This article was originally posted on Medium.
The current COVID situation is probably the biggest phenomenon most of us have registered in our lifetime. There are a lot of firsts — first time working the whole month from home, first time celebrating a birthday without family and friends, and the list has almost no end. But the tech advancements we are catapulted towards are here to stay. If the way we interact with machines is changing, it will change at least 10x now.
So, let’s see how our interactions with gadgets have been evolving.
Flashback to my childhood, a single Black and White TV lay in our living room. It has just dial for control. Every time we needed to change the channel, one of us (read — the youngest in the room) had to get up, change the channel.
And let me tell you, getting up every time you need to control a machine does not feel very nice.
Slowly, we graduated to a colored TV with a remote. It had buttons (too many buttons I must say, and to date I have never used all). As I grew up, more and more machines started finding home around me — Basic phones (Read the sturdy Nokia), Refrigerator, VCR. We could make most of these gadgets listen to us while sitting on our couch or bed via the press of a few buttons.
Having the machines obey our command, while I sat on the couch, was nice. But why so many buttons on my remote!
Fast forward a few years, Smartphones, TouchScreen laptops, and LED TVs, and Fridges have become the default in the majority of our homes. We slowly learned the new method of interacting with our fellow machines — Touch.
From more buttons to a few, and then to Touch. It felt way more intuitive for us.
While millennials and post-millennials have found it very easy to adapt to the constantly changing mode of communications, the pre-millennials have at times found it difficult to adjust, but that’s a topic for later.
There has been a new player in the market for some time, the evolution of which will change lives, change the way we interact with machines and with humans.
Yes, I am talking about conversations.
Our current level of usage of conversations to interact with machines is limited to our usage of chatbots, Virtual Assistants in the form of Google Assistant, Siri, Alexa, etc. We take interest in chatting with, issuing commands to and be heard by each of these Products. Our mode of communication can be written, or voice but what talks back to us behind the scene remains the same — an Artificially intelligent machine.
Human-Machine interactions through conversation are the future!
Why are we switching to the conversation? Three major factors are driving this change: - Chatting/talking is how humans are used to interacting. It’s intuitive. There is no need to battle multiple buttons, or logos or icons. All we need to do is talk. - We don’t need to learn to talk. We already know it. - Internet connections have never been better. It is easier than ever before to be on the grid.
While conversations are fast becoming the default mode of communication for many, the advancements in AI will constantly keep raising the bar. The aim is to leave little to no difference between how humans converse with each other and how they interact with machines. Machines can become a better human — better listener, advisor, and actor.
So what are some of the essential characteristics that we look for in someone we would like to talk to? Well, there are many but let’s focus on the top three things that make up a good conversation.
Ability to have an engaging conversation
There are multiple facets involved in having an engaging conversation. Major are: 1. Attentive and active listening 2. Personalized responses 3. Knowledge of multiple topics
If you compare humans vs machines, responding in an appropriate and personalized manner comes naturally to most humans. Knowledge of multiple topics — the jury is still out on that one. But machines are, and always will be better listeners!
Active listening comprises both — listening and understanding what was talked about. Machines, being more docile, are always better listeners. Hence, understanding is the key differentiator. You could listen to a person talking in Spanish, but you might not understand it. The same applies to machines. A machine would not be able to actively listen and understand you unless it is familiar with: - The language you are using - Nuances of the language; formal/colloquial/regional modifications - In the current situation, machines often do struggle if you decide to use a mixture of multiple languages.
Knowledge of multiple topics
Majority of machines that can converse today, have knowledge bases limited to a particular area. Say, a chatbot for a Bank typically answers questions limited to the Banking domain and only questions it has been trained for. Similarly, Alexa, which is built to handle a lot of informal conversations, would be unable to have conversations around, say, Stephen hawking’s Black Hole theory.
The reason is that, like humans, the machine does not know that the topic exists, and does not possess knowledge on the subject. However, unlike humans, the machine has an almost infinite capacity to learn different topics. A lot of active research is being done in this regard.
Machine comprehension and Text Summarization are active research topics you could read more about here.
Empathy makes a big impact on the conversation. You could be saying the same thing, but the way it is phrased or toned makes all the difference in how it is received. One could display empathy by appropriate phrasing of response or/and using the appropriate tone.
Showing empathy, though easy for most humans is not a simple task for machines. It calls for machines to possess common knowledge (which does not come very easy, unless you are human) to know and judge the kind of responses appropriate for each situation.
Let’s take an example.
Human: I walked 5 km in the sun.
A person listening to it could respond in multiple ways like — Why did you walk? Was it that hot? I wonder when it will rain. The central assumption in all responses being — Sun is hot, heat is uncomfortable.
This central assumption that comes so easy to humans as a result of years and years of gaining knowledge, is typically missing in machines right now.
How to imbibe common sense in machines is another challenging research topic in Artificial Intelligence. Read more.
A consistent personality
Every person behaves differently in different situations. After a tough day at work, a person might be irritable. Ice cream might make her extremely delighted. She might talk to a subordinate in a stern manner and to a year old kid in a loving fashion. However, a person retains a consistent personality and you can predict how a person will respond to (in terms of phrasing a response to tonality) in most situations.
The same situation should apply to conversations with machines.
While being flexible in its responses, the personality should maintain a level of uniformity. And considering that each one of us prefers to converse with a different personality, maybe we get a choice. In most of the currently existing conversation channels, personality is something that is hardcoded. Exact responses to every situation are known beforehand and seldom do they change based on users.
Natural Language Generator can help adapt the phrasing and tone of conversations based on its users. Want to read more? Here is something to get you started.
Making machine conversations better and more human is a goal that both Academia and Tech industry is chasing alike. The race is ongoing amongst all the big Tech behemoths and startups specializing in Artificial Intelligence.
The latest achievement has been logged very recently by Facebook via its new open-domain chatbot “Blender” which claims supremacy in terms of knowledge of multitudes of topics and displaying empathy.
See for yourself! Find more details here.
Conversations are the future of Human-Machine interactions. Machines can be dependent on, in the future, to fill the gap where humans have often lagged. The applications of Human-Machine conversations are many, but that is a topic for another time.