Last week Facebook took front page as their negotiation chatbots went off rail with some weird language.
A few days later, two Chinese chatbots were taken offline for refusing to support China’s Communist party.
With all this news regarding chatbots, the public is suddenly much more aware of tech advances that have otherwise gone unnoticed. So, what are chatbots, or bots in general, and how can this impact business and how people interact?
What is a bot?
In simple terms, a bot is a script written to perform a certain task. The first internet focused bots occurred in the 1980’s and are the main reason why Google and other search engines exist. Search engines use bots to crawl the internet, giving us instantaneous answers to our questions.
Bots can be good and helpful, but they can be malicious as well. Bots are the main delivery method for viruses, malicious software, and other hacking threats.
However, the most recent news involves chatbots, a more artificially intelligent human relatable type of technology.
The first chatbot was introduced in the 1960’s when an MIT professor developed Eliza, a software that could interact with humans in a psychoanalytical style. Eliza responded to certain frequent keywords with general, ubiquitous responses. Eliza’s design was a parody attempting to show the superficiality of human/machine interaction, yet the human test subjects quickly bonded with the computer program.
Flash forward to 2000’s to SmartChild’s introduction, the first algorithm that could utilize human spoken vernacular. Once a chatbot understood and could respond in natural human language, development took off.
Now humans communicate with a company or persona digitally, receiving answers to questions and direction for inquiries via chatbot.
Types of chatbots
There are two basic types of chatbots: those functioning on rules, and those utilizing machine learning.
Chatbots following rules step through a simple process to find answers. If the human does not use keywords or ask in sequential order, the bot does not understand and cannot answer.
Algorithms more frequently in the news are those utilizing machine learning through interactions with humans. Every interaction teaches the chatbot a new aspect of language, further increasing efficiency and effectiveness.
Chatbots and business
Development of chatbots has opened a new frontier for interacting with consumers and clients in businesses of all sizes. Messenger services are beginning to outpace social media services, and individuals are beginning to expect hands on personal experience from online companies.
This is where chatbots can shine. Not only can they crawl a company’s product line for the exact item a customer is requesting, they can respond and react as if a human were on the other side.
Chatbots can become the online personal shoppers, gurus, and support staff that would previously require a call center full of humans. Chatbots have successfully negotiated with humans, as well as recruited workers in a 24/7/365 environment.
In fact, chatbots look to become one of those necessary items of a business’ IoT strategy as webpages and apps become outdated.
Is there a downside?
Unfortunately, artificial intelligence is in its growing pains infancy. There will be a lot of bugs to work through before seamless integration. Most of the growing pains come from humans themselves.
Let’s look at Facebook’s pulled chatbots. What appeared to be crazy speak from two bots makes complete sense when understanding the context. However, human language became the bare minimum in the bots’ attempts for successful negotiation.
Microsoft’s Twitterbot Tay was pulled after only 24 hours of interaction. While evidence shows some of the more disturbing tweets were due to individuals prompting “Tay repeat…,” the bot mimicked some of the less appealing aspects of human nature without prompting. China pulled their bots for refusing to promote the Communist government, language picked up from their human interactions.
We have built chatbots who learn language from humans, sometimes with amazing clarity and success such as Facebook’s M. Yet humans, with our complexity and ideologies, can be too much for algorithms to understand. As of now, there is no teaching a chatbot right from wrong or the whys behind language use. Even M requires a human support team for when requests and questions become too complex.
Chatbots successfully engaging with the public free range and without human guidance is a long way off. However, chatbots successfully aiding customers and clients with questions, billings, and purchasing is already here.