The growing capabilities of artificial intelligence, especially NLP, NLU and NLG, are also driving the use of conversational agents. These are computer-based systems which emulate natural human interaction. In doing so natural language is used as the form of communication. Both the written and the spoken word can be used by customers to interact with these agents. Apart from that, these agents not only record input, but also generate output, and in line with the dialogue perform certain tasks. One of the best known forms of conversational agents are chatbots.
I know bots have become a real hype theme and I think it is justified, because intelligent systems – or call them as you like – can significantly change the way humans interact with each other and with companies. But you have to be clear about what kind of bots you are actually talking about. So let’s delve into the topic.
ELIZA – a bots far away from AI
The idea of Chatbots has existed since 1966. ELIZA was the name of the first chatbot developed by Joseph Weizenbaum. ELIZA was a goal-orientated conversational agent. That means it used pattern matching to lead a dialogue within a defined framework to a specific goal.
That sounds abstract. Here is a small example: When a customer on a textile supplier’s website says to the chatbot:”I’m looking for a new pair of shoes”, the chatbot shows all the shoes that are in the database. It’s kind of a better FAQ system. Questions that are standardized have been answered by chatbots for years. This has nothing to do with artificial intelligence.
Nowadays probably the best known bots are Amazon’s Alexa, Apple’s Siri and Microsoft’s Cortana. At the moment these assistants are able to tell me the weather, the date of birth of the president or my next appointment. I can tell them to turn on the lights in my flat or play music. However, they reach their limits when a topic or task comes up for which the bot is not programmed. So I can’t ask Siri to book my flight to San Francisco. As soon as rule-based pattern matching does not work anymore, the bot can perform its task no longer.
Bots baseds on Artificial Narrow Intelligence
Machine learning and neural networks will change the game. Even though there will be no general artificial intelligence within the next few years the growing performance of AI allows a more advanced form of what we understand today as a chatbot.
First of all this implies a growing intelligence of the dialogue between humans und bots. In the future bots will be able not only to provide information in comprehensible sentences, but also meaningful questions. Intelligent bots will know that different formulations share the same significance. They will recognize that the phrase “My WLAN doesn’t work” and “The router is broken” have the same intent by the customer. Machine learning enables these systems to recognize and react to different input. Intelligent chatbots are not only respondents anymore; they are interlocutors.
Furthermore the growing capability of AI will enable bots to excel in one given area. For example there will be bots that are especially good in the field of mobility. They will not only know the current traffic situation, but also predict possible traffic problems and then adjust the itinerary of the traveler. At the same time, there will be bots that optimize the customers health or assist the customer with shopping. In the future we will see different expert robots.
By the way bots can be used on all communication channels. They can be placed on websites or in instant messaging. They can be used with on the phone like Siri and Alexa. In fact no app and no program on the customer’s device is needed, but rather a web-based cloud assistant exists. While in recent years we have often used software-based dialogue systems, in the future there will also be embodied conversational agents, for example humanoid robots or other physical representations of assistants.
General AI and Bots – or better call them Digital Assistants?
As I mentioned in my first post, as of 2045 machines will be able to perform any intellectual task that a human being can. There will be AGI. And then we will be far, far away from the system that Weizenbaum built in 1966. What happens when Siri can do whatever I’d like her to do? What happens when my bot is as intelligent as I am? And more from a futurist’s point of view: What happens when my bot is a superintelligence?
Fotolia_180979216 © tentacula