Artificial Intelligence And The Future Of Work
Michael Housman: Hey everyone thank you for tuning in. My name is Mike Housman and I’m the Co-founder and Chief Data Science Officer of RapportBoost. Today for our first ever interview I have the pleasure of speaking with Roy Bahat who is the head of Bloomberg Data. He’s a long time investor. He’s also an entrepreneur. He’s built companies in the past and Bloomberg Beta specifically has an interest both in the future of work as well as in artificial intelligence. So we’ll be digging in a little bit about both of those. But first Roy, thank you so much for joining us.
Roy Bahat: Thank you.
MH: So just to start, I thought we should talk a little bit about workplace, people analytics, and things of that nature. You and I both have an interest in it. I’m wondering from your perspective, what are the major factors that are influencing and changing the future of work?
RB: I think there are two ways to think about the future of work. The traditional one is if you went to a conference ten years ago about the future of work, the kinds of things they would talk about were working remotely, and new management structures, and how the organization of the company is changing. So I call that the future of working for us. Meaning for companies like ours. And then the other way of looking at it, which is the way I’ve been researching recently is more like the future of employment. Given how technology is developing and cultures are developing, what are the kinds of occupations there are going to be, what are people’s career paths are going to look like, etc. And it’s really hard to talk about one without the other because what happens inside organizations is affected by what people’s choices are in the broader economy. And the number one factor in all of these things is technology. The falling cost of communication, the increasing sophistication of the judgments made by computers. These are factors that drive both the future of working for us – so things like Slack and the many choices we now have for video conferencing and collaboration tools like Github etc. And the future of employment in the sense that it affects what occupations there are going to be both supply and demand for – that’s number one.
Number two is demographics and the single most obvious and overlooked demographic change, meaning the change in the composition of the population, is the aging of the workforce. You know we all look and see statistics about how the population in general is aging. Yet, somehow when we start talking about the future of work we go down this path of talking about millennials. Millennials are going to be essential because they are young and they will be here in the future. But if you think about the composition of the workforce it’s going to be much more important to figure out how older workers function because as a proportion of the total workforce there will be more of them than any other group.
MH: Does it touch on some of the fears that people have around automation and robotics?
RB: The Bureau of Labor Statistics does this official categorization of occupations and they classify some occupations as routine. And it’s not a perfect categorization but it’s pretty good – and it’s official which gives it all kinds of nice standardization benefits. And let’s just assume that the jobs that are routine are the ones that are more likely to be affected by automation. Well it turns out that the jobs that are routine are disproportionately held by older people. And the kinds of work that are being done as alternatives to full time jobs. So meaning not a salary W2 employee but all forms of 1099, the gig economy, the cash economy you know, call it whatever you want. Those roles are also disproportionately held by older people. There’s this myth that we’ve got this massive gig economy from all these online platforms like Uber and Lyft and Task Rabbit etc. The reality is that the lion’s share of growth in the alternative work economy has not been that – that’s a tiny sliver. It’s just been older people taking on alternatives forms of work.
MH: That’s interesting. I guess here in Silicon Valley in San Francisco everyone thinks about the gig economy but you know this has existed before right. There’s a proportion of freelancers and contractors has been growing year after year.
RB: Yeah, I mean it goes back and forth over the course of history where before we had a formalized work week and in a sense everybody was a piece worker. And the word job actually like the etymology of the word refers to a lump of work. The origin of the word has to do with doing projects and gigs. With Bloomberg Data we’ve done this research project for the last year together with New America called “The Shift Commission”. We just tried to study the fundamental uncertainties about the future work. You know when many of us talk about the future, we tend to do this awful thing, which is we make a prediction and then we plan based on our prediction. Except that almost all the history of official forms of knowledge and particularly academic knowledge shows that we’re terrible at predicting. So why we continue to insist on doing this is beyond me. What we did [in the report] was instead of saying ‘how many jobs are the robots going to take away?’ and ‘how much is the gig economy going to grow?’ We said something different which was: let’s do scenario plans.
Instead of making a prediction, let’s assume a lot of jobs would be taken away by the robots. Okay now let’s assume that won’t happen. And now let’s assume that work will continue to shift toward alternative forms of work. And now let’s see that stopped. Then we compared and contrasted all those scenarios and the truth is that you find there’s not actually a lot of difference between those scenarios. In the sense that under any of them we’ll have to deal with instability of income, under any of them we’ll have to deal with an aging workforce, under any of them we’ll have to deal with encounters with technology and the workforce, we’ll have to deal with growing geographic inequality. There are all these background conditions that we don’t focus on as much because they’re not as easy to talk about. They’re actually much, much more important.
MH: I loved the report and I thought the different future economies that you guys had identified makes a lot of sense. It’s scenario planning, and you’re exactly right. It could take on any one of those four combinations – or more likely a combination of them. So it’s good to sort of project that out and think well you know one of the things that occurred that might lead us down this path or another one. So a lot of what we spend our time thinking about at RapportBoost.AI is the future of work as constituted by human and machine. So we’re very focused on the technology that’s enabling humans. What do you think are the most promising and exciting technology innovations that will equip us for the near future workplace?
RB: In the near future I just think it is – this going to sound boring – but it is the continued computerization of all the things. Plenty of people’s daily work experience is still moving paper from one place to another. Doing a repetitive task. And that’s all going to just transform. Over a long future – which is to say the next ten to twenty years – I don’t think anything holds a candle to machine intelligence. The idea that machines can make judgments that were previously too difficult for anything but a human to make – if a human could even make it. This idea is super powerful because it will make the work experience more enjoyable for those who are working. It will threaten to eliminate certain kinds of occupations. It will create a wind behind it that’s going to make all software today look small by comparison. The economists never see computers show up in the productivity numbers, productivity just hasn’t gone up by enough in order to show us that computers have an effect, and I just think that’s because we haven’t seen the beginning of it yet.
MH: That’s interesting. You do spend a lot of time reading the literature on information technology. And you’re right – you called it the productivity paradox – because everywhere you looked computers were changing the world and the one place you wouldn’t see it is in research papers where they find computers were having no effect. I think that makes a ton of sense. As it’s shifting to more and more white collar work you’re going to see more and more of these massive transformations. So let’s shift gears a little bit. Let’s talk a little about artificial intelligence. You’re an early stage investor in a few AI companies. What do you think are some of the most exciting developments in this space.
RB: So first I think that it’s hard to say because the numbers aren’t public and many of our investments are unannounced. But there’s a chance we’re actually the most active investor in artificial intelligence. And the reason for that is that about six months or a year before the rest of the technology world started jumping up and down about AI, one of my partners spotted the trend and was reading some academic papers and talking to people who are a little ahead of the curve. And she said hey, we should take a look at this. Her name is Shivon Zilis. And I actually discouraged her from thinking about it and I was totally wrong. Because she did a few months of work and then we looked at it and said oh boy, we should just kind of turn our whole fund over to start focusing on artificial intelligence. And you know the phrase artificial intelligence is actually misleading. We tend to use the phrase machine intelligence when we write about it just because it [AI] carries a connotation of a robot that thinks and acts and looks like a human. When really what we’re talking about is a class of technologies that predict things that you have not yet seen based on things that you have seen and do all kinds of other judgments that previously computers couldn’t make. What I tend to focus on are the very mature forms of this technology – I mean in a sense it’s just a linear regression in the most mature form of it – and how we could use those technologies to deliver immediate business value. So companies like Textio where we’re investors that look at a job description as you’re editing it and use data about past job descriptions to make recommendations to real time about how to change your writing so that you can get better results for applicants, more diverse applicants etc. That to me is a quintessential example of this kind of applied machine intelligence which is exactly what we’re really excited about now. Orbital Insight does this looking at satellite imagery. Applied doesn’t mean easy. These things are still very, very difficult to do but the technologies are mature we’re just figuring out how to make them useful. The technologies are relatively mature and we’re figuring out how to make them useful in a business context. Sort of like the Internet in the late 90s you know the technology was there, it’s been used for a while and now it’s about making useful.
MH: And solving a problem. That makes total sense. I think it’s funny you mention how artificial intelligence kind of became the buzzword of the day and it sounds like you were ahead of the curve, so kudos to you guys. I think relatedly I’ve seen that in venture communities and I’m wondering are there any technologies that you think are over hyped or scary or just things to be cautious of.
RB: Artificial Intelligence in general is overhyped in the sense that lots of people are claiming Artificial Intelligence for things that do not have AI in them. And you know if I can claim that anything that has a regression in it is AI, well then all of a sudden it’s like, we’re an AI company, let’s add .AI to our domain name. So in that sense there’s too much hype. That said I think it’s hard to overhype the actual potential of this technology. And in that sense it’s also very similar to the internet in the late 90s where any average company you looked at was probably overselling itself but in totality all of the companies were actually underselling the totality of the opportunity.
MH: I tend to agree that it’s not AI or any regression it’s, what problem is it’s solving? How is it making life easier? How is it automating work? So on the subject of optimization I know you mentioned Textio. You invested in Textio, Digital Genius. They’re really about optimizing language saying the right thing to the right person at the right time and it’s a similar thesis to what we have. What gets you excited about those companies or just communication in general?
RB: Well you know language is the kind of tool that is definable enough. You can collect data on it and you can share results on it because you’re ultimately talking about strings of text. And yet it’s so powerful especially in a work context. So if you’re looking at customer service or sales – – you know we’ve all had that feeling of you’re on a sales call. You’re talking to somebody trying to convince them of something and have this feeling of, if I only knew the magic words. Or if I only knew what was going to happen here. I’ll give you an example in a sales context which is if you’re on a sales call and you know the call isn’t going to work, well all of a sudden you could do a lot of things like stop talking to the person and save your time for other things. We’ve had a version of this predictive technology for a long time in the sense that we’ve been able to A/B test things and send them out into the world and see what happens. But once you make the feedback loop fast enough where you can use it as you were doing work for example if you’re doing it with live chat, it’s much more powerful.
MH: We definitely agree and we think learning from millions of conversations or in the case of Textio, job postings – doing that at scale is really exciting. You can potentially do better than a human copywriter right or a human agent.
RB: Of course we all know that you still need at humans right now for so many parts of this and I think seeing where that interface is, and seeing what a partially machine partially human system come up with. You know in the world of board games everybody’s been watching Google’s Deep Mind and its progress with winning at Go which is again lots more complex than chess. The interesting observation a lot of people have made lately is that the machines have suggested all these creative strategies that humans never thought of and it’s now making humans better players and my guess is that the machines are going to have to ‘up’ their game in order to beat these newly inspired human players.
MH: Sure, it’s human versus machine. You know, it’s great.
RB: Yeah, I watched War Games with my son over the weekend and it reminds me exactly of that where we’re encountering all these questions about us and our tools and where is that line? They’re age old human questions. We’re just dealing with them in an accelerated way.
MH: Yeah, great movie by the way, good call. It’s kind of becoming more relevant right?
RB: I have to say, we’ll have to see whether Russia wins the Cold War.
MH: I love it. So last question for you. For founders of early stage AI companies, what advice do you have, and if someone was approaching someone like you for funding, what advice would you give them?
RB: I’m not great with advice because I kind of think all advice is either the person trying to justify the decisions they’ve made in their own life or to use you as an experiment. So I don’t know that I have advice for what people should do in part because the great founders are going to selectively ignore great advice and then do their own thing. That said, it’s no different in AI from any other kind of business which is the faster you can make something useful for someone, the better off you will be; end of story. Everything else is nonsense around making something useful.
RB: I can say if somebody comes to pitch us we’re very early and so it can be a concept they don’t have to have made anything. But if they could have made something in the time it took them to make their pitch presentation, if they could have built a product, then that will cause a red flag. And at the same time if they’re more in love with why it’s useful than with the tool that’s generally a really good sign. And then, we post our criteria for investment publicly on our website. We try to be must transparent mentor investor and we hard to keep those updated so that as I learn what we’re actually deciding I update and amend those public criteria. It’s really no different for an AI company than for any other company which is ultimately we want to tick some boxes. Like do we trust this person, is the deal fair? What we’re really deciding is based on: is there one concrete reason to believe that this company has a strong shot to be an outlier.
MH: Cool, well super helpful and I disagree with your assertion that you’re bad at giving advice because you’ve given me advice personally and professionally and I found it all to be very, very valuable.
RB: Thank you. It helps when smart people are doing smart things anyway because I can take credit.
MH: I appreciate it. Roy thank you so much for joining us. This has been great and yeah we’ll obviously check back in with you at some other point and we’ll pick your brain a little bit more about AI and the future of work.
Learn more about the live chat agent training from the team at RapportBoost.