The Exchanges

Q What's, uh, what about what you feel is played out a little bit? If you're an entrepreneur in this room and you're thinking of, uh, leveraging your, your skills to start a company, In big data, you should not start company X.

A I think you have to be careful if you're just looking at the visualization component. I mean, look, we all, Apple's indoctrinated, all of us, you know, build beautiful products, and I certainly had that hammer in my head when we built WebOS. Um, and you need, certainly, like, you need to have a great way to manipulate data. There's no doubt about that, but it just feels like between D three and other tools, that that area, it's, it's, it's getting more difficult to differentiate just on visualization. You have to have other pieces. You know, earlier on, you know, data governance was brought up. I think that's a, Still, like, a big open area on, in the lineage side, especially when you start linking into things around, like, SOX compliance, and, like, what does that actually mean? Um, what are the other areas? I think.

AI assessment note: “getting more difficult to differentiate just on visualization”

Q Things that creates this, uh, this need for it. Um, great. Um, you talked about data science, people winning Kaggle competitions. What's your, um, thinking on data science being democratized? At the end of the day, do we have a bunch of tools that enable anybody to be a data scientist within a company, or do we, uh, as startups, do we build tools to enable data scientists, or Both?

A So, I think that the data scientist term may be a little bit overloaded. Um, maybe just a little, ah, like big data is a little. Um, I mean, look, there's, I think there's always going to be a need for someone who has a background in statistics and can understand how to use tools to understand what are the questions that should be asked, how to do the feature engineering, to understand your business, whether it be in financial services or in healthcare or whatnot. Um, I do think that the incumbent players like SAS, Uh, I mean, I think over time will hopefully be dislodged by, you know, some number of startups. I think at the same time that there will be more and more self-service tools to enable people to, maybe not everyone becomes data scientists, because I don't think that, at least in my definition, I don't see that really working, but there'll be more accessibility so people can ask more questions of the data. One of the things that we built at Twitter, and it was absolutely borrowing off of the, A third of the people that I recruited from Google was how could we, you know, let non-engineers run MapReduce jobs, ask more questions of the data, because if we let more non-engineers ask questions, there is a higher probability that we would actually iterate on things, whether it be the product or just how we look at the business faster, right? So it's kind of like you end up d…

AI assessment note: “maybe not everyone becomes data scientists... but there'll be more accessibility”

Q So, so where, where do you see, um, uh, from your VC perspective, what's played out? Where do you see the white space? So you started alluding to this through, uh, apps.

A Yeah, and actually I didn't even finish. So, uh, I'm at Kleiner Perkins, a general partner, joined us three years ago, and this is certainly an area that I, I invest in. I think, um, an area that we're starting to see, I think some really interesting applications being built around some of the data mining is in specifically around CRM and in, in, in, in Salesforce. Salesforce recently bought a company called RelateIQ, and I think there's some really interesting work that's going on to think through how do we actually do better Salesforce casting Within companies and using machine learning so you could actually see, you know, something that's more complicated than, hey, you haven't reached out to that sales lead in, in three weeks. Go reach out to it again. Versus what's the dialogue? Mining email. I think just mining email in general. I think we're seeing a lot of interesting, um, more B to B type applications emerge. Um, and so I think there will be others, but that's certainly one obvious one right now. I think that's actually really pretty quite exciting.

AI assessment note: “an area that we're starting to see, I think some really interesting applications being built”

Q What worked, what didn't, what was surprising maybe?

A I think, you know, it's an interesting thing right now where the, ah, the cost and ability to store tons of data is, is so prevalent both in consumer and enterprise companies that I think oftentimes you end up in this kind of data obesity state where you start storing data for reasons you don't even know why you're storing it. And so we try to be reasonably disciplined around thinking through, we're going to store this data, whatever that data be, whether it be logs on users or systems, do we at least have an idea of how we might use that in the future, if we were not going to use it right then and there? Because I think it's very easy to store a bunch and you go, oh, you know, great, like, I've got this big Hadoop cluster. Um, but it turns out that, like, if you don't use it, Um, it doesn't really matter, and actually, I'm an angel investor in Cloudera, and I know Jeff really well, and I think one of the things that's been a challenge with Hadoop, whether it be, you know, Cloudera, MapR, or Hortworks, is that, um, they're solving great problems and making it really efficient, but now a lot of larger companies are asking, where's the ROI, and, you know, how do I get production workloads, like, how do I move those production workloads? I think we're just early on getting, getting that done. I think it will happen. Um, but I think we have to kind of, as a collective group of both…

AI assessment note: “one of the things that's been a challenge with Hadoop”