Alice Albrecht - From Neuroscience to Data Science (#28)

Welcome to another episode of Develomentor. Today's guest is Alice Albrecht. 

Alice is currently an independent consultant. She is also building a brain-computer interface startup. 

Prior to this, she led the strategy and advising portions of Cloudera Fast Forward Labs, where she was also a research engineer.

Before joining Cloudera Fast Forward Labs, Alice worked in both finance and technology companies as a practicing data scientist, data science leader, and - most recently - a data product manager. In addition to teaching machines to do cool things, Alice is passionate about mentoring and helping others grow in their careers.  

Alice holds a PhD from Yale in cognitive neuroscience where she studied how humans summarize sensory information from the world around them and the neural substrates that underlie those summaries.

Click Here –> For more information about tech careers

Episode Summary

“We’ve had this huge surge in the use of deep learning and neural networks and all of these more complicated and very, very data-hungry techniques and organizations have seen a lot of benefit from those. But I think that in the next couple of years we’re going to see a plateau in the advancement there and we’re going to have to rethink the way that we’re developing these algorithms.”

—Alice Albrecht

In this episode we’ll cover:

1. How Alice started college when she was 15 and was on the fast track to becoming a professor
2. Why Alice left academia to work in the private sector and technology
3. How a neuroscience background gives Alice an advantage when working in data science
4. What Alice looks for when hiring data scientists
5. Why Alice believes deep learning will hit a plateau in the next couple of years

You can find more resources and a full transcript in the show notes

To learn more about our podcast go to https://develomentor.com/

To listen to previous episodes go to https://develomentor.com/blog/

Follow Alice Albrecht
Twitter: @AliceAlbrecht
LinkedIn: linkedin.com/in/alice-albrecht-6379868/

Follow Develomentor:
Twitter: @develomentor

Follow Grant Ingersoll
Twitter: @gsingers
LinkedIn: linkedin.com/in/grantingersoll

Transcript

intro: 0:18

[inaudible]

Grant Ingersoll: 0:18

good morning and good afternoon, good evening, wherever you are and whatever time zone you are in. Thanks for listening to the Develomentor podcast. I'm your host Grant Ingersoll. My goal in each and every episode is to bring you interesting guests across a variety of careers and technology and really showcase the practical examples of how you, our listener, might carve your own path regardless of whether you go to the traditional four year college degree route or not. In fact, today's guest has done just that. She, it is the second time I've had on a guest who started off getting an a degree in psychology and in this case she actually got her PhD in cognitive neuroscience before switching over to data science where she has grown from an individual contributor to a leader of data science teams. Along their journey, she's done research on child development, postdoc work, and even been a product manager. Please welcome to the show Alice Albrecht. Alice, great to have you here.

Alice Albrecht: 1:17

Okay, great. Great. Thanks for having me. I appreciate it.

Grant Ingersoll: 1:20

Yeah, no happy to. Uh, you know, you've built this, quite a career around, you know, first doing that PhD like I mentioned. And then these days leading the data science team. Why don't we just kick things off and have you actually walk us through a bit, introduce yourself, walk us through a bit of your career and your background.

Alice Albrecht: 1:39

Sure. So, um, I'm Alice and, uh, my background, like you said in the introduction is pretty varied actually. So, um, I started off on an academic path, so I did my PhD and got a postdoc in cognitive neuroscience. And, um, in that I had a really strong focus on understanding how humans perceive scenes, um, and spent a lot of my time thinking about, uh, what data we could collect, uh, that would sort of illustrate all of the secrets that are hiding in your brain. Um, and so when I decided to leave academia and sort of make the jump into industry, uh, data science was this sort of still nascent field, um, and was a great place for academic ex-pats to sort of find their footing. Um, so I worked for a while as a data scientist and like you said, I have led data science teams. Um, and uh, at this point in my career, I help other companies that are trying to build data science teams and data science products and use machine learning in the organizations to um, understand how they could build those products and what strategies they can employ to make sure that they're successful in that.

Grant Ingersoll: 2:55

That's great. I love that term by the way, an academic expat. Walk me through a little bit there. Uh, uh, you know, this is often a critical decision. You did all this work to get a PhD, right? I mean, that's a serious investment of time and energy and, you know, you went off and launched a career there and then, you know, at some point you decided, Hey, you know, I need to make a change here. Walk me through a little bit. What was involved in that change? How did that come about?

Alice Albrecht: 3:28

Yeah. Um, it wasn't an easy decision. I had really gone the academic route, um, wanting to have an academic career and really thinking that was going to be the thing that I would do for the rest of my life. So, um, I know some people that um, get graduate degrees and whatever it is, um, may or may not plan to go on in academia. So with that foundation of wanting to be an academic, um, the decision really, uh, was based on a couple of things. I think, um, as people go down these paths in whatever space they're in, um, you gained more and more understanding of what that job is actually going to look like later down the road. And for me, I realized that my academic research was going to continue to narrow to an extent that, um, it felt almost constraining. Um, so yeah, there's this, uh, sort of misconception that as an academic you can just do whatever you want, but you really are still beholden to grants and to sort of[inaudible] fulfilling your story for tenure and whatnot with your papers. Um, so industry at the time that I made the transition was a place, again where there were lots of academics that were doing research outside of academia. Um, and data science was a way to still have that sort of experimental nature and to be exploring, um, and to be thinking about new ideas and new things with data, but you weren't sort of stuck in a more narrow path.

Grant Ingersoll: 5:05

Interesting. Yeah, no, that, that makes a lot of sense to me. And it's, it's funny cause, uh, you know, as you introduce yourself, uh, you know, it's almost like a complete contrast these days, right? Because as you said, like you go out and spend a lot of time with your customers doing consulting on data science and, and you know, and I've done a lot of that in my own career and, and one of the cool things is like, you get to just see like so many different use cases, right? Like you get to explore all these different ideas. So if you were being narrowed before, it's almost like these days it's just wide open. Is that, is that the case?

Alice Albrecht: 5:40

I think so. Um, and I think the, the wide openness in the industry, like as a practicing data scientist or someone leading teams is within organizations. You have this ability to have this wide breadth of things that you work on. Um, but also there's obviously lots and lots of different companies and almost every company is using this now. So you can have this experience of being in lots of different industries within. But I'm saying it's industry.

Grant Ingersoll: 6:08

Yeah. So let's see, you know, for our listeners who are perhaps aren't as familiar with data science, you know, perhaps you could give just a quick definition and then more importantly like talk through kind of what's the day to day like for a first a data scientist. Cause I know you've been in that role and then, you know, perhaps, uh, towards the end, talk a little bit, what does it mean to be a data science consultant? What does it mean to be a manager of a team of data scientists?

Alice Albrecht: 6:36

Yeah, and data science is an interesting thing because I think it's evolving really rapidly the term. Um, so as it stands today, the way I see data science is I'm a function within the organization where people are using the data that the organization has or enriching it with other data to cleverly try to predict things. Um, I think when I had first started in data science, it was much broader than that. Um, but I see that as the main sort of function now. And we have seen this delineation of, um, roles where we now have data analysts and machine learning engineers. Um, and so I think it'll be interesting to see where data science is, what it means to be a data scientist. Um, you know, in a couple of years from now. But I see the day to day for data science being a lot of working really closely with business stakeholders. So those could be usually product managers, um, to understand what is being built in the organization or what needs to be known by the organization to operate or to sort of understand where they're going. Um, and so that collaboration with the business stakeholders is key. Um, and I think it takes a little more time than we think about for the data scientists. Um, and these days, although it wasn't that way when I started, um, the data scientists aren't doing as much of the sort of pipeline creation and management work depending on the size of the company, um, but are really more focused on, um, running specific analyses or experimenting rather with a couple of different analysis, um, to either build something or to get the answers they want.

Grant Ingersoll: 8:22

Yeah, that makes sense. You've seen evolve kind of this specialization of data engineers and data scientists like what's, you know, without naming names, and of course, I don't require any of my guests ever talk about actual what they, you know, their current employer or any of that. But like w what's, for instance, some high level examples of a problem a data scientist might solve for a company, let's say I'm a, uh, I don't know, I'm an eCommerce site. How, how might a data scientist work at an eCommerce site like an Amazon or a Walmart or a fill in the a retailer.

Speaker 3: 8:57

It's interesting. Um, so even beyond e-commerce, I think it really depends on the function they're supporting within the organization. So, um, if a data scientist is working, um, let's say in ecommerce, u m, some of their work could be dedicated to, u m, better surfacing recommendations. So, you know, doing some modeling around, u m, different ways that you can recommend products to consumers coming in. U m, there's also a lot of work though around, u h, content generation, so, you know, predicting what sorts of merchandise will be more successful versus others. U m, and then really in any company, there's a lot of work that's done by data scientists still around, u m, sort of the internal workings and helping to make that eCommerce site more efficient. So helping with call centers or, u m, automating different workflows.

Grant Ingersoll: 9:51

Yeah, that makes a lot of sense. That's a, I think that's a great example. And you know, I liked that space from a data science perspective because it's often so very cut and dry around, Hey, we either need to be, you know, we're either going to try to make more money by selling more things or we're going to try to save money by helping customers find what they need without actually having to pick up the phone. So very concrete examples, that data's often usually pretty good. Is that fair to say? You don't have to necessarily do as much, uh, scrubbing and cleaning and there's often a lot of data available to[inaudible].

Alice Albrecht: 10:27

Yeah, I think that's definitely the case these days. I think we've gotten past that real, real struggle with getting any data.

Grant Ingersoll: 10:32

Yeah. Whereas some of these internal enterprise projects, you know, uh, you know, sometimes it's like, all right, where am I even gonna get the data around? Like employee behavior, those kinds of things. Uh, shifting gears maybe a little bit, uh, you know, so you've got this neuroscience and psychology background. How has that helped you? I mean obviously part of getting your PhD is you need to build your research and you need to learn stats and all of that. But what are perhaps some other ways it's, uh, or along with that, what are some ways that that background has helped you be successful in, in data science and in these new roles you're in?

Alice Albrecht: 11:11

Yeah, that's a great question. Um, I think there are probably three ways and some of them are less obvious, um, for people that weren't in the field or the really specific field that I was in. Um, so I think the first thing is that, um, I spent a lot of years understanding humans and their brains and all of the data that we're talking about or most of it really, unless we're talking about, um, machine maintenance is coming from human beings. So, um, I, I think that background and understanding, you know, to some extent, um, how humans are making decisions or what data would be relevant to gather is a huge part of, you know, my success in doing data science. Um, and I think another piece is, um, the work that I did because I was so focused on, um, human vision, um, I spent a lot of time back in the day in MATLAB, I'm creating visual displays. So I was doing all of the work to create the stimulate from my experiments. Um, and so that, yes, I actually, that's why I taught myself how to program was to create these very, very specific displays with um, often really short latencies and I had to have those controlled, um, down to like milliseconds. Um, and you also have to hook those up to various devices. So because I have this neuroscience background, the FMRI and EEG and TMS and all of these acronyms, um, basically all of these different ways of recording brain data or recording ive movements, even, um, you had to be able to integrate hardware, um, and create the programs around that because they largely didn't exist. We didn't have as many sort of open source software projects as we do now. Mmm. So it was a strange forehand to doing some sort of, yeah. Engineering work and learning how to program, um,

Grant Ingersoll: 13:12

necessity is the mother of, uh, in this case education, right? Uh, you had, you had to teach yourself because especially in a PhD role, like it's not like there's a programmer sitting next to you is going to do it for you. Right?

Alice Albrecht: 13:26

Yeah, yeah. You're really, you're really on your own and you were in a race too sort of make these scientific discoveries. So if you can't do it efficiently, you're really, you're your disadvantage. Um, and, and I would say the third piece that is often not super obvious to people is, um, for the work I was doing, um, let's say back in 2009, 2010, um, in neuroscience, we had started using machine learning to try to classify what people were seeing from their brain data. Um, so, and back in those days, I was basically using support vector machines to try and see if I could classify like different images that you saw on a screen based on lots and lots of exposure to those different images.

Grant Ingersoll: 14:19

Interesting. Yeah. So another good example of some of the data science work people do, uh, around, uh, you know, classification of content, which is a huge area for data science, uh, at least in my experience. So, yeah. So, you know, flash forward here, you're, you're out of academia, you know, looking on your profile, you know, you've worked actually across a lot of industries as well, finance from what I can tell, internet search academia and now you're in this consulting role. What do see as some of the key lessons you've learned that have helped you be successful in, in all those different roles and, and kind of led up to where you are now?

Alice Albrecht: 15:02

Yeah, I think, um, one of the things that has helped me be successful is to be able to see the commonalities across really disparate, uh, industries and fields. So, um, at least from a data perspective, a lot of the, um, you know, the algorithmic approaches, but also the ways of thinking about use cases. Um, you can learn things, let's say from my time at Yahoo that, um, would be applicable when I'm thinking about building other sort of consumer facing tech products and finance, even if it's fantastic. So yeah, I think the biggest thing is being able to, uh, C and uh, sort of fold those commonalities and where they exist. Um, and then in terms of the consulting, obviously working in different fields is really helpful when your job is to work with lots of different industries. Um, and to be able to hop back and forth between say, you know, finance and oil and gas and agriculture. So

Grant Ingersoll: 16:04

yeah, I know that's one of the things I love about that role is, is you get to go see, you know, there's never a boring day, right? Cause you're never, you're never working on the same thing for just years and years on end. Now the flip side or the downside is often you never actually quite get to see things to the finish either because you're handing something off. But you know, you hit a, you spend a good amount of time with clients, you know, and a lot of technical folk aren't always necessarily the most outgoing, you know, talk about some of the skills you had to develop to be in a client facing role or w ere you always just k ind o f naturally comfortable, you know, building relationships and doing, you know, some of the soft skills that often come with, with that space.

Alice Albrecht: 16:50

It's interesting. Um, so I don't think I was naturally good at doing client facing work and I, I think if you'd asked me 10 years ago if I would be doing client facing work, I probably would have left. Um,

Grant Ingersoll: 17:02

yeah, me too. I went through the same thing. I totally know what you're saying, but yeah, keep going.

Alice Albrecht: 17:09

Yeah. And, and I think, um,[inaudible] yeah, I think it's, it's interesting in the space that I'm in now doing the sort of machine learning and data strategy work, I have to have some client facing skills like basic um, etiquette and um, be able to hold a conversation that's not a technical. Um, but in fact I find that[inaudible] people give me more of a pass if I don't have the sort of glass and veneer of an actual consultants of somebody that really just went into consulting as a career and wanted to sort of climb that ladder. Mmm. So I give myself a little less credit for developing this really great client facing presence I guess.

Grant Ingersoll: 17:57

Yeah, no that's interesting because I mean what I've often found is like a technical consultant is there's often, there's this meeting where like it happens in every process I think where like there's the business people and the technical people are in the room and they're all kind of looking at you as the, the, the lead liaison for your company, right. Where you've got to, you've got to translate, for lack of a better word, math to non math people, but then you also have to kind of prove your credibility with the math people or the the computer science people. Right. And, and that requires an interesting dance if you will, of kind of going up and down the stack. So perhaps you're selling yourself a little short there because I'm assuming you've been in some of those meetings just based off of your role and

Alice Albrecht: 18:50

I, yeah, I definitely have been. And I think actually, um, for those out there that are doing data science work in an organization, part of that role even is being able to take really technical things and translate them. So it does set you up to be in a more client, let's say, facing role, even if that client isn't, it is not an internal client.

Grant Ingersoll: 19:10

Yeah, it makes sense. Um, I love to ask this question of, of a lot of my guests, what's been the most surprising? Like if you look back on, uh, you know, uh, I'll assume you're older than 18. Let's look back on Alice at 18. What's been the most surprising thing that if 18 year old Alice were to say, wow, I'm going to do that? Uh, what, what would she say?

Alice Albrecht: 19:37

Yeah, I think, I don't think there's any way 18 year olds myself could have envisioned that this is where I would be. Um, and I think, yeah, I think my former self would have probably been surprised. Mmm. And I've gotten this, especially as I've, you know, as I've gotten older and as I've been working with, um, you know, even much earlier in my career working with people and asking them, you know, are you surprised at where you're at? How did you get there? What was the formula? But it's been always the advice I've gotten is you can never plan these things and you just will end up in a place someday and there was no master plan to get there. Um, so I think, yeah, in that theoretical conversation with my 18 year old self, that's sort of the, that's the realization I think is that was true. And you do end up in all of these interesting places later on.

Grant Ingersoll: 20:32

Yeah, I mean you probably, I mean, I don't even know, did 18 year old Alice have the plan to go do academia even at that stage or did that come later as well?

Alice Albrecht: 20:43

I was an incredibly precocious 18 year old. Um, so I have this plan to be an academic. Then, um, I actually started college when I was 15, so I was wrapping up, um, at that point and had really realized that that like, I really wanted to be a professor because it meant you could keep wearing to school for a really fun and learning.

Grant Ingersoll: 21:05

Oh wow. Interesting. That's really a, um, I'm a, I didn't even know that, that, that's fascinating. And, and so you must have had been your PhD by the time you were done with that was pretty early as well. So the fact that then you pivoted out of it is, is says a lot about, you know, Hey, um, this is just isn't right for me, which is awesome. I love hearing those kinds of stories of people who are, Hey, you know what? I know I have these expectations, I had these expectations for myself, but you know what, they just don't fit and I don't want to do that anymore and I'm going to find a way out or find a way to go do something different.

Alice Albrecht: 21:44

Yeah. And I actually did get that advice as I was in my postdoc. I remember going to a talk and someone said you had something like seven lives to live and each of them are about seven years and it was take a lot of different farms, but you've done one for seven ish, maybe nine. Uh, depending that sort of, this is an opportunity for you to think about what do you want that next seven years to look like.

Grant Ingersoll: 22:08

Interesting. And I guess we're only living to 49 then. Ah, okay, fantastic. I'm like, wait, what are you telling me? I'm almost there. Uh, that's a, that's awesome. I actually, that's, I'm thinking back on my own and uh, yeah, probably about right. Um, you know, at least kind of you go through these phases as you, as you move through a, you know, I did straight up software engineering and I did some consulting, I did sales engineering and these days I did research for a while I did a, and these days I'm a CTO. So, uh, you know, it's, that's definitely interesting. One last question, kind of looking back a little bit. You know, whenever I have managers or leaders on the show, I really like to take a step back and then ask, you know, what do you look for when you're hiring data scientists? You know, what are some of the key skills and traits that for somebody coming into this, you know, Oh, Hey Alice, you know, in her observations of data scientists, they have some of these key skills. Like, you know, how do you think about hiring skills, personality, uh, you know, that kind of stuff.

Alice Albrecht: 23:30

Yeah. And I've said a lot about this. Um, especially when I was killing out the team when I was at simple, I feel like, um, I wanted to be really cognizant of, uh, data scientists having some sort of like commonality and then the specific individuals each having really different perspectives and different backgrounds. I think that just works better when building out teams. But for data scientists in particular, I think the number one thing that I look for is intellectual curiosity. Um, and I think so much of the job is thinking about a problem and trying to think about it different ways because there's always a lot of different solutions. Um, and if you can get really excited about lots of different things and lots of different problems, all of the technical pieces will come with training, especially now, there's so much out there to help people with that. Um, but that I would say is the fundamental piece that I look for.

Grant Ingersoll: 24:34

Yeah, no, I mean, and that you hit the nail on the head of how I look to hire a two. And it's not even just data scientists like anybody, right? Is like, I want that, let's go figure this out. Attitude. Yeah. Right. Like, and like, like you said, like the skills, the programming languages. I mean, sure, you need to check some boxes. Do you know Python, do you know, you know, do you know this language if you're a programmer, do you have this specific set of skills within reason? Like you're not going to train somebody up from scratch obviously, but, um, that let's go figure this out. Let's, let's go look at this in different ways like across the board, you know, so if our listeners, you know, listening and like if you take nothing else away from any of these interviews, this is the one that I'll tell you across a lot of the people I've interviewed, especially all the managers, they all hit on, right? And so how do you actually, like, let's dig in a little bit. Like how do you actually interview for that? Like, how do you approach finding that out with somebody?

Alice Albrecht: 25:39

I think I have the most fun way of finding this out. Um, and we'll see as you ask this question to everyone else. Um, something that I always like to do, uh, is have a sort of made up question, um, with lots of different answers to it. So, um, for data scientists in particular, this could be a question that I'm going to give them a dataset to solve. Um, and for now, um, you know, these days I hired a data strategist, so the question, um, ends up being about this fictitious company and all of these things. Um, but I think having a standard question that I ask everybody, um, with some reasonably well thought out answers on my point, so I know it's answerable. Um, it's a really fun way, um, for everybody to be on the same page and talking about the same problem even if they haven't been in that industry or haven't sort of done work in that area specifically. Um, and then in terms of measuring the creativity of the answers, I don't think there is a sort of data driven way of measuring that creativity, but I think there are proxies. I think that, um, the number of different answers given or, um, the types of questions people ask, Mmm. Or, um, to your point of sort of just figuring things out, um, their comfort level with going down different paths that were not obvious that might be placed into that.

Grant Ingersoll: 27:13

If you, if maybe at the risk of sharing your secret question, would you be willing to like actually give this as an example and then, you know, listeners of this podcast will have the, the inside on interviewing with you? Is that, is that okay?

Alice Albrecht: 27:28

Yeah, I mean, I can make one up. So, yeah. So I don't have a standard one that I use every time. Um, but I generally, I actually have the most fun, I do this in my tax too, making up fake companies with lots of different know problems that I've seen across the industry. But, um, let's see. Um, so for a data scientist, if I were hiring one today, um, a potential question could be, uh, say you're working with a product manager, um, for a new product that's going to be launched, um, soon, uh, and that product is brand new to the suite of products that the company has. So we don't have any prior data on it. Um, but, uh, it's got some characteristics of other products say, uh, in our line of products. And, um, maybe for some Empliciti let's say that product. Is it an article recommender? And this is a media company, sorry. So, um, as a data scientist in that space, uh, I think the first line of questioning is always what would you do to help that product manager or whoever's meeting to make the decision, decide on what[inaudible] sort of paths they might need to go down. Um, what features they would want to make sure that they're instrumenting. Um, and um, what do they think is the potential for this product? And those questions are all going to be very vague because the product doesn't exist yet. Um, but in that situation, Mmm. Sort of stepping out of the question prompt, I expect people to then to start think about what are other datasets that I can think about? What are other examples of this that might be available to me? Um, proxies that I could use to build up my understanding of the problem. Um, and any customer data that we would have or ways that we could quickly get some customer data without building the product out.

Grant Ingersoll: 29:26

I love it cause you have this cold start problem for sure. And you have, yeah. And like how do you, how do you, uh, get to an MVP? Great question. I love it. Yeah. Well we'll, we'll do this. We'll save the answer. Uh, so they, they still have to go to the question. They still have to, uh, they still have to go interview with you. Yeah. That's fantastic. I love it. Uh, I think you hit on, you know, at the end of the day, the key to it all right, which is, show me your curious and show me you want to figure it out. Let's shift gears. We've been talking a lot about the past and kinda, uh, those things. Let's shift gears a little bit and look at the here and now and forward of being a data scientist or, uh, or I guess now you're calling it a data strategist, uh, uh, which is, which is also a great name for this role. Uh, you know, the word cognitive, artificial intelligence, machine learning, and these are kind of everywhere these days. Like if you're in tech, you can't turn a corner without hearing a lot of, some of its hype, some of it's real. Uh, what do you see as kind of the key opportunities and challenges in this field especially cause you, you know, you often spend time looking forward around what's, what's next, what do you see as the key opportunities and challenges for data scientists, for data strategists for this role of managing data and turning it into something useful?

Alice Albrecht: 30:53

Yeah, I would say the opportunity space is actually large and growing because I think, um, one benefit I'd say of all of the hype around this job and around, um, you know, data and machine learning and you know, whatever we call AI right now, uh, in general, is that a lot of companies have begun to integrate this. I'd say maybe a year or two ago, a really, I saw a big shift, um, and most companies having a planner on this and having people and having the data acquired. So I'd say the opportunity is really large right now too. Build really meets products that are powered by data. Mmm. And so I think for data scientists out there, there is this huge opportunity to um, help influence the organization that they're in, understand what's possible with data and machine learning. Um,[inaudible] I think that there are a couple of key places where that possibility is rapidly becoming more and more possible. Um, one of those is in NLP. We've seen a lot of advancements in that area, um, in terms of algorithmically but also in terms of packages and ways that people can use this without necessarily having, you know, a PhD in anything really.

Grant Ingersoll: 32:14

And the NLP just for our listeners is natural language processing is what we're talking about there. Just to jump in there. Yeah. But keep going. Yeah.

Alice Albrecht: 32:23

Thank you. Thank you for keeping me honest with the acronyms. Um, yeah. And so I think, uh, the opportunity space there is really to say if we were going to make better products today, given this tool that we have, Mmm. Where are the places we could use that as a tool? Um, and, uh, like I said, the packages are getting better and also we're getting, um, we're not at a point where we can commoditize these models. Um, so there's no sort of point and click solution to a lot of this, but we are getting to a point where the libraries and packages are getting or making it easier and easier to do this without having to get so deep into the actual math behind it. Um, yeah. And yeah.

Grant Ingersoll: 33:10

Yeah. And in fact, I kinda think like, uh, I often talk, uh, you know, especially up and coming developers and data scientists like there, there's just this function now that says that that you have at your fingertips that's like make this application smarter by learning from user behavior, right? Like that's, that's the new world of the next generation of applications, right, of, Hey, there's this machine learning function. There's this predictor that you can just integrate all throughout your application and it will[inaudible] you know, semi magically learn.

Alice Albrecht: 33:42

Yeah. And I think this is, this is a huge opportunity and it's also sort of tipping into the, maybe the more pessimistic end of this when we talk about, um, you know, what are the costs or what are the things to look out for? Um, as we see more and more what people I think are calling auto ML or, um, these ways of kind of sprinkling data or machine learning and Mmm. If organizations stop understanding what it is that they're working with as a tool, then it becomes a problem. So if you have people that have no data experience, that are just trying to integrate this, it can become a huge issue because they don't really understand the implications of what they're implementing. Um, and it's different than other forms of engineering, I think because of the, the sort of self-learning nature of this. So the, the model can drift and only things that can happen. Um, do you want to look out for? Um, and I think on the other end of this hype cycle too, one thing to look out for and the next couple of years is we've had this huge surge in the use of[inaudible] deep learning and neural networks and all of these more complicated and very, very data hungry techniques and organizations have seen a lot of benefit from those. But I think, uh, that in the next couple of years we're going to see a plateau and the advancement there and we're going to have to rethink the way that we're developing these algorithms. Um, and to really, before we get anywhere near that road to AI, if there's gonna have to be a lot of work done.

Grant Ingersoll: 35:23

Yeah. Interesting. Actually, I think you stole my question there a little bit because I was going to perhaps a provocative statement that, you know, in the era of big data, there's this diminishing returns or there's this saturation point, you know, can Google really, for instance, learn anything new about users? I don't know the answer, but like, you know, you see a certain amount of behavior and a certain amount of it's constrained by the application you're using. Like there's only so many ways you can ask a question of say Google and you know, like after awhile they've just, you know, they've, they've effectively seen the entire space. Does that then limit us in terms of uh, to your kind of, your point of how much we can then go and build on her or is it that we need to develop better techniques for, you know, the kind of like what the human brain does, which is be better at pruning out? Uh, you know, clearly we can take in a lot of data as humans, but we also prune out a lot of it and, and I'm sure your, your neuroscience background could correct me where I'm wrong there, but I would love to hear your thoughts on that.

Alice Albrecht: 36:30

Yeah, I think we will see a point of diminishing returns in terms not in terms of the applications that are already in production now. I think those will continue to get faster and they'll continue likely to use less and less data. And there is a lot of work being done to make that true today. I think where we're going to see the diminishing returns is in novel or new things. So, um, the analogy that you used in terms of pruning data is apt in a sense. But I, I think what we will see happen is a little bit of a backlash against deep learning as being the answer to these things. And as, as people make that natural connection between human neuroscience and the techniques that have been developed or rather have been made possible by the technology we have today, they were developed a long time ago. Mmm. What we'll see is the need for us to think really differently about the types of machine learning algorithms that we use. So I think that we're going to see a big surge probably in a lot of probabilistic techniques, um, and people hopefully thinking outside the box, uh, in terms of where they could develop things for us to see any kind of novel or different applications of these.

Grant Ingersoll: 37:47

Yeah. Very interesting. And, and for our listeners, you know, we'll link up in the show notes, some, some getting started info on things like data science and deep learning and uh, and, and some of the things Alice's talking about here cause it truly is a fascinating part of technology and there's a lot of investment going on there. I think Alison, the interested at a time, you know, I want to shift gears a, you know, thank you so much. So, so many rich, uh, useful pieces of information in there, both around your career path as well as around how you think about hiring and where this field is going. Let me pause and just take a take this last final question of just, you know, like if you sum it all up, uh, you know, what's kind of the best career advice you have for somebody who's interested in getting into this, around where do they start? What do they, what do they need to, to be successful?

Alice Albrecht: 38:38

I would say for somebody that's just starting right now, um, I have maybe some unintuitive advice, which is that don't, uh, I think this will evolve and change, uh, to the point that when someone getting started feels like they're up and running the field and what it looks like will be different. So my advice actually is for people to start to dig in and to find what they're really inspired by now and whatever that is. And if that includes data, that's amazing, um, to find ways to use that and to build things with that now, but not to be too worried about a prescribed path to get there because everything is going to evolve as does everything else.

Grant Ingersoll: 39:26

Yeah, no, that makes sense. I mean, it's that open mindedness, which goes back to your, I think the key takeaway and perhaps the title of this show is this, this intellectual curiosity fair to say.

Alice Albrecht: 39:39

Yup.

Grant Ingersoll: 39:39

Yeah. Great. Hey Alice, thanks again for joining me and to all our listeners, you can find the show notes from today's episode and all of our interviews at develomentor.com there. You can also leave us feedback or perhaps even submit your own story to be featured. So please do join us, ah, on the site and Alice, once again, thanks for joining me and, and uh, we'll leave it at that. Thank you as always to our listeners for taking the time to listen. If you’d like to show, we’d love for you to subscribe on Apple podcasts or whatever your favorite podcast app is. You can also visit us at develomentor.com to hear older episodes as well as find other content on careers in technology. Most importantly, if you like the show, please tell your friends. Referrals are the lifeblood of any podcast. If you have any feedback on this episode or any episode where you’d like to be a guest, drop us an email at podcast@develomentor.com Finally, we here at Develomentor hope that each and each and every episode helps you move that one step closer to finding your path

Outro: 41:06

[inaudible].