10 - Fred Goff, Jobcase: How is ML shaping the future of work?
Machine Learning has been used to increase profit for businesses and solve tough problems in research, but what if ML was being harnessed to help people? Well, there are few things that affect people’s lives more than work.
On this week’s episode, Fred Goff, CEO of Jobcase, talks about harnessing the power of ML to empower workers and build community, with some powerful economic and political repercussions.
Read on for the transcript!
Hey, it’s Monte. If you'd like to see my newest project, a real-time feature store, we're holding a hands-on workshop at TWiMLCon on January 28, where you can actually play with it. Learn more at mlminutes.com/events.
Monte 0:10
Hi, I’m Monte Zweben, CEO of Splice Machine. You’re listening to ML Minutes, where cutting-edge thought leaders discuss Machine Learning in one minute or less.
Monte 0:30
This episode, our guest is Fred Goff, founder and CEO of Jobcase, a social media platform dedicated to empowering workers. Fred is a leading voice on issues related to worker advocacy, and how AI will impact the future of work. Welcome, Fred.
Fred 0:49
Thanks, Monte. Great to be here.
Monte 0:51
Well, tell us about your journey. How did you get to where you are now?
Fred 0:55
So look, where I started out was I'm a first generation college grad that just had the good fortune of finding his way to Carnegie Mellon. I then ended up in New York, I had the good fortune of finding my way into Wall Street and managing about a billion dollars. I launched a machine learning based hedge fund, and eventually then ported the technology that was underneath that into the consumer internet space. But I think what's most relevant for the journey is a moment: In the consumer internet space, I found myself realizing that most people are talking data, and they mean people. And I've been in rooms where they talk about commodity labor, and I think that's people. That's my cousin, my dad, and people are not commodities. And I started thinking about how could we leverage AI and machine learning and big data for individual people? How do we be a complement that while Neil Bussery is finding actionable insight and big data, to help companies, we can find it to help individual people. And so that's how I find myself to founding Jobcase and running this company.
Monte 1:59
That's fantastic. There's been many uses of data, and machine learning to help companies, help enterprises, help governments. And I love the fact that you're focused on helping people.
Fred 2:10
That's right.
Monte 2:13
Tell us more. What's the problem you're trying to solve with Jobcase?
Fred 2:18
So in 2013, Oxford came out with a study that McKinsey amplified that said 47% of jobs can be automated through computer models, computerization, that's pretty scary. That's scary to economic systems, that has fallout into political systems. When we got focused on how you manage this whole future of work. Increasingly, we realize that people don't really have the tools or a community of other people to help them go forward. But if we could have a platform for that, the same elements that might be incredibly scary, the elements that might take you as a cashier and displace your job, if you actually lean into that mode, it can be incredibly empowering, free online education to teach you skills that can come with a salary and benefits and etc. But how are you going to solve that? So we decided that if we could form and democratize data to help people help themselves navigate the future of work, this wouldn't be just important for them, but it actually has system wide economic impact.
Monte 3:18
Excellent. So you're helping people navigate work? Maybe you can give us a day in the life of a job case member?
Fred 3:30
Yes, well, I'll give you a day in the life today, as I'm sitting here talking to you still in the middle of the pandemic. We have an office manager, say, Teresa, who's lost her job recently. And she's she's over 40. And let me tell you, that is not old, but you face ageism. And so what can she do to move forward? She'll come to Jobcase and start talking about this in our community. Other people will help her navigate how to deal with ageism. In our community, other people will will bolster her emotionally in terms of you can do this, our team is going to look at her behavior on site is going to look at what kind of jobs she's clicking on are going to find the kind of job alerts to help with what she needs. Now in Teresa's case, it's not five year career navigation or education to get to a new career, if she needs a job now, and we will help her land that with our internal technology and all the employers that are active on our site, and she'll be encouraged to achieve it because of the community that supports her.
Monte 4:34
Terrific. So you're really helping people, especially in today's economic constraints and economic pressures, find positions when they may be left out of work. You're doing it through community and not just job listings. Maybe you can tell our listeners how community helps.
Fred 4:57
Sure Monte, I'd love to. A lot of your listeners are probably very familiar with LinkedIn, LinkedIn is great for knowledge workers to connect with their first and second degree connections. And they leverage those to figure out what to do next about opportunities that are open and how to pursue it. You know, for a lot of folks in the workplace, all those that we celebrate today as frontline workers, medical billing specialists, and hospitals, waiters, everybody bringing the packages to your door: they're not on LinkedIn. And as they look to figure out what to do next, it may not be their first and second degree connections that'll help them; it may be, but there's frequency of job changes, and keeping in touch with people is hard. Or it may be that you decide you want to be a nurse, and you don't have anybody you know, that can do that. Having a community that can help you navigate that and support you and say you can do it is incredibly, incredibly important. And so those are the aspects just at the beginning of how community and network helps. Jeff Wiener talks about the networking gap in this country holding people back and being part of the economic injustice. Fundamentally, we're fixing the networking gap problem.
Monte 6:00
That's fantastic. Using a community social media site, to connect workers with people that can help them sounds like a fantastic mission. Now, this is ML Minutes, so let's get to the machine learning. Where do you apply models in your community to help workers and help employers?
Fred 6:25
Monte, ML is at the core of everything we do. When when we had roots as a hedge fund we were offering and inventing algorithms, right now our philosophy is to be at the cutting-edge and be the best at the world at harnessing ML and AI for the benefit of our members. So we harness deep learning networks like TensorFlow to better predict lifetime value of a member, which helps informs our bids on how we advertise opportunities. We shamelessly steal NLP practices to embed in clever ways into into recommender systems. Regarding recommender systems, it can be off the shelf, I mean, leveraging Redshift for simple Jaccard distance calculations on our collaborative filtering that help us understand when someone says, I've been a cashier, and it's up to us to decide how do we what jobs are relevant for that. It's not just cashier, right, as you get to lower wage jobs, it's a very broad, diverse set of jobs that might be applicable. And leveraging ml to understand implicit and explicit signals to land relevance is critically important.
Monte 7:30
That's great. So you seem to have deployed many different technologies from data warehousing technologies like Redshift, all the way through to deep learning, like with TensorFlow, and maybe some of the new NLP approaches that are attention-based. This is a robust deployment of AI. Can you identify one specific challenge that you faced along the way of deploying ML?
Fred 8:02
We've both been around long enough that we have memory, where your gating factors are compute power, or your gating factor as storage, or maybe even the cleverness of the heuristics or algorithms themselves. Our problem the HR landscape is data, old, good old fashioned data. It's not stitched together, most people's work histories are in silos at individual companies. There's no common ontology to stitch together. And I'm not a believer in unstructured data having a lot of value, despite having seen business models for 20 or 30 years that it does. So if you think about the--it's not so much in the deployment of models, but in when we try to form how do we forecast the right career navigation? And as you try to find whether it's collaborative building, whether it's an expert system, whatever recommender system you have, we have difficulty in stitching together the data all the way down to the end result that we're really looking for: hires, retentions, growth. And so fundamentally, at the end of the day, it's not a technology problem, it's not a deployment or engineering problem. It's still data problem. And that's what we're focused on.
Monte 9:05
Well, I can honestly say that just about every guest on ML Minutes has said almost the same exact thing: it's data. And I think that we continue to be in a world where it's not the algorithms, it's not the math, it's the data that's going to drive signal. And it's still challenging us to get data into a form that is predictable, usable, clean, and repeatable. And it sounds like you've also got that challenge. Have you done anything in particular at Jobcase to overcome the data challenge?
Fred 9:46
Well, we have overcome the data challenge to a great degree, some of its philosophical and some of it some new partnerships that's promising. With regard to philosophical, I think of our data, Monte as explicit and implicit. So when we're thinking of-- to go back to the example, we talked earlier, what what can a cashier do next? You know, it's not about looking at a resume and saying you've been a cashier before, we have to look at behavior in terms of what jobs are looking at what you're clicking on, what articles you're reading, what kind of people are you interacting with, to inform us of your intent. Because for a lot of folks, it's not about a requirement or a skill, it's, "Would you be interested in doing this?" So opening up our understanding of data to intent, not just explicit information, and then forming models to predict that has been really effective. In terms of stitching together that end result, we're really pleased that Workday ventures is now a part of the company, which opens up a potential of us taking what we know about employees, and marrying it perhaps with what they know about employers to really close that data gap for the entire system.
Monte 10:51
Well, that's fantastic. Being able to make models that can score intent clearly can help drive behavior and benefit your members. And now being able to leverage data from a corporate partner should really help inform your models. Okay, moving on to the future. What's next for Jobcase?
Fred 11:18
Well, at the at the risk of of hubris, we're gonna rebalance capitalism, my friend! I think that we will succeed in having an agnostic platform where people keep their portable profile, this will accelerate labor markets, reduce friction, and through this network, we now will have a tremendous volume of workers. Our objective is to take that volume and advocate for workers in the workplace, not through division, but addition. An example would be when Delta Airlines decided to do a billion and a half windfall bonus to their employees, instead of doing another shareholder buyback, we went out and told all of our members fly Delta. Not just work Delta, fly Delta, because they treat workers well. We think we can use capitalism to balance capitalism, because as we can harness consumer dollars, to reward employers that treat workers well, then their competition, they have to treat workers well as well. So we think we can have a meaningful impact on individual people's lives. And through that we can have a meaningful impact on trying to get rid of the economic disequities in the system.
Monte 12:24
That's fantastic. And what what a passion, you're exhibiting for the role you play in helping workers into the future. Fred, how do you predict AI is going to influence the future of work?
Fred 12:41
Monte, that's a that's a profoundly important question you're asking. The way that I see AI impact in the jobs landscape rhymes to me with nuclear scientists in the 50s. If you're a nuclear scientist in the 50s, you can spend your time and energy building bombs, or powering France. The AI scientists today can spend their time helping people or displacing people, it's going to be a choice, it's going to be a choice. And so very specifically within the jobs landscape, we can try to harness it to displace tasks, you know, within a radiologist, can we displace the task of reading that chart, and free up the radiologists to broaden their job description to be a whole bunch of other things? Or do we use it to displace that radiologist? At Walmart, you could use it to displace the workers or you could use it with AR and VR as they're doing in some cases, to better train your employees to be better engaged with the consumers. It's a choice on how we harnesses technology. And what the big vision of Jobcase is to hopefully have everybody hear us: we've been saying it at MIT, and we'll say it on the West Coast now too, you have to put people first but it has to be a cognizant, purposeful choice. Otherwise, it could be a very dark future.
Monte 13:58
Well, that's an inspirational view of AI. And I appreciate that, and agree with it. It's been a pleasure, Fred, thank you so much for joining us on ML Minutes.
Fred 14:12
Thank you guys. Great to be here.
Monte 14:21
If you want to hear Fred talk more about how virtual reality will shape future jobs, check out our bonus minutes. They're linked in the show notes below and on our website, mlminutes.com. Next episode, we'll be exploring applications of machine learning in space with Dr. Danielle DeLatte. To stay up to date on our upcoming guests and giveaways, you can follow our Twitter and Instagram @MLMinutes. ML Minutes is produced and edited by Morgan Sweeney. I’m your host, Monte Zweben, and this was an ML Minute.