Artificial Intelligence is finally in mainstream news almost daily. From self-driving cars to intelligent drones, medical chatbots to insurance analysis algorithms AI has gone mainstream in a big way.
We web developers, of course, have known all about this for some time. We use AI every day to build sites, don’t we? What would we do without our trusted bots providing helpful advice as we pick colours and typography, create content, define the layout and launch a site? How would we optimise our sites without having our bots tap into big data, fire up their neural networks and provide us with the perfect website recipe in order to maximise conversions? What would our users do without a bot to converse with them and lead them to the best choice every time. Thanks to AI web development has never been easier, and every launch is a success!
Sadly, unless you are from another dimension that is not actually what web development looks like right now. So where and, more importantly, what exactly is the grand AI promise for web development. I have been looking at all the different ways AI will influence (and has influenced) web development. This talk prepares you for the inevitable changes AI will bring to web development and points out what technologies you can take advantage of today as well we what you need to be thinking about so as to be able to better handle what is coming.
The bots are actually here and more are coming. Let’s make them our friends.
DENISE: Hello everyone, and welcome to track B. Please pick up your T shirts from the registration desk downstairs. We have a ton of T shirts left you can even get an extra one if you want I want to thank our sponsors Timpani WooCommerce, Jetpack. HeartInternet for our sales sponsors. SiteGround, 34 SPW com, WP Engine and GoDaddy and our Dress Circle and grand circle sponsors as well as balcony and patron sponsors. So I am so excited about our next talk which will be by Ronald. Ronald has a wonderful story about how he got interested in technology, he is based in Italy right now and in Sicily an amazing place where he works remotely at the digital agency Deeson. Ronald actually has a PhD and studied artificial intelligence and then when he moved to Italy started getting involved wit web design and Drupal and his work has come full circle he is looking at art finitially technology inside web development. So let’s give a warm applause and welcome to Ronald (Applause).
RONALD: Thank you. I am going to have to start with an apology. I had a cold, pretty bad cold and I have a weird very deep voice. This is not my normal voice. And then it sometimes turns in to a 12-year old voice and can go really high. We will see how this goes.
So yes, I work for Deeson I am the technical strategy director there. Deeson is hiring if you are into UX, looking for UX Director, WordPress develop developers. Solution art et ceteras and so talk to me after this talk.
I am going to try to convince you about two things, that AI, artificial intelligence is going to change the way we develop websites and that this is going to happen quickly. That is the important bit and then I am going to talk about what we do about that.
Before I go on how many of you are web developers, build sites day in and day out. So great this is going to be very relevant to what we will be doing in a few years.
So we are going to talk a bit about what AI actually is. Then look at technological progress from a slightly abstract stand point how technology progresses in general. And that is important to understand the consequences. We talk about how web development is changing and what should we do next.
So what is AI? It is not Sky net it is not the terminator. This is important because when people mention AI, when I finished my studies and people would ask what did you do and I say artificial intelligence, they say oh you are building a terminator or something. No it is kind of more mundane and worse at the same time, it really is going to change things. I am going to give you a view of AI that is not philosophical or anything like that. Not about consciousness, it is a view that helps us think about how we build things and it comes from age based software development. How do you build software that has intelligent software within it.
So what is an agent. An agent is anything that can perceive the environment through censors and modify. Think of a thermostat. It perceives the temperature of the environment and then it switches the heating on and off to change that environment right through effectives. First level in agent based software is something called reactive agents, they are very simple and I am going to be using chat bot as an example here. How many know what chat bots are? Okay cool. It is programmes that live within something like messenger or Slack and you can talk to them and they respond and do things for you. So a reactive agent chat bot would do something like just sit them and it is only going to react once you talk to it. So if you say “hey how are you doing” it is going to reply “I am doing good how are you doing?” It is not trying to achieve anything it is just sitting there waiting for something to happen. And from an AI perspective that is not necessarily interesting, but it is the first stepping stone. Then we get into something a bit more interesting which is a proactive goal oriented agent. Now a proactive agent is a piece of software that has a goal. What is a goal? It wants to change the environment from state A into state B. And it is going to do that proactively. So it is not just going to sit there and mate for you to talk to it, it is going to actually try and do it. This is when things start becoming interesting. Because we are relinquishing control. We are letting programmes do things for us. So in a chat bot scenario you would have something like a chat bot with a goal to take the world from a state where Ron did not read the news where a state where Ron read’s the news “you will not believe what happens today, click here to read the details” and this is things we see happening on chat bots and we write applications that do this. Let’s take it a step further. So you can have a proactive goal oriented agent that is also learning. It wants to change the world but it is not just going to do it trying to do the same thing evidence and over again. So with our chat bot example it is going to say “you will not believe what happen today, click here to get the details” and I ignore it. So try something else “click here for the latest news” no result. It learns what is going to make Ron click. And it changes a different approach “Dear Ronald, all the things you need to know here” that does not work either. It keeps using some facility function to figure out what makes said. If it said “My masters will delete me if you don’t click here” I would click. You have a piece of software proactive, trying to change the world in some way. And it learned based on very much things it tries and it is still very simple. 20 lines of code and just something that assigns values to, but this is still AI. It is very simple, but the moment you start relinquishing control you are building something that is changing the would on its own. Now, next level and we are done, are autonomous agents. And autonomous agents are more interesting they are not just driven by goals where goal is a very specific thing change from state A to B they have motivations and they can be things like increase user engagement. Just in general. And then they have specific goals like get the user to read the news or subscribe or sign up to facebook, stuff like that. And there can be any number of complicated mathematically interesting function there that try to figure out how am I going to get maximum utility for this motivation of increased user engagement putting in place actions that are going to achieve those goals. The basic level it is simple and that is what I am trying to drive through. That relinquishing control. Not just trying to get the user to read the news, we are saying there is this high level thing which is user engagement. And then it can get like transistors they start simple and they get increasingly more complicated. Integrated circuit. And then you imagine you can have different motivations and those motivations had to different goals so increases engagement but it is also increase add income. So then do you know do you get the you store click on an ad, that is going to take them away from me and my ability to increase engagement. And conflicting things and you are really increasingly losing control. You don’t know what the programme is going to do. You know you have overall utility functions that are going to lead it in the right direction. Take this into car driving, why do we call them autonomous vehicles and phone everybody is excited and scared about how autonomous vehicles are going to change the world, because we are letting these cars figure out how they are going to get us from A to B. And the goal maybe get the passenger back home but the motivations can be things like do this in a fuel efficient way, so you don’t know exactly what route it is going to choose. It can be don’t kill the passenger before you get them home, that is a big motivation. Don’t kill anyone else while you are taking the passenger home. And those motivations are going to conflict at some point. This car is going to have to decide there is someone in front of me and there is a wall and a passenger, who do I kill, that is a real question that autonomous driving vehicles are going to have to answer. So that is where AI starting from simple building blocks becomes a more interesting thing because we are relinquishing control, letting a piece of software decide what it is going to do. Okay. So leave that to the side for a bit and talk about technological progress and how progress happens. There are two concepts I want to talk about here. Exponential growth and interconnectiveness. So what is exponential rate of change I think you have seen graphs like that, things start slowly and take off. A great book called “the second machine age” which talks about what exponential rate of growth means, I definitely recommend. And where we hear about exponential rate of growth most is MOORE’s law. The number of transistor in an integrated circuit doubles every two years, this line is straight it is a logarithmic scale. So this has been extremely consistent over a long period of time. Now the problem with exponential rate of growth is that we as human beings are really bad at dealing with numbers, specially big numbers like we don’t get it. So, I will tell you a story about chess to talk about exponential growth. And the story goes like this. Once upon a time a ruler of Indian was bored and asked a mathematician to come up with a game and he invented chess. I don’t think it was how it was invented put it is is a great story. The court was really pleased with chess it was a fun game they were not bored any more and they asked the mathematician how do you want to get paid? The mathematician said I want to be paid in rice and what you are going to do is place a grain of rice on the first square of the Board and then on the second square of the Board you are going to place two grains of rice because you are going to double what you did before, and you are going to keep doing that so it is four, and keep doubling the amount of rice that you place on each square of the Board until the Board finishes. So it is like 64 steps right. So I think where are we now. 128, 256 grain of rice and the ruler was laughing here is some more rice, okay 256. 512 grains of rice. Okay, here is more rice. And I wanted to do a great slide where this went crazy with grains of rice. But keynote bombed it can’t handle more than 512 grains of rice. So here is what happens, the numbers grow very quickly. So the first line think that is like the 70s the first two lines and then the 80s and it is okay here is a bag of rice and here is a Barrow and here is a truck. By the fourth line that is two billion grains of rice. A big rice field a lot of rice and then it is four all of a sudden and then by the time you have finished the fifth line it is 274 billion blah, blah, blah. So how does the story end. There are two endings the first one is that the Royal goes bankrupt, they cannot supply enough as to complete the Board. The second ending is that they just chop the mathematicians head off and went back to playing chess which is probably the most likely outcome of that scenario. Now why this is interesting? This is interesting because we are actually 32 doublings in. As humans in terms of computer technology we have done the first half of the Board. And this is not me saying this, this is Ray Kurzweil Director of Engineering at Google and a really famous AI scientist and Director of Engineering at Google. So make the connection there. We are starting the second half of the Board. Every next step change is going to be so much bigger than everything that happened before. That is what exponential rate of growth means. That is why I think this is going to happen very soon because the rate just goes like this. So everything we have seen happen in the past 30 years it is just going to double. Think of when the iPhone came out. What was that ten years ago, eleven years ago. Now everyone has a Smartphone, those Smartphones have all sorts of censors and we just use them and get annoyed when they don’t quite work. That device is crazy, that thing recognises your voice and does things in the real world because of that and that just happened in the past few years. Exponential rate of growth. So that is one concept to keep in mind the future is going to change much faster than the past did. The other concept is things are inter connected. And another great book called the inevitable. And it is talking about technological processes, things that are by their very nature leading technology into a certain direction. And I am going to give you one example of inter connected, it has to do with machine learning. Machine learning has actually been around for a very long time, since the 1950s, let’s just put it there. It has not been very exciting. So a few things had to happen at the same time for machine learning to get exciting and they have happened. The first one is cheap parallel computation. In order to train a neural network you have to run a lot of processes in parallel and that takes a lot of computation. And that is expensive. But, at the same time as neural network scientist trying to make efficient neural networks teenage kids were really excited about shoot them up games and when they shoot someone they should see that head explode in full realism. As a result we got graphical processing unit chips GPUs from Avida that are extremely efficient and the cost went all the way down because teenage kids were buying them. So 2011 Andrew NG who used to be Head of Google, and is now at. So he said let’s get a bunch of GPUs and put them together and solve neural networks that way and problems that would take weeks to solve they solved in a day. So that happened. Other thing that happened is big data. I am very pleased with that. How do you represent big data. So AI feeds on data. In order to train a neural network you just need a lot of information and we did not necessarily have that information, we had it a training set, then storage got cheaper and networks got faster and database technology got better and Google and facebook started collecting data and suddenly you have a lot of data to train neural networks. The other thing that happened better algorithms. In 2016 someone called Jeff Hinton working in Canada with his research group came up with a better way of running neural networks and they called that deep learning, so when you have deep learning it is another algorithm for running your networks and that in increased the efficiency by a factor or several factors. One of the first paper they wrote, I don’t expect you to read this, something called image net classification with deep convolutional neural networks. A competition to see how well neural networks perform. Just look at the numbers there. 102 million high Resolution images categorise across a thousand different classes, the neural network 60 million parameters and 650,000 they use GPUs in order to solve the problem and they got an error rate of 15. 3%. So 85% of the time they were correct. Current error rates for image recognition is 5%. That is better than humans. So computers are better than us at recognising an image right now. So, exponential this all happened in the past five years. Computers went from being other for image recognition to better than us. They are better than us in recognising cancer in scans. They are better than doctors. There are all sorts of things in the past two years computers started being better than us, and that is going to impact all sorts of things and how we do them. Okay, so we have AI, we know that the rate of change is exponential and we know that changes are inter connected and months we come together the shift is factors of magnitude more. So how does this impact web development and what we do day-to-day? Well let’s think about what problem we are actually trying to solve. What is it that we do? And if the answer is build websites that is the wrong answer. Because who needs a website? There is no intrinsic need for a website in human beings. People need solutions to their problems, they want to find clients, they want to sell widgets and hire people and retain et cetera et cetera. So what if we can solve all of these problems without websites, that is the question we need to be asking ourselves because we cannot depend on people needing websites and if they don’t it is a bit of a problem. So, we have two questions we need to answer. Do we need a website and if we do need a website how will it be built. Let’s tackle the first one. I am going to give you an example with Deeson’s own website. Here is my user story. As a user I want to find where your business is and I type Deeson agency into Google and it gives me some results fantastic. Click on the first one. I go to Deeson’s website and it does not say where they are, so contact maybe I will click on contact, it is a very nice picture of our managing director. But I have to scroll down and finally here is where they are, London and Canterbury and remote. But actually I lied because when I typed Deeson agency in Google this is what Google actually gave me. Deeson is exactly there, this is the phone number and their opening times and you know if you are a restaurant or a hotel or any sort of business of that sort you did not need a website to solve that specific use case. And it can get better. Because if I have Alexa I can say “Alexa where is Deeson agency?” And it is going to tell me. I did not even use the web. I used the web but I did not type anything. We need to ask ourselves what are the problems we are solving and how are they currently solved. Once you are in a situation where you also have dedicated market places and by the way all these things work efficiently because they work with artificial intelligence in the back end. Google knows where Deeson agency is because there is a tonne of things in the back end. Take airBnB. It has all the data and the processing power it will give you a best price why should you build a website for that. The same with Amazon and so on. So if I want to book a flight now I can talk to kayak. I can say give me some ideas, tell me when flights are available, tell me how much they are going to cost. I am not using a website. At Deeson we are working with museums and building exhibition experiences instead of building a website for an exhibition you build a message board and it can have like fun character of the one we are working with now is King Charles the second, so there is a King Charles Charles the second and asking about the life and eventually you can book tickets. Devices are changing. Getting smaller, getting bigger, they are moving, static or hidden. You have voice and motion and text entry this is all going to impact the utility of websites. And we can understand the user in ways that we could never before. So, websites are becoming a part of a much bigger picture. Does not mean they are not completely going to go away, but they are going to change in terms of what they are. But let’s deal with the other problem. Okay we do need a website. Convince ourselves the best solution is a website, how are we going to build it. There is a bunch of dedicated SaaS tools including WordPress.com, so go to here and get a website built and it works nicely. I include facebook there, because that is what facebook pages are, they are websites and there is a bunch of businesses right now that say I don’t need a website I have a facebook page, I know who comes there and what they click. I see their face and all of that. Then you dig a little deeper and you have things like wix has their AI based design and you can laugh and say it is not really, it is, you are relinquishing control. It is deciding how to set up the picture the page based on context, on what you are try doing achieve. It is not just wix. You have this services completely AI based. So fire drop says “build your site with Sascha the AI web designer. You have tools like Adobe SENSEI. This is across all their products where they are building artificial intelligence into every single thing they do including web design. So this is just nuts. The fact that we can do this. A designer can sit and say an umbrella here and a person and then the programme is going to go away and say okay those things match exactly what you need. And anyone who does design knows how much time designers spend to just find the right picture. Now with AI you have compressed the time into literally nothing. This is just going to get better because exponential rate of growth. Two years from now it is going to be much much better than what it is now. In the Drupal world you have Acquia doing personalisation Acquia lift and it allows you to try out different images and words and keeps building that user profile. You have ad designed management. A great story CosaBella. They do lingerie and they have outsourced their ad management to an AI and the CEO is crazy about this thing, saying I would never going back to my agency, I love my agency they are beautiful people. I will never go back to humans I will just use Albert and Albert is going to figure out the right phrase and image and manage my advertising and does it so much better than those lovely humans that I really like. Logo design. There is something called Mark maker that spits out designs and you tell it what you like and it just keeps doing it and you can say well this is not as good as a real designer, but exponential growth, it is going to get better. Automated content creation. Association press writes something like three to 4,000 stories now per quarter using bots AI, it feeds in facts and outcomes a story. No journalist actually interacted with that. Initially there was a monitoring phase to make sure it was okay. There are no humans involved the story gets released. It would be awesome if it were robots typing at a keyboard but it is not which is sad. So every aspect of what it takes to build a website is going to impact by AI. And the question becomes okay what now? I think there is a few things, a few answers. We need to start thinking now about all those other interfaces like chat based, motion, virtual reality, augmented reality and how those solve problems. We need to stop selling websites, because websites are not going to be the most interesting thing, you have to sell solutions to problems. Tell your client, which is a good idea anyway. This how I am going to get you more clients. There is going to be new professions, new things we can’t think of right now. So stuff like data scientists and machine learning experts and stuff like bot designers and conversation writer thousand people that write the conversations for bots. I love this one “chief listening officer” some one who tries to interpret what is going on. Assistive ability expert. By definition we don’t know what the new professions are going to penal they get there. And something else that I want to mention is it has to come down to policy. How many know what universal basic income is? It is an idea, a complicated idea summarised into there’s enough abundance and automation that the basics are covered without you doing actually go work. It is not welfare. It is literally saying we don’t need to do these things any more as a society. Because we are changing how society works so we need to change how we run society, the two things can happen separately and then it is going to break and you get a number of things that we are already seeing, this process is happening. So in conclusion, it is going to happen, because there are underlining forces that are pushing things in this direction. We can either pretend it is not going to happen or accept it is going to happen and it is really exciting. I love the idea that my retirement plan is universal basic income. I have all sorts of things I want to do and they don’t need to involve computers any more. That is fantastic. There is opportunity there. Okay.
Thank you (Applause).
DENISE: Thank you so much, that was a lot of information. Did everyone learn something? A lot are you afraid of your machines now? No okay. So we have some time for questions and I am sure you all have many. So raise your hand and somebody with a mic will come to you.
FROM THE FLOOR: Do you think that jobs will vanish and jobs will be created. What worries me certain jobs will be automated, and you will find people not working, do you think there will be a balance between job created and those that will vanish?
RONALD: There is an intense debate about this, some people say you don’t know. I fall on the side of there will not be the same number of jobs created as opposed to jobs that go away. There will be less jobs because we are automating. So by definition less things to do. That is why the change is actually bigger it is political and sociological and so on.
FROM THE FLOOR: Yeah so one of the things that occurs to me at the moment when you are doing old school search with Google et cetera you present went present with a bunch of results and you can choose which shop you buy Nike shoes from, with this bot revolution you are only given one option. That is a lot of power you are handing over to the provider of the results. Do you have any thoughts on the rights and wrongs and what can be done about that?
RONALD: I think we are going to see two things developing. On the one hand we are going to have companies like Nike having their own bots and interact with that and that is the shoes you are going to get. What is going to be more interesting is independent bots that at the back end they are powered by a search engine and they learn who you are and what you like. And you have a conversation literally having a conversation with someone in a shop and they offer different options and so on. I don’t think that bots are going to limit that side of the equation. You are still going to get the single ones and essentially search engines disguised as a bot.
FROM THE FLOOR: Have you got any practical starting points for this kind of work. I have played with something called super script and there are a few tools out there. Have you got any practical favourites starting places or resources you can share?
RONALD: You mean for something specific like bots.
FROM THE FLOOR: Yes the museum bot.
RONALD: What we are playing around with is bot Kit an open source GS base framework and API services like API AI from Google for example that does the natural language processing and figures out the intent of what the user typed in and that you type that back into your bot and give the reply.
FROM THE FLOOR: So you are telling us that we are using bots to inspired by us to create content, writing text and create graphics but what will happen in the future bots will be inspired by other bots so we are dropping the human out of the equation. What is the creativity process in that?
RONALD: I don’t know what is going to happen with that. There is different views right. So it is kind of the view of AI abandoning humans in a way and worrying about itself and things inter acting between themselves. But I think what is more likely to happen and it is very hard to predict right, but is a combination, so there is going to be an augmentate augmentation of what we can do. So not just bots interacting between themselves it is us with bots and it will be hard to tell the line of where the bot finishes. Right now if you take away my Smartphone I can’t go back home. I am going to sleep under a bridge. That is the kind of thing I mean, it becomes codependent on things. The same time my Smartphone is useless without me. Ray Kurzweil calls it a singularity. I think it is 2029, two things together, who knows.
DENISE: Any other questions.
FROM THE FLOOR: For in order for AI to take over more and more stuff does that mean for the consumer to give up privacy more and more.
RONALD: Yes that is where the interesting social questions come out of it. I think the question is much bigger than do you need websites or not. It is happening if you look at elections and how they are run and the level of influence. We need to start asking ourselves what are we actually, where do we want to end up. I will just add I don’t think the answer is oh let’s just stop. We can’t stop. But we need to deal with it. And the problem is that the rate of change might just be a bit too fast for us to adapt. I don’t want to get all dark and so on. But there might be a breaking point. So the more we think about it now the less likely that is.
DENISE: You have a question?
FROM THE FLOOR: Just to follow up on that, talking about influencing and what we just saw happening, is there any controlling instance for AI that we can think about already? Like fake news and all that stuff.
RONALD: There isn’t really. If you look at the news you will hear there are Parliamentary commissions that are calling the managers of Google and facebook and asking questions and saying you need to do something better about this. But there is no legal framework to actually handle it, so there are no penalties to fake news currently.
DENISE: Did you have a question? Did anyone else have a question.
FROM THE FLOOR: I just wondered looking at this in a WordPress context are you aware of anything where bots are being used to interact with WordPress and help people carry out high level functions using this kind of bot technology.
RONALD: I have not seen anything like that. Although I would be surprised if that did not happen soon enough. The person you need to talk to….(Indicates someone)
DENISE: I think we have time for one more question. Does anyone have a question.
FROM THE FLOOR: Don’t you think the problem is what worries me the fact that we are not all unified as humans we have nationality and there are factions in the world, if we use this technology those bot will realise they are more united and have more in common that we do (laughter)
RONALD: That is evolution.
FROM THE FLOOR: They have more in common with each other than we do.
RONALD: It is entirely feasible very intelligent people, Stephen Hawkin and Ellen .. Specifically said this might be the end. These are not silly people. The issues are completely real.
DENISE: So we are going to leave with that positive thought enjoy the rest of WordCamp and we will be back here in 20 minutes.
FROM THE FLOOR: Where can I find your slides.
RONALD: I will put them up on slide deck.
DENISE: He will be tweeting it out and this talk will be on WordPress TV.