Chatbots are some of the phenomenal breakthroughs in the world of artificial intelligence (AI). These AI-enabled conversational agents are designed to hold auditory or textual dialogues with humans. Chatbots appear intelligent, but they still have a long way to go in terms of perfectly replicating human-like conversations. Even so, chatbots have remarkable adaptive, learning, and predictive capabilities.

Chatbots require training to function as expected. Users also need thorough training to understand how the interactional software works. In spite of these bottlenecks, the ability of chatbots to turn complex processes into simple dialogues is a notable merit. Though we are yet to have a chatbot that can pass the Turing test, a number of recent AI advances are worth noting. Following just two weeks of training, Elon Musk’s AI bot easily beat professional human gamers at Dota 2, one of the most renowned multiplayer online video games in the world. Google’s recently published paper on its gaming program AlphaGo Zero embodies even stronger AI capabilities. The program was built without reliance on any human gaming data, and turned out to be stronger than its previous versions. The launch of AlphaGo Zero marks a brighter future for AI as it could revolutionise how AI software is built. In the future, data is likely to be less important in AI software development. Instead, AI software developers will focus on finding applications with a high potential for repetitions. This will be a valuable advantage for the development of chatbots given the huge quantities of dialogues chatbots could hold with users.

The Future

The prevailing virtual assistance technologies restrict us to just two options: screen-based chatbots or device-enabled voice assistants. Whereas these assistants work just perfectly at the moment, the future of chatbots goes beyond screens and devices. The future will be about interactions with virtually all manner of surfaces that surround us. We will interact not just on screens and websites, but also on walls, windows, and glasses. Machine learning will without a doubt shape future interfaces.

Importance of Natural Language Processing (NLP) and Natural Language Understanding (NLU)

Advancements in bot technologies have been instrumental to the evolution of NLP and NLU technologies. Not so long ago, NLP-NLU technologies were mostly comprehensible to scholars, but they are now a crucial part of the foundation of AI platforms.

Though numerous NLP and NLU innovations have been made in the last few years, much of the attention has been on innovations launched by dominant players, such as IBM’s Watson Conversation Service, Microsoft’s Language Understanding and Intelligence Service, and Google’s NLP API. These platforms have gained popularity among customers and developers worldwide in large part because most of them have integrated bot applications, such as Google’s Allo and Facebook’s Messenger.

In spite of tremendous breakthroughs in NLP-NLU technologies, it should be noted that most of these technologies are still in their infancy, particularly with respect to replicating human dialogues. Save for a few technologies, most NLP-NLU innovations are only capable of discerning the intent of the discrete words, phrases, or sentences that comprise a conversation. Majority of NLP-NLU technologies are yet to comprehend natural language within the broader context of conversations.

It is important for NLP and NLU stacks to enable not only intent-based analysis, but also context- and flow-based analysis. That is, NLP-NLU technologies ought to be able to determine the context, state, and flow of a dialogue. Context denotes environmental conditions, while state denotes previous data points in a conversation. Flow-based analysis basically encompasses comprehending the flow of a conversation based on its state and context. This ability has the potential to enrich interactions between humans and bots.

Screens to Surface(s)

Obsess over each second, each phrase, each user — context and personalisation is not just key but a necessity in AI

The world of chatbots has undoubtedly come a long way since 1966 when the idea of a chatbot was first conceptualised. Nonetheless, we still have a long way to go. Wide scale adoption of chatbots in business will mostly be shaped by AI breakthroughs. Chatbots can only replicate human-like conversations through much more advanced natural learning capabilities and machine learning algorithms. Key to the achievement of this would be the accumulation of a vast repository of data that can be manipulated continuously. This could make chatbots much better replicators of human-like conversations. Only then can we actualise the imagination of surface interactions.