite it’s own code, can come up with a completely new version of itself, will it eventually become a general intelligence and therefore take on a mind of its own and rebel against it’s masters?

 

Anthony Lake:

[00:17:30]

 

 

 

[00:18:00]

There’s a couple of schools of thought on that. I guess I ascribe consciousness and free will, in inverted commas, as human rather than machine traits. I think consciousness is more than just an aggregation of algorithms. I think there are other things happening there, not necessarily spiritual things but other things that we can’t yet model. In terms of getting at least to the point of seeming human, I think absolutely computers will get there. Quite quickly I would say. In terms of developing free will, it’s a very difficult question, but I don’t think so.

 

Andrew Ramsden: I guess if a genetic algorithm, for example, has an idea of what success looks like, it would be a fairly long circuitous route to success that caused us problems.

 

Anthony Lake: Yes.

 

Andrew Ramsden: Of any scale worth nothing.

 

Anthony Lake:

[00:18:30]

Yeah, absolutely. I don’t think we’re going to hit that single point in time where a machine suddenly wakes up and then goes on a rampage.

 

Andrew Ramsden: It’s interesting isn’t it? Would it even need to be, and this is before getting into the realms of philosophy.

 

Anthony Lake: That’s okay.

 

Andrew Ramsden: Would it even need to be conscious to be a problem or to be a hazard?

 

Anthony Lake:

[00:19:00]

No, not at all. Take for example the whole nano-tech potential end in ‘grey goo’ where the bots just keep replicating, replicating, replicating because that’s what they were designed to do. Until they run out of materials.

 

Andrew Ramsden: Unintended consequences.

 

Anthony Lake: Yeah, exactly. You just need to be that sort of thing.

 

Andrew Ramsden: Yeah, there’s risks along the path, but you don’t think we’re likely to end up with some sort of singularity any time soon.

 

Anthony Lake:

 

[00:19:30]

Not any time soon and I think when it comes, there re already a whole lot of very smart people looking into how we might manage that and what we might do about that. Some of the best minds in the world are doing that right now.

 

Andrew Ramsden: Nice. Do you think that the artificial intelligence or the machine learning approach is the answer for our irrationality? Is that solution for man math?

 

Anthony Lake:

[00:20:00]

Wow, that is an awesome question. I think man math is a little bit of, I already know the answer that I want so I’m not actually going to listen to the answer that I know is the right one.

 

Andrew Ramsden: I’m not going to check in with Siri for a second opinion on this one.

 

Anthony Lake: Yeah, no. I do believe that we absolutely could solve man math and give reasonably good advice, but I have a hunch that we’re not seeking reasonably good advice. I have a hunch we’re seeking justification to do what we want to do in the first place.

 

[00:20:30]

Andrew Ramsden:

 

Fair point. This is very true. Is there a benefit then to counteract some of our irrationality in a context where … Because I think as human beings we don’t necessarily like being told what to do.

 

Anthony Lake:

 

[00:21:00]

No, absolutely not. Just jumping back into philosophy, again for a second. The idea of free will is one that’s held very closely by absolutely everybody. Not absolutely everybody, because there are quite a few books around that question whether we actually have free will or whether we do alto of post op rationalization on everything that comes out of our mouths, every behaviour that we take. That rationalization is sense itself, it sort of fantastic.

 

Andrew Ramsden: It is. Things like that again in the quantum physics and the trying to rationalize what’s going on in that space.

 

Anthony Lake: Absolutely yeah.

 

Andrew Ramsden:

[00:21:30]

I guess in a general usage we probably not going to give up our free will to consult the AI as to the best course of action. All day every day.

 

Anthony Lake: I can definitely predict when you’re not going to take the best course of action.

 

Andrew Ramsden: Right and help maybe protect yourself.

 

Anthony Lake: Intentionally, yeah.

 

Andrew Ramsden: Nice. There’s always such a cross over into philosophy and ethics these days, when it comes to artificial intelligence.

 

Anthony Lake: Absolutely.

 

Andrew Ramsden: We’re going to automate these things and give our decisions over to computers.

 

Anthony Lake: Yeah.

 

Andrew Ramsden: What guides that?

 

Anthony Lake: Yeah, exactly.

 

Andrew Ramsden: What limitations do we put around it?

 

[00:22:00]

Anthony Lake:

 

 

 

 

[00:22:30]

 

It’s a huge question, and that’s one that needs to be answered. On every instance, on every time you’re looking to do something like that, you do need to have a look at some fundamentals that aren’t about the data as to per se, the fear about privacy. They’re about utility, they’re about fundamentally discrimination. Should we discriminate, whether it be by price or opportunity or something else, between one set of people and another set of people? Is that a fair thing to do? Is it a moral thing to do?

 

Andrew Ramsden: This is very true. Even though little drive business margins.

 

Anthony Lake: Yeah, it’s that ultimately the best thing in the world to do. Should we enhance our ability to perform semantic analysis on text for experiments?

 

[00:23:00]

Andrew Ramsden:

 

Yes. There’s always a way to take the technology and perverse it.

 

Anthony Lake: These things are just talks. Form way back when we had sticks. You can use a stick for good or evil, it’s no different at all with anything that we do now. These are tools and we as people need to understand the best way to use that.

 

Andrew Ramsden:

[00:23:30]

I’m really looking forward to that age, that period of time that we come into our where we get to have these conversations and we get to see those conversations play out about the ethics. I think it forces us to move into a space where we are clearly articulating, almost codifying, literally codifying what we believe to be fair and equitable.

 

Anthony Lake:

 

[00:24:00]

Also what we’re doing is we’re deciding on the future that we want to have, rather than just accepting the future that happens to us. There’s a lot of power in doing that.

 

Andrew Ramsden: I think there’s been so many grey areas when it comes to ethics and that’s fine and that’s good and I think there needs to be, but I think there’s also a lot of room for codifying. I think there’s a lot that we can make black and white, and we probably haven’t to date.

 

Anthony Lake: No, we’ve had sort of a shared understanding culturally based, generally, but you’re right, I don’t think we have codified a lot of that stuff. It’s a great opportunity to do it.

 

[00:24:30]

Andrew Ramsden:

 

I think the typical example, it’s often raised as around self-driving cars and if there’s that moment in time where the car needs to decide, do I crash into the car in front of me or do I serve onto the footpath. There’s a question here as to, is it going to do damage to the driver or is it going to do damage to a pedestrian and how does it make that decision? I think that’s … There’s probably a whole range of these sorts of ethical questions to be grappled with.

 

[00:25:00]

Anthony Lake:

 

There’s a fundamental streetcar question that’s very much in line with that. Who do you save? If you can only save one person, do you save the young adult, do you save the child, do you save the old person, do you save the middle aged person, the mother or the father.

 

Andrew Ramsden: These are the philosophical trolley scenarios?

 

Anthony Lake: Correct.

 

Andrew Ramsden: Okay. We can put some links to some of that material in the show notes for people if they want. Go and read up more about that.

 

Anthony Lake:

[00:25:30]

That’s a very good moral question. It’s one that we actually have to come up with a solid answer for. To have these autonomous vehicles running around.

 

Andrew Ramsden:

 

 

 

 

[00:26:00]

 

 

 

 

 

[00:26:30]

Absolutely. They’re not easy questions to answer. Sometimes on face value they seem really really easy. I think there’s one trolley scenario, and I was hearing this talked about on a podcast. It was Sam Harris, but it may not have been his podcast. Anyway, I’ll link to it, but he was to answering a listener question about one of the trolley scenarios and he gave a fantastic explanation. On the surface, maybe I should just link to it, rather than try to rehash it, but I guess to now that I’ve raised it I should at least have a go at recapping. On the surface there’s a scenario where the trolley’s coming down the track towards three people and you have the power to flip the switch and send the trolley down a separate track towards one person. You have the power to control whether three people get injured or one person. It seems like a pretty obvious choice. Most people decide to flip the switch and injure one person.

 

 

 

 

 

 

[00:27:00]

Then there’s another scenario where the trolley’s coming down the tracks towards three people and you’re standing on an overpass above and there’s a heavyset man standing next to you. You have a choice of pushing the heavyset man into the path of the trolley to save the three people. I guess there’s this philosophical argument that if we could all get to that point where we push the heavyset man, then maybe the world would be better off, but we don’t, because it’s a different sort of scenario and I think Sam does a fantastic job of unpacking why that is a different scenario.

 

Anthony Lake: Cool, I’d like to listen.

 

Andrew Ramsden: That raises the question. Do you just jump off yourself and into that trolley to save the three? Rather than pushing the poor heavyset man. I don’t know what the parallel there is to code and artificial intelligence. I feel like I’ve wandered us …

 

[00:27:30]

Anthony Lake:

 

 

 

[00:28:00]

 

That’s a really clear parallel. These autonomous vehicles will be in that sort of situation and the logic by which the car makes that decision is going to be predetermined by the people that write the code for the car and those people will have to come up with that decision. Whether that’s influenced externally, whether Google company policy is to kill the one person and IBM company policy is to kill off three people.

 

Andrew Ramsden: Right.

 

Anthony Lake: I think it should be elevated a little bit higher than that. There’s a direct parallel there. It’s just, we’re embodying the technology with the ability to both drive itself around and we’re also putting it into that position where we need to tell it what to do and we need to tell it very clearly what to do. Otherwise we’re going to end up with unintended consequences.

 

[00:28:30]

Andrew Ramsden:

 

I think there will be parallels to other context as well. That’s a really nice example with self-driving cars because it’s about life and death and it’s …

 

Anthony Lake: It’s very dramatic.

 

Andrew Ramsden: It’s very dramatic and it’s very clear that that’s a problem, but I guess there will be more subtle, insidious problems, as you point out. Is it fair, is it ethical? This machine is making a decision to charge you a bit more than someone else, based on some context.

 

[00:29:00]

Anthony Lake:

 

Based on what device you use to access a website. Should MAC users copy MAC user’s text?

 

Andrew Ramsden: I guess this has been happening for a long time?

 

Anthony Lake: It has yeah. We’re just getting better at it. We’re getting better at differentiating, segmenting, micro segmenting people into little groups and then making business decisions about those groups, for better or for worse.

 

Andrew Ramsden:

[00:29:30]

I think that’s what’s great about the conversation that we’re having, specifically here but also more broadly in the industry is, it does force us to, these practices that’s been going on for a long time, it forces us to question them and have the conversation.

 

Anthony Lake: Yeah.

 

Andrew Ramsden: We may decide that it’s perfectly fine and go ahead and that’s great and I think that’s fantastic.

 

Anthony Lake: Previously it’s been a little bit blindly. At least people are now have a bit of an awareness and making it conscious decision. I think that’s a lot better.

 

Andrew Ramsden: What do you geek out about these days?

 

[00:30:00]

Anthony Lake:

 

Oh wow, okay. I geek out about comparing different algorithms. It’s really nerdy, I know. About performance, about accuracy, about quality of analysis. I geek out about state vector machines.

 

Andrew Ramsden: State vector machines. Wow. I’m well out of my depth there. You’ll have to explain to me what that is.

 

[00:30:30]

Anthony Lake:

 

It’s pretty simple. It’s basically just a matrix manipulation, a matrix calculator.

 

Andrew Ramsden: It sounds perfectly simple.

 

Anthony Lake: It’s a really fast way of doing maths.

 

Andrew Ramsden: Okay, nice, thank you for dummying it down for me.

 

Anthony Lake: I geek out about deep textural analysis. That’s a lot of fun.

 

Andrew Ramsden: Is that about the sentiment analysis then?

 

[00:31:00]

Anthony Lake:

 

Yes, sentiment, quality, concept extraction, all those fun things.

 

Andrew Ramsden: You were telling me about, you work in comparison, some sentiment analysis and algorithms.

 

Anthony Lake:

[00:31:30]

Yeah. Some sort of home rolled Python, natural toolkit. Some APIs from the big boys. An IBM API, a Google API and a couple of less well known APIs. Just seeing which one does the best job on, I’ve got about a million and a half records.

 

Andrew Ramsden: How do you asses which one’s doing the best job? How do you measure that?

 

Anthony Lake:

 

[00:32:00]

Here’s the interesting thing. Humans generally agree on sentiment about 80% of the time. If two humans read the same sentence, it’s only four out of five times they’ll both classify it in the same way. What I do is take a random sample out of that one and a half million of coded records compare all six, seven different engines to see how much agreement there is with them and I’ll also look at the actual text itself and rank it myself.

 

Andrew Ramsden: Based on your interpretation?

 

Anthony Lake:

[00:32:30]

Based on my interpretation. Then what I’m doing is I’m comparing the distributions of all of those rankings against each other and against the ranking that I give because it’s only me. I’m basically comparing APIs and machine learning sentiment analysis against Anthony Lake’s sentiment analysis.

 

Andrew Ramsden: Nice, the gold standard.

 

Anthony Lake: Yeah, of course.

 

Andrew Ramsden: Cool, that would be fun.

 

Anthony Lake: Yeah it is. It’s a ball. It’s very nerdy but it’s a ball.

 

[00:33:00]

Andrew Ramsden:

 

Are there any that are really outstanding or they all sort of, they have their pros and cons?

 

Anthony Lake:

 

[00:33:30]

Surprisingly there’s a less well known API. I’ll get you a link for the podcast, that seemingly outperforms Google, Watson, etc. What I’ve done with it though is instead of using it’s classification, what I’ve done is I’ve pulled the raw data that is used to make that classification back and tweaked a bit at maths that it uses to give me a bit more granularity of sentiment and that’s far and away better than anything else that I’ve been able to find so far.

 

Andrew Ramsden: I wonder what they’re doing differently?

 

Anthony Lake: I’m not sure.

 

Andrew Ramsden: It’s a bit of a black box.

 

Anthony Lake:

[00:34:00]

It’s a little bit of a black box. I’m starting to unpack it a little bit because they had based it on the Python natural language toolkit. I’m in my own sort of research on trying to create something that’s equivalent in performance with what I can get from the API.

 

Andrew Ramsden: I assume when you’re connecting to an API it is a black box. That’s a proprietary algorithm.

 

Anthony Lake: Absolutely.

 

Andrew Ramsden: Whereas this one is based on open source?

 

Anthony Lake: Yes.

 

Andrew Ramsden: Okay, so you can peak under the hood, so to speak.

 

[00:34:30]

Anthony Lake:

 

They haven’t exposed what’s under the hood, but because it’s built using some freely available libraries and tools, you too can grab those freely available libraries and tools and start coding up stuff to try and get a similar performance.

 

Andrew Ramsden: That will be hard though. I imagine they’ve invested a lot of time and effort in how all those tools connect together.

 

[00:35:00]

Anthony Lake:

 

Yeah, absolutely. Which is why they can sell access to the sentiment API.

 

Andrew Ramsden: Right.

 

Anthony Lake: They’ve put a lot of time and effort into it and done a very good job. It’s marketable.

 

Andrew Ramsden: Okay, that solution is an API based solution and it’s propriety as well.

 

Anthony Lake: Yes it is.

 

Andrew Ramsden: Yeah, that’s very cool.

 

Anthony Lake: Yeah.

 

Andrew Ramsden:

[00:35:30]

How do you find you’re able to communicate the value of what you’re doing through to … Because I know you talk to the customers at the digital strategy summit earlier this year. How do you communicate the value to them, because they wouldn’t necessarily care.

 

Anthony Lake:

 

 

[00:36:00]

They don’t really care what I do data wise. For them the value is around that increased revenue, better margins, more volume, optimal solutions, less resource wastage, less marketing and promotional wastage, more effective marketing. It boils down to, there is a fundamental dollop value to what we do. Otherwise we don’t do it.

 

Andrew Ramsden: I guess there’s that understanding of the business that comes into play and what represents value for them.

 

Anthony Lake: Yes.

 

Andrew Ramsden: What about the communication side of it? Do you ever find that there are people that are resistant or challenging to deal with?

 

Anthony Lake: Absolutely. Yeah. Absolutely.

 

[00:36:30]

Andrew Ramsden:

 

How do you approach influencing them?

 

Anthony Lake:

 

 

[00:37:00]

 

 

 

 

[00:37:30]

It depends, to an extent, on whether you can understand what their motivation is for being challenging. It’s quite often the case where, because we talk at a C level and generally CEO … More so the CEO, the CMO, so the chief marketing officer, if there is a chief sales officer then that person. The chief financial officer. Those more business orientated people. Occasionally what happens is there is some resistance from one or two areas. There’s an existing data analytics area within the business or depending on the style of IT department there can be resistance there as well. One of those is driven on, I think on the most part feeling threatened. The approach that we take with them and I’m really passionate about is, we never go into take somebody’s job away. We go in to help them do their job more effectively. You enable the person that’s thinking, oh God, what am I going to do when these guys show me up? No, we’ll help you, we’ll teach you.

 

[00:38:00]

Andrew Ramsden:

 

You’ve worked out what’s motivating them and that fear and you’ve then been able to reframe that problem.

 

Anthony Lake:

 

 

[00:38:30]

Yeah, so then they no longer see a threat, they see an opportunity. With IT departments it’s occasionally the case of just inertia and the not invented here type stuff, because a lot of the stuff that we do is cloud based, it’s essentially little or no footprint within the customer’s side. There’s not much IT resistance generally. Sometimes it can be very difficult to get data out of people, because they don’t want to give it to you, don’t know where it is, don’t have the appropriate access to the appropriate systems, that sort of thing.

 

Andrew Ramsden: Some more technical challenges there than deliberate interference.

 

[00:39:00]

Anthony Lake:

 

Deliberate interference is rare. In those sort of circumstances I do feel that that’s a matter better dealt with internally. The idea there is to just to communicate the reality of the situation to the management team and allow them to manage.

 

Andrew Ramsden: Nice. Anthony, what do you think are the unique challenges of being an influencer in the technology or the digital space?

 

[00:39:30]

Anthony Lake:

 

Being sure of yourself is a key challenge. Actually knowing to an extend what you’re talking about.

 

Andrew Ramsden: Is that about keeping up with the changing landscape?

 

Anthony Lake:

 

 

[00:40:00]

Yeah, it’s definitely about keeping up and it’s about having thought about stuff from more than one angle, rather than just following a straight path. It’s about having a look at alternate paths and positives and the negatives in that paths and making a decision about the paths you take. Rather than just falling from technology to technology or solution to solution.

 

Andrew Ramsden: Right, can be hard to generate those other perspectives. How do you do that? How do you look at doing that?

 

Anthony Lake: It’s brought up something very interesting and I’m just going to side-track for two seconds.

 

Andrew Ramsden: No, absolutely.

 

Anthony Lake:

 

[00:40:30]

Google’s really good at getting to know what interests you, knowing your political position, knowing what you like to read, knowing what you don’t like to read. You’ll notice if you use anonymous or incognito mode in a browser, type in a search and then search under a logged in profile, you get very very different answers. The answers that you get logged in as yourself are self reinforcing stuff.

 

Andrew Ramsden: Interesting.

 

Anthony Lake: What I generally do is I try and avoid …

 

Andrew Ramsden: Is that confirmation bias?

 

Anthony Lake: Yeah, its confirmation bias. We’re human and we love confirmation bias. Look how right I am.

 

Andrew Ramsden: Yeah, it’s fantastic.

 

[00:41:00]

Anthony Lake:

 

Generally from I am doing some sort of research, I will generally do it in incognito mode or in an anonymous browser.

 

Andrew Ramsden: Nice.

 

Anthony Lake: That way, try to avoid whatever personal bias that any sort of information gathering service would like to supply me.

 

Andrew Ramsden: That’s a great answer to the question, how do you get the fresh perspectives. You do turn off the automatic confirmation bias.

 

[00:41:30]

Anthony Lake:

 

Yeah. You talk to people with different opinions, because you can always learn something.

 

Andrew Ramsden: Absolutely. That can be quite challenging too, can’t it?

 

Anthony Lake: Oh hell yeah.

 

Andrew Ramsden: Everybody sees the world through this lens of their own background or their own history, their own baggage.

 

Anthony Lake: Everybody is never a villain in their own story. They always think they’re doing the right thing and the best thing.

 

[00:42:00]

Andrew Ramsden:

 

I think that’s a good observation. I think they generally believe that they usually have good intentions.

 

Anthony Lake: Absolutely, yeah.

 

Andrew Ramsden: They might be intentions that aren’t necessarily aligned with yours.

 

Anthony Lake: Understanding that is helpful in understanding why somebody’s gone down a particular path, taken a particular job.

 

Andrew Ramsden: Or why they don’t agree with you. Why they’re arguing with you. why they’re making your life painful.

 

Anthony Lake: Yeah. You’re never going to get on with everybody.

 

[00:42:30]

Andrew Ramsden:

 

How do you get that alignment between intentions and … Is that too abstract? Sometimes I ask these really floaty, abstract questions, but if you’ve got someone that comes up to you and they’ve got the best of intentions but they’re heading down a completely different path. How do you address that?

 

Anthony Lake:

 

[00:43:00]

 

 

 

 

[00:43:30]

It’s often contextual, depending on where they’re trying to go, what they’re trying to do. Some techniques that I’ve use previously in similar situations is, to walk through with the person step by step what they’re going to do, what they expect to happen and how they’ll know that they’re heading in the right path. If I want success in this, I’ll know I’m successful by these four things, and how will I get to those four things? Okay, let’s pull that apart to get to the first thing. What steps are you going to take to do that. Why do you think to get that out? What if you don’t get that out, come for step one. How does that affect the next one?

 

Andrew Ramsden: It sounds like a really open conversation.

 

Anthony Lake: Yeah, and it’s very much reflective and I really enjoy that type of conversation and I like it in reverse as well. Where somebody can just point out, but yeah and if you do that you’re going to crash the car, or whatever the case may be. Oh yeah, okay, thanks.

 

Andrew Ramsden: I guess, in a sense, you’re helping walk through their thinking and [crosstalk 00:44:00].

 

[00:44:00]

Anthony Lake:

 

Yes, to help them understand it.

 

Andrew Ramsden: There you go. It’s not about challenging their thinking necessarily, although that might happen through the process.

 

Anthony Lake: It might, yeah.

 

Andrew Ramsden: You’re trying to genuinely understand it.

 

Anthony Lake: Yeah, maybe they have a better approach to it than I do.

 

Andrew Ramsden: Right, you’re also genuinely open to your perspective being challenged?

 

Anthony Lake: Oh yeah. You got to be.

 

Andrew Ramsden: So true.

 

Anthony Lake: Nobody is always right.

 

[00:44:30]

Andrew Ramsden:

 

No, as much as we’d like to think that we are. Do you ever recruit for other technical staff to join your teams?

 

Anthony Lake: Yes, I’ve recruited extensively.

 

Andrew Ramsden: What do you look for when you’re trying to recruit great staff?

 

Anthony Lake:

 

[00:45:00]

Attitude. The single most important thing is having the right attitude. It doesn’t matter about the level of technical skill or the area of technical skill. If the attitude is bad, it will never work, and if the attitude is good, skills are learn-able.

 

Andrew Ramsden: So true. What would the qualities be or the attributes be of that sort of attitude if you had to describe them?

 

Anthony Lake: Positive, humble, hard-working, interested in learning.

 

Andrew Ramsden: I think you’ve nailed it. I agree.

 

Anthony Lake: Those are the type of people I love to work with.

 

Andrew Ramsden: Yeah, open to lifelong learning.

 

[00:45:30]

Anthony Lake:

 

Exactly right. Because you have to be in the technical game. You can’t stand still.

 

Andrew Ramsden: This could be confirmation bias and I’m very aware of this, but I see this trend time and time again, of that idea that you talked about earlier which was having the challenge in the technology space is having that confidence, but then also the humility and having both of them.

 

Anthony Lake: Having the balance. You’ve got to be confident in your choice but humble enough when somebody goes, that’s wrong, to go, okay, I see that now.

 

[00:46:00]

Andrew Ramsden:

 

There’s that other mantra to … Do you need to get out?

 

Anthony Lake: No. What’s that? The other mantra?

 

Andrew Ramsden: Yeah, there’s that other mantra which I think … I’ll have to look up who it is that says this, or said this originally, but there’s that other mantra about strong opinions loosely held.

 

Anthony Lake: Yeah.

 

Andrew Ramsden: Being really passionate and confident and yet being open to being proven wrong.

 

Anthony Lake: Being open to being proven wrong. Absolutely. Yeah.

 

Andrew Ramsden: Then jumping on the next thing with passion.

 

Anthony Lake: Yeah.

 

[00:46:30]

Andrew Ramsden:

 

I think there’s a bit of a conflict there because we want to be seen as people that have some loyalty to ideas and concepts and see things through. We don’t want to be seen as this type of character, this flighty character that jumps and flip flops, does 180’s.

 

Anthony Lake: Yeah, there is that. However the counterpoint to that is, always use the best thing you possibly can.

 

Andrew Ramsden: Based on the knowledge that you have at the time.

 

Anthony Lake: Based on the knowledge that you have at the time.

 

Andrew Ramsden: It’s very scientific.

 

[00:47:00]

Anthony Lake:

 

Yeah, there’s nothing wrong with being very scientific.

 

Andrew Ramsden: That’s the scientific method isn’t it?

 

Anthony Lake: Yes, it’s the way it work.

 

Andrew Ramsden: Nice. This is probably a bit of side line and I should have probably asked this upfront. Did you ever have any major illness or injury or health issue?

 

Anthony Lake: Yeah I had.

 

Andrew Ramsden: Do you think that sort of changed you or impacted the way you see the world?

 

[00:47:30]

Anthony Lake:

 

Yeah, it may have. It did force me to reflect a bit and to look at my priorities and think about what I did actually want to achieve. I guess it took me out of idle, that’s for sure. I was just a little bit crazy as in what I should have been and it took that away.

 

Andrew Ramsden: There’s a distinction between before and after?

 

Anthony Lake: Yes, there is a distinction.

 

[00:48:00]

Andrew Ramsden:

 

What changed for you significantly afterwards? What did you then go and do? Did that signify a transition into roles where you were influencing more roles where you were managing more or roles where you were …

 

Anthony Lake: Yeah, it took me out of the purely doing space and into more the managing, the leading, the doing.

 

Andrew Ramsden: Cool.

 

Anthony Lake: Just through …

 

Andrew Ramsden: Maybe I should frame that one properly and then let you go for it.

 

Anthony Lake: Yeah cool.

 

[00:48:30]

Andrew Ramsden:

 

What sort of detail should we talk about in terms of the health impacts?

 

Anthony Lake: I fractured my spine. I squished three vertebrae and lost about two centimetres in height.

 

Andrew Ramsden: Wow, was that an accident?

 

Anthony Lake: Yeah. Had, obviously, a lot of mobility issues at that point in time and also just being able to sit down for long enough to be effective in the workplace.

 

[00:49:00]

Andrew Ramsden:

 

Anthony, I know you also had a significant accident earlier in your life, which was a horse riding accident.

 

Anthony Lake: Yeah.

 

Andrew Ramsden: You had a whole range of impacts from that fractured spine and all sorts of broken bones. Tell us a bit about that.

 

Anthony Lake:

[00:49:30]

Okay. That was, it could have been in, I think, 2008. My wife’s sister had a boyfriend who were over visiting and we took him horse riding. Being a country kid I’ve obviously been horse riding quite a bit before. However, this particular horse was a little bit stiff, a little bit jumpy and it stepped into a gully, quite a deep gully and as it went down, I went up and then as it stepped out of the gully it came up and I went down very very hard.

 

[00:50:00]

Andrew Ramsden:

 

Ouch.

 

Anthony Lake: Yes.

 

Andrew Ramsden: That’s a significant injury. You said you lost two centimetres in height. How did that happen? Was that as a result of surgery or compression?

 

Anthony Lake: Compression. Just compression on the vertebrae.

 

Andrew Ramsden: Wow, that’s phenomenal.

 

Anthony Lake: It hurt.

 

Andrew Ramsden: Yeah. Nerve damage?

 

Anthony Lake: Yeah.

 

Andrew Ramsden: How did you recover from that?

 

[00:50:30]

Anthony Lake:

 

Pretty slowly actually. I spent a few months in a brace and a lot of time not having long periods of sleep, because long periods of in the prone position, it would exacerbate the issue, the injury. Changing position often and sometimes embarrassingly, having to stand up in the middle of business meetings.

 

Andrew Ramsden: I think everybody does that these days. They’ve all got Apple watches that tell them …

 

[00:51:00]

Anthony Lake:

 

Yeah they do now. They certainly didn’t back then. I wish they had it. A whole lot of mobility issues, a whole lot of carrying issues as well. I couldn’t lift anything more than about five kilos, which obviously have a lifestyle impact.

 

Andrew Ramsden: Did that have an impact on you otherwise?

 

Anthony Lake:

[00:51:30]

 

 

 

[00:52:00]

Yeah absolutely. Previous to that I think I thought I could do everything myself and take on the world and do it singlehandedly. That really confirmed to me that yeah, no. No I can’t. Really to have more impact and have more effect, I needed to use the help of others to multiply what I was trying to do. To get them on board, give them a hand, point them in the right direction and form a team, a coherent team with coherent goals. That was a better way to get more impact.

 

Andrew Ramsden: What was it about the accident or the injury or the recovery process that led you to that conclusion?

 

Anthony Lake: It was really discovering my own limitations and that those limitations were far more significant that they had been previously. Philosophically it led me to realize that no, I can’t do everything singlehandedly, even if I’m at 100%. I do need help.

 

[00:52:30]

Andrew Ramsden:

 

Nice. You tell me you don’t think of yourself as a leader?

 

Anthony Lake: No.

 

Andrew Ramsden: You’ve got wonderful insights into leadership and what it takes to build a team, certainly in terms of those recruitment attributes that you look for. What’s your approach to building a team and getting them to work together? How do you think about that?

 

[00:53:00]

Anthony Lake:

 

 

 

 

 

[00:53:30]

 

There’s so much that I could say about that. I’ll just give the high level story, I guess. Ensuring open lines of communication. I love a flat team structure rather than hierarchy team structure. I like flexibility and the ability for one team member to be able to learn skills from another team member. Role swapping is great, but ensuring that everybody’s aware of the direction that we’re headed in, the steps that we’re going to take to get there is, I think the most important thing. Bought into the journey as it were.

 

Andrew Ramsden: Yeah, how do you do that?

 

Anthony Lake:

 

 

[00:54:00]

You have a compelling journey. The goals that you set for a team must be attractive goals, and they must be in line with the people that you’re trying to use to achieve them. If you’re asking people to do stuff that they don’t necessarily agree with, it’s going to be challenging for them to be a super productive team member and happy in the team. They can still, to an extent, work effectively but to have a really high performing team, everybody in the team has to be in alignment on where the team is going and what sort of things they’re going to get from there.

 

Andrew Ramsden: Is there almost a value space there, value alignment between what that individual feels significant and valuable?

 

Anthony Lake: Yes, absolutely.

 

Andrew Ramsden: Aligned with what the business sees as significant and valuable.

 

[00:54:30]

Anthony Lake:

 

Yeah, and it’s important to try and evoke those values very very quickly, very very early, so that you can see if there’s any misalignment.

 

Andrew Ramsden: You include that as part of your recruiting for attitude?

 

Anthony Lake: Yes, absolutely.

 

Andrew Ramsden: Attitude and values.

 

Anthony Lake: Yeah, absolutely.

 

Andrew Ramsden: Have you done that in the past?

 

Anthony Lake: Yeah.

 

Andrew Ramsden: How do you evoke those in an interview situation or in a recruitment situation?

 

Anthony Lake:

[00:55:00]

I’m asking people what they’re passionate about. It’s a great way to do that. What lights your fire, what gets you out of bed in the morning? When you’re not doing this job, what do you love doing and why do you love doing it? Those sorts of questions, trying to get a sense of what drives people. It’s often a good way to start understanding what they value as people.

 

Andrew Ramsden: Absolutely. Nice. Well, we’re probably getting to a point now where we need to think about wrapping up. I just have one more question for you.

 

Anthony Lake: Sure.

 

[00:55:30]

Andrew Ramsden:

 

What does the future hold for Anthony Lake?

 

Anthony Lake:

 

 

[00:56:00]

That’s a big question. That’s a really good question. I’ve moved from technical inter management and now I’m back in a technical role again. For the moment I’m really happy to be in a technical role again. I’m loving getting my hand dirty as well. I’d like to continue to do that while I can and I just want to look for harder and harder problems to solve. That’s my future.

 

Andrew Ramsden: Nice. Keep throwing down the gauntlet.

 

Anthony Lake: Yeah, exactly. I want challenges. I don’t want it to be easy. If it’s too easy, there’s no sense of accomplishment.

 

Andrew Ramsden: So true. Very true. Nice, and it keeps you stretching and learning and growing, which I know you’re a huge fan of.

 

Anthony Lake: Yeah, absolutely. Whole life’s learning is where it’s at. Absolutely.

 

[00:56:30]

Andrew Ramsden:

 

Lovely. Thank you for that today. It’s been a real honour to get to hear your story and thank you for sharing with us.

 

Anthony Lake: It’s great talking with you Andrew. It’s just been a pleasure.

 

Andrew Ramsden: Is there anywhere you’d like to direct our listeners attention to or any final advice?

 

Anthony Lake: Final advice? I guess one piece of advice is, be passionate about what you’re doing.

 

[00:57:00]

Andrew Ramsden:

 

Nice. That’s great advice. Thank you again.

 

Anthony Lake: No worries.

 

Andrew Ramsden: Until next time.

 

How do your leadership skills stack up?

Take the 6 minute test
X