Eva Dimitrova is a chatbot consultant. She helps companies take the right decisions when considering an automated engagement solution, thinking about what would the whole interaction look like, what should the bot say first, what should be the personality of the bot.
Before talking to Eva my chatbots experience was zero. I only could try to come up with a decent idea somewhere between the two extremes of Spike Jonze’s Her, where this lonely writer develops an unlikely relationship with an operating system and Xian’er, the tiny robot who can chant mantras and answer questions about his faith via a touch-screen display. Little I knew chatbots were an entirely new emerging universe, a tettiroty to be explored, cultivated and imagined. There were even Botfathers, as I discovered later in a post, found mentioned in one of Eva’s written pieces.
I was lucky to have Eva in my Dialogues series and be able to learn (and share) more about chatbots.
Enjoy Eva and the fascinating depth (coupled with a very down to the earth explanations and expectations) with which she approaches this brave new territory of human-machine conversation.
Why are you so fascinated with chatbots?
I was raised with the idea that details matter, that words matter. Coming from Marketing – a creative, yet analytical field – I suddenly saw this chatbot movement going on and thought a brand new channel in Marketing is being created.
4-5 months ago there were still no best practices available, no guidelines, use cases or even expectations about what a chatbot is supposed to do. Apart from returning some messages back. So you had to come up with everything from scratch – explore, test and invent, take part in building the industry.
You have to be creative when you do a chatbot, yet very precise with your wording and how you manage the conversation. Unlike websites and apps, there is no graphical interface.
I find all this very interesting and see it as a way to contribute with my knowledge, as well as to learn more myself. It’s fascinating to be working on something that has never been done before.
Somewhere Between Iyengar yoga, Downhill Biking, Sherlock and Chatbots
Where did this all start? Tell us a bit about yourself and your background.
As a kid I changed a couple of cities before coming to Sofia. I started doing sports in order to find new friends and have been doing that ever since. I went on to compete and win some national and European titles, but at some point decided it was not enough. Colored metal pieces only do as much and I gradually moved from physical activities to working on my knowledge. I suppose watching Sherlock helped as well. I don’t think I’m ever going to quit sports, just don’t have as much time now. There are two activities I find very convenient – running and Iyengar yoga. The first is totally free and quick to do, and the second helps me deal with our modern desk-bound lifestyle. It’s also very special, because of its attention to details.
In terms of education and work, I studied Economics in Sofia University, though I thought about switching to Journalism at some point. Nearly didn’t graduate, but pushed it for good. Later worked in a startup company, where my introduction to this community began. At the end of 2015 I decided it was a good time to try and do something myself and so I quit my job. I was accepted in the Founder Institute program, where I learned in a structured way all about doing a startup, and where was mentored by some big local names in the digital business. I made my startup, built a prototype, but later shut it down. Too many things were not as they should have been and so instead of repeating my mistakes, decided to give myself time and figure things out. And this is when chatbots came in.
Where do chatbots and downhill biking intersect?
They don’t, haha! Yeah, I used to do downhill for 7 years before I quit in 2014. I guess, they intersect in me. Other than that, I don’t see much in common.
What Are Chatbots Made Of?
You are designing the chatbots flow of interaction. What does the process involve?
That’s right, I design the conversation and the experience. I think about what would the whole interaction look like – from start to end. What should the bot say first? Should it use buttons? What’s the best way to navigate the user? How to make the interaction easier and faster. How should the bot handle fall back cases or bad language? What should be the personality of the bot?
Depending on the chabot, there could be many variations. And depending on the platform you develop for, you could have different tools on hand. For instance some messaging apps allow bots in group chats, other offer certain customization or analytics. The interface is also slightly different.
All this needs to be decided and re-adjusted with time. Platforms change all the time, so you have to stay on top.
What are the main technologies behind a chatbot?
There is a lot of tech behind a chatbot and actually most of it exist for the past 30-40 years. Nevertheless, there was one big improvement in recent time that made chatbots so popular. It’s the Natural Language Understanding (NLU), which is under the hat of the Natural Language Processing (NLP), and it refers to the way computers process human language. It’s at the core of Google’s search engine and translation services, as well as in the way speech recognition works. Think of the way you can talk to Apple’s Siri or Amazon’s Alexa.
All this is called Artificial Intelligence (AI) and is achieved with complex systems which you feed with loads of data, like text. This process is called Machine Learning (ML).
What tools do you use?
My biggest tool is my imagination, nothing can substitute that. From there on I try to explore different tools and see what works for what need. Wit.ai for example is good when you want to tell a story, because you can really navigate the whole conversation. Api.ai and IBM’s Watson on the other hand have pre-built scenarios that you can use and don’t have to create yourself. Chatfuel.io is good if you want to build a keyword-based chatbot and has some personalizations.
It all depends on the bot you need to build and on your budget. Language is also important – if you want to build local, you would need an NLP engine for that specific language. It’s something me and my partners of Futurist Labs are working on for the Bulgarian market with BotCraft.
What is the most exciting part of the process of creating that flow, the pre-defined to an extent interaction?
It’s like thinking and telling a story – you can make it really exciting and keep the interest of the user. To make him want to know what’s next. It’s also a story with many branches and people can hop from one to another, so you have to make all of it click well together. It’s more complex than it first sounds.
Where do you gather inspiration from when creating chatbots?
Everywhere, literally. I’d walk on the street or do something and in the background my brain comes up with ideas on how to automate this and that. I even thought about a bot that would remind your flatmates it’s their turn to clean the house. Would be awesome, a lot of nagging saved for good.
The Good, the Bad and the Chatbot
You mentioned Iyengar yoga, can you imagine your instructor as a chatbot?
Interesting, how would I do the asanas and text at the same time? Maybe if the bot had a voice interface would be possible for some styles of yoga. Not for Iyengar, though. How you do the posture is very important in Iyengar yoga and the teacher have to constantly monitor and correct the students. That’s why I like it so much.
This could work for more general instructions, but then again – video is already existing and much superior for this use case.
When preparing for our Dialogue, I stumbled upon the below cited. What is your take on it? Do you see people being reluctant to make chatbots part of their businesses?
The reality is more trivial. This March, Microsoft introduced Tay – an AI-based chat robot – to Twitter. They had to remove it only 16 hours later. It was supposed to become increasingly smarter as it interacted with humans. Instead, it quickly became an evil Hitler-loving, Holocaust-denying, incestual-sex-promoting, ‘Bush did 9/11’-proclaiming chatterbox. Why? Because it worked no better than kitchen paper, absorbing and being shaped by the nasty messages sent to it. Microsoft apologised.
source: Should we be afraid of AI?
Of course, many people are sceptical about it. After all, it’s a developing technology and far from complete. What I’ve learned, though, is you shouldn’t make decisions for the future based on past data. Because past data can only tell you about the past.
As the article says, Tay was designed to learn from her experiences. As is every human being in the matter of fact. So it’s normal when exposed to bad influence, to turn bad as well. Same experiment is running in Asia with a bot named Xiaoice and the results there are completely different. People talk with it all the time and even start to feel emotional toward it. It’s a different culture and different public approach.
My experience so far shows that people in general are curious about new technologies. They want to try it, see how it works and make some dumb-proof tests. With chatbots this is very easy, because you can limit the exposure, monitor the performance and have human in the loop, who can take care where the bot cannot. It’s pretty accessible and safe.
I believe people need to have the right expectations. Chatbots don’t have a human level intelligence yet. But they are very good at automating simple tasks and personalizing the experience for the user.
Speaking of automation, something comes to mind: people hate automated responses. How do you escape that feeling of something talking to you with pre-defined questions and answers? Or do you have to escape it at all?
I imagine you talk about those times when you call your mobile operator and need to wait 2 minutes before you can select your option or talk to a person. Very frustrating indeed. This is exactly where chatbots can be extremely helpful, because you don’t have to wait for an automated voice menu. You can go straight with your problem and get the solution nice and quick. And if you want – you can directly ask for a person to assist you. No waiting time.
It’s not the automation that people hate, but that the interaction is slow and impersonal. Chatbots solve both of those by default, because they already know your name and can learn your preferences. You will be surprised how many services around us are automated and we don’t mind it. Just think about the vending machines, the ATMs and how we want to avoid meeting the grumpy people in the administration.
Do Chatbots Need Content Writers or Vice Versa?
Where do you think, in the process of designing, is the place of content writers?
This is something I discussed in one of my blog posts (Bots are just a facade, here is the real problem) – the new jobs that will appear with this new technology and who would be suitable to do them.
Storytelling, word choice and expression have always been crucial for our communication and I think content writers are a sort of personality designers really. It’s the final touch of a great functionality and one that can make a difference, especially when all you have is a blank screen in front.
One prediction of the future is that bots will communicate between themselves and we, humans, will only interact with a few virtual assistants. True or not, until we get there every chatbot has a voice and should use it well.
If I am to have a chatbot (a machine interviewer) for my Dialogues series, what should I think about in the first place?
I don’t think you have to worry about it, there is way too much personal approach in an interview. A machine can never have the full scope of the picture the way a human does. It cannot catch hints or humor, cannot compare stories or speculate on different ideas.</p>
Machines cannot produce their own thoughts. Well, not in the foreseeable future for sure.
So no automation for you, sorry :)
All You Need to Know About Building a Chatbot
Where is the chatbots industry heading to?
I’d say smartening. With their accessibility and scalability, chatbots are something brand new for us – we’re constantly discovering new ways to use them. And on the other hand the platforms are also constantly improving, adding features and setting up standards. So it’s a big, complex, moving mechanism that slowly speeds up. The earlier you start, the more speed, understanding and data you would gain. It’s a competitive advantage.
If a company wants a chatbot (would they? :P), what are the things they should be considering first?
I’m happy to have more and more people contacting me everyday, asking about chatbots and how to build one. That’s awesome! Still, there are things to consider before jumping into chatbot development:
Who are you building this chatbot for? Who will use it and what for?
Is it better than your current solution? Is it faster, cheaper or a lot more convenient?
Why would the users prefer your chatbot to another? Can you differentiate in a meaningful way?
What are the craziests examples of chatbots you’ve met?
I’ve heard of a chatbot that you can personalize as someone you know and has died. It’s a bit creepy… talking with dead people or so. Other than that, most of the chatbots I’ve tested are standard, I suppose because to be online, they first need to be approved by the messaging platforms.
Who’s Gonna Interview the Interviewer
Teodora: And now, Eva, I would be grateful if you could spare several minutes to ask me a few questions. :)
Eva: That was unexpected! Okay, let me see…>
<h< span=””>2>Eva: What do you find exciting about the chatbots?
The opportunity for a man-machine dialogue in the first place. Also this strange but fascinating idea of creating something through the interaction with it.
Eva: How do you imagine the future?
There’s no future. Only now. :)
How do you relate your work and interest in communication with the bots?
<p>I still don’t see chatbots as something that is directly related to what I do. But I feel this is an territory where web writing will go through yet another metamorphosis.
A Chatbot Genie…
Here’s something curious. When I asked Eva to ask me some questions, she wrote a few, of which I chose three. But then again, when thinking about them, my mind slipped into answering some kind of a forth question.My writer’s imagination said: a chatbot genie kicked in somewhere between the lines. So the question Eva didn’t ask but I read or felt somewhere was:
The Chatbot Genie: Teodora, what would you use a chatbot for?
I would use one as a spare version of my true self – the one that knows and is wise enough to
Go placidly amid the noise and haste,
and remember what peace there may be in silence.
I would feed it meditative thoughts, experiences and words whenever I am in such a mood, and then in tough times, I would use it as a better version of myself to remind me to keep going and stay inspired.
Eva: What do you think is the key to a good chatbot?
A chatbot designer with a heart :)
Last words: Who Knows What Should We Expect…
You might have noticed that I like to speculate on ideas related to the man-to-machine dialogue. Would you too elaborate on that a bit.
Let’s do that! I’ve been watching the Westworld series lately and the storyline is very interesting and relevant to AI and chatbots. (Spoiler alert!)
So there are robots who look and act exactly as humans, but they have pre-defined scripts and because there are hundreds of them, it all looks very smooth. Their personalities are build by hundreds of layers of characteristics. At some point, though, the robots start to improvise and develop unscripted behavior. This happens due to each robot being given different personality with time, but his old memory is not well removed and his characteristics are just partially changed. So the robots start to remember who they were before, start to realise there is more than their immediate reality. This really got me thinking about the future and how maintainable would robots be, if they have human personalities. It’s very complex to first try and build one and then to change it.
Another movie – Elysium – had machines for all public services, like police and medical treatment, and they lacked any humanity. They were brutally honest and direct. So it was the other extreme here.
Who knows what should we expect, perhaps we will first go to few extremes before finding the balance.
Thank you Eva!
Everybody’s got the same tools for chatbots. Just like we have the same computers at home. How we use them is what will make the difference. Execution, creativity, in-depth tool and domain expertise, social and people knowledge. Those are the skills that will matter.
You might also like: