Thursday 7 November 2019

Graphistania 2.0 - Episode 1 - This Month in Neo4j

Hello everyone!

it has been deadly quiet on this page, hasn't it. That's really oh so true, and I am / was not happy with that. This blog, the podcast, and everything around has always been my humble contribution to our awesome Neo4j community, and in the past 6+ month or so, I have not been doing my part. Sorry for that. Lots of excuses that I will not bore you with, but I am going to try to do better.

Part of the reason for the silence was of course that I thought that the podcast formula (in which I always asked for the three same basic things: who are you, why graphs, what's coming in the future) had kind of run its course. 100+ episodes had given me lots of fantastic conversations, but it was time to move on. I needed a new formula.

A couple of weeks ago, while doing absolutely NOTHING graph related - unless you want to imagine a graph of a bathroom, a shower, soap, and yours truly - I came up with an idea. What if we did episodes about all of the cool, innovative things that are popping up in our community on a daily basis? Sure. But where could I find those? Well, on the Neo4j developer relations "This week in Neo4j" (TWIN4J) newsletter probably, right! But who would I talk to that about? Well... this is where I found a great partner in crime. I thought about one of my most creative colleagues, someone who is paid to be creative and is really good at it - and came up with noone other than Stefan Wendin. Stefan leads our Innovation Labs in EMEA, and has presented on that topic extensively in the past.



So we have lots of innovation. We have someone who KNOWS a lot about innovation. So let's then have a chat about some of these innovative graph database applications, shall we? Here goes.





Some things DO stay the same, so here's the transcription of the podcast recording:

RVB: 00:00:00.000 [music] Hello, everyone. My name is Rik Van Bruggen from Neo4j and it has been a long time in the making. Entire summer without any kind of podcasting. But I want to do it again. I wanted to start this stuff again. It's been so much fun in the past and should really get back to it. To do that I've kind of come up with a new idea to produce some interesting episodes for you guys. The idea is basically that we'll be looking at all of the stuff, the interesting stuff, the interesting use cases, the innovative use cases that are coming up in our community on a daily basis really, and talk about that. Just highlight some of those, discuss what they might mean for us, how people could explore it and all those types of things. And we're going to do that using some kind of a baseline content which is in the This Week in Neo4j (TWiN4j) newsletter. I'm hoping you're all subscribed to that. It's a fantastic, useful newsletter produced by the Neo4j developer relations team and it's a fantastic source of beautiful content. 
RVB: 00:01:27.843 But a podcast without guests would be extremely boring. I know you don't want to listen to me all the time. So I've actually invited a dear friend and colleague from our Stockholm office to join me in this journey, and that's Stefan Wendin. Hi, Stefan. 
SW: 00:01:46.912 Hi, Rik. Very nice to be here. Really looking forward to this and to all the listeners that did not subscribe to the newsletter, don't worry. I didn't either. But I've found it actually extremely helpful. So this can also be a starting point for a very beautiful journey together. So... 
RVB: 00:02:05.900 Let's see. Yeah. 
SW: 00:02:06.926 Let's see. 
RVB: 00:02:06.803 Yup. And so that's what we're planning to do. So literally we're hoping to do this on a monthly basis. It's probably going to be more on a whenever I feel like it basis. But that's okay too. This is a community thing. This is not a Neo4j thing and we'll dive right into it. Stefan, let's talk a little bit about some of the posts that we saw in the past month or so. I've been looking through the This Week in Neo4j newsletter and I found some really interesting things. But what did you think? What did you find? 
SW: 00:02:40.613 Oh, yeah. There was a couple of ones that really stood out. I think we got the chatbot of course, the political ones because also that's a foundation of building a new society in a sense. But then again we got also the AI and ML kind of thing adding context to that. I think also super interesting. And then I actually missed the last one which was about the map. Funny story, I see Rik playing around with it in the office and I go, "Holy crap! What is this cool feature? I've never seen this one before." And then Rik, very relaxed said, "Yeah. It was in the list I sent you [laughter]." 
RVB: 00:03:19.841 So Stefan, let's talk about those a little bit, right? So let's start with the last one. Visualising nodes on a map. There was a post a few days ago, I think, with someone, I think Stella-- what's her name again? I'll put it on the blog post. She created this Neo4j desktop application that basically takes graph data from a Neo4j database and visualises it on a map. What do you think is most interesting about that? 
SW: 00:03:51.058 I mean I think it comes down to the foundations of what I love with graphs and there's so many things. But I always keep coming back to this kind of Alfred Korzybski. Was that the name? This kind of 'the map is not the territory' kind of thing. So before joining here I always have a hard time kind of seeing my data represented as a table because it didn't look anything like the real world. It couldn't handle complexity in anything. And then just kind of adding this kind of real kind of map layer on top of it and the possibilities to, as we found out by just trying-- which is again the best way of trying. Adding different kind of layers immediately you start seeing, "Oh, but we have the shops in these kind of areas, we have the clusters of possible prospects of customers." And then by just overlaying that in like a millisecond, you have a visual kind of explorative thing that everyone could take part of. And I think this is one of the beauties which I like where the lot of the new stuff is that it enables people, not necessarily from the data domain or the data science or developer community to take part in discussions around data. And I think if we can solve that and bridge that and anything that helps in doing that I think is adding business value and-- 

RVB: 00:05:07.464 Fantastic. Well, I mean I will post it on the blog post but I do think that with some of the spatial data types that we have in Neo4j and these beautiful, very simple desktop apps or visualisation integrations that we have these days makes it really easy. Let's talk about another one. For example, the political post that you mentioned. There were two of them, right? 
SW: 00:05:30.217 Yeah. Two of them. 
RVB: 00:05:30.927 One of them around British MP voting and then there was another one-- what was it again? 
SW: 00:05:37.197 It was this kind of lobbying in the US. 
RVB: 00:05:39.228 Oh, the lobbying in the US. In the US. US lobbying data. I mean both of them are super impactful, right? To influence politics really is-- 
SW: 00:05:48.686 Yeah. And also to show what is actually being done at the moment. We have a lot of these kind of discussions. I think it's pretty much all over the world that we have this kind of far, far to the right blaming a lot of let's say sometimes immigrant and a lot of other people's blaming them for the thing but as we know that's most likely not the case. So we tend to have this very biased discussion towards different sides of the arena and not looking upon the visualisation and seeing what's actually going on. And I think to have this it would actually be mandatory for every single country. Imagine where you can see this is the person, this is how they voted, and this is the bill they actually kind of pushed forward. And then also like a following up. This is what actually happened with that. If that would be visible and kind of remove the friction in looking upon it because I mean the information is there as of now, of course, but for a normal person it would be pretty much impossible to kind of see any patterns or follow-up anything or you need to spend a lot of time. So I think just creating that would be just an amazing. So Swedish government, if you hear this, call me up [laughter] and let's put it together. No, but as a citizen of Sweden, I would love this. I mean this is-- 
RVB: 00:07:08.321 I mean there are so many examples of that. I mean we obviously had the Panama Papers a couple of years ago but I have this friend of mine in Belgium that basically created a website called "Open the Box" that brings together corporate data, corporate information, of who's in which company, shareholders of those companies and then, and this is unbelievable, political mandates. Which politicians hold mandates in which companies, in which organizations, da, da, da, da. This is the type of stuff that the government has been trying to do for decades and that a guy in a shed in Belgium is able to do in three weeks on his holiday, right? It's crazy powerful how open data and obviously the right tooling around open data is making these things just so much more transparent and hopefully kind of also responsibilising citizens, right? I mean if we have the data then we can do something about it. If we don't have the data then we just going to sit on our couches, right, and we're not going to do anything. So really interesting articles there. 
And maybe let's talk about another one which there was a beautiful presentation, I think it was at the Nodes online conference that we hosted a few weeks ago by Alicia Frame on the intersection of graphs and artificial intelligence, machine learning. There are some really, really interesting insights there. What's your perspective on that? How do you see that evolving? 

SW: 00:08:50.340 This is one of the big paradigms. I think we are, again, as-- at least I always do. In the first year, I say, "Oh, holy crap! This will change everything." And then in a year - a year goes quickly - nothing really happens. But in 10-years time it's going to rewrite history as we know it. So I see this kind of idea and how it actually adds context because pretty much all of the AI stuff is now coming from tabular structures or built on the foundation of those. And they're good for certain things. Again, a graph is good for other things. So I think it-- it actually reminded me of one of my favourite stories. I'm going to do a derailing here and [crosstalk] 
RVB: 00:09:32.383 Go for it. Yeah. 
SW: 00:09:33.245 Yeah. So I remember, I think it was back 2011 - and we can post this in the comments down below, there was this kind of thing that the scientific community tried to solve like uncovering the cure for HIV. So they have a problem with protein foldings. So they have been trying to solve this for roughly 10 years. I think they solved 45% - the scientific community working on this. And we know that they have pretty much unlimited budget in this, right? They had time. 10 years is a awful lot of time. But then they kind of-- and that's a long story but basically what they did they created an online game which is called 'Foldit' and then they let people online basically play a game of protein folding. And in 10 days, I think, they solved 35% or 85% of this. 
RVB: 00:10:26.578 No way. 
SW: 00:10:26.714 And over the weekend they solved 100%. So in 14 days, they solved something that the scientific community couldn't have done. So for me, this is a story about this kind of deep knowledge but then added a new perspective. And every time I look on transformation, this is what I look for. Do I have the deep knowledge? In this case, we had the usual stuff that we usually do, right? And then we add a new perspective to it. And that new perspective's going to create and double the value of that deep knowledge. And if we have a system that can handle it, it can be a process of an organisation or in this case a graph and machine learning-- 
RVB: 00:11:02.020 Machine learning. Yeah. Exactly. 
SW: 00:11:02.900 So for me, this is kind of-- I could just see how this can literally blow the ball out of the ballpark. Yeah. 
RVB: 00:11:10.789 Fantastic. 
SW: 00:11:10.363 So super excited about that. 
RVB: 00:11:12.256 It's a really cool use case. We've also been able to move quickly in that space thanks to the Neo4j labs, we've got algo library now, we've got the NEuler prototyping tool in the Neo4j desktop. It's actually becoming really accessible. Even a walk-about salesman like myself [laughter] I can play around with it and actually do some interesting things. 
Let's talk about that one maybe as a last topic. Something that's actually quite salesy in the sense that I've seen it come up in a number of conventional projects for Neo4j. And some of our most amazing use cases are actually related to it. It's the chatbot, right? Max De Marzi was building a chatbot in Neo4j. Wrote it up in a bunch of articles. I think it's such an amazing use case because it's basically-- it's so easy to calculate a value for it. Right now people are calling up the call centre. Everyone needs to get him to go set the queue. You're the next in line. Blah, blah, blah. And then you get bad service, right? That's basically how it works. And these chatbots, eBay is using it, a bunch of insurance companies using it. They're actually providing tremendous value based on that. What do you think about that? 

SW: 00:12:46.494 No, I think that obviously this kind of money value as a company you can say a lot of things, but if you double click on this and I have this kind of you know what happens when you double click on it. It unfolds, right? So I usually use that a lot in my actually normal work. I double click on PayPal [laughter], [inaudible], and stuff. So I think this chatbot thing, I think what stands out, if you spend a little time researching this, I think it's the interesting part is we tend to actually be more open to chatbots than we are to a normal person. This will, of course, change over time, but this lets me into this other way of thinking because-- you talked about eBay, right? So let's think about eBay. So if I go to eBay, I have already an eBay behaviour. But if I got to an eBay chatbot I will most likely not have a behaviour for it because it's a new thing. So then I will default to how I normally talks and this allows, I think, for brands also to understand how are we perceived. So coming from almost this kind of brand and strategy perspective because if we can understand how words create sentences, right, since they are mapped in nodes over time, then we can start to also understand also how to talk to customers in the way they actually would like to be talked to not in the way that we designed it five years ago because I think this is also this kind of the consumer webs kind of big paradigm. I mean let's do Snapchat as an example. I think that's a great one. Every single one of above the age of 12 complained about it. "I can't get it." No, the reason you can't get it is because you have a history of left [inaudible] the web menu in your backpack. So kind of our exploring the new possibilities is kind of also limiting in your kind of mindset. So I think this is one of the double click. It looks like a very simple thing but insights that you can draw from this based on the conversations that you also have analysing those, I think this is a very untapped potential from a whole other part of the organisation actually, to be honest. 
RVB: 00:14:49.254 To me it's just a win-win-win, right? 
SW: 00:14:51.659 Yeah. 
RVB: 00:14:51.944 It's literally the consumer of the chat wins because they don't have to queue up in the morning or hold the line in the call centre queue, right? They get quicker service. They can get it whenever they want, they can probably get it faster. The provider wins because it's actually very cost-effective, right, and there's really-- I don't see a reason why people wouldn't be jumping all over this. The technology is getting there. It's really something that we should be exploring a lot more into and we've seen a bunch of really great cases there. 
SW: 00:15:32.211 Yeah. Now just thinking of it also like-- because sometimes we actually want to talk to a person but most likely, in the most cases we don't. We just want to have our problems solved. 
RVB: 00:15:41.432 Problem solved. Exactly. 
SW: 00:15:41.831 The less friction of solving the actual problem which will then create more time for the customer service agents to actually talk to the people that needs that conversation, right? So it's a win-win. Or as you said, a win-win-win-win [laughter] in most cases. So I think yeah, if you haven't tried this out, go check out Max posts and just set it up. I sent it over to a client the other week who was like, "Stefan, can we build a chatbot with Neo4j?" 
RVB: 00:16:10.044 Yeah. By tomorrow [laughter]. 
SW: 00:16:10.266 Yeah. By tomorrow. Here's a post. And when I send he was like-- and actually he was like, "This is so easy now. I don't want to do it." 
RVB: 00:16:16.295 Yeah. Yeah. 
SW: 00:16:16.429 And I think this is a process kind of cool. 
RVB: 00:16:19.305 All right. That was a great first pilot, I would say, of our-- 
SW: 00:16:22.960 Yeah. This went quick. 
RVB: 00:16:23.384 This went quick, didn't it? And I hope it went quick for the listeners as well. Obviously, as usual, we'll get a transcription up on the website and the links to This Week in Neo4j newsletter and then hopefully we'll be able to do this on a monthly basis going forward. It was a lot of fun. Thanks a lot, Stefan. 
SW: 00:16:42.304 Yeah. Super fun. Okay. Thank you. 
RVB: 00:16:44.688 Thank you.
Hope you guys thought this was a good start - as always, your feedback would be very welcome!

Subscribing to the podcast is easy: just add the rss feedfind us on Spotify or add us in iTunes! Hope you'll enjoy it!

Cheers

Rik

No comments:

Post a Comment