Brandon Cross Logo
Brandon Cross Logo Brandon Cross

Is A Chatbot Still Useful If It Fails The Turing Test?

Is A Chatbot Still Useful If It Fails The Turing Test

The famous Turing Test was first proposed by British computer scientist and codebreaker Alan Turing in 1950. Turing suggested than any flowchart-based digital artificial intelligence of sufficient complexity and detail should eventually be able to mimic behaviour and speech in a way that is indistinguishable from that of a human being. If the two responses from a computer and a human were anonymised, a third human should be unable to tell the difference for the machine to pass.

The first chatbot considered to have passed the Turing Test was the Russian-designed chatbot ‘Eugene Goostman’ in 2014[A1] . The point of a chatbot, however, is not to mimic a self-aware human operator, but to provide easy access to a range of relevant information. The Turing Test is, therefore, not directly relevant to the development of genuinely useful, cost saving chatbots.

Standards Of Communication

While the most sophisticated neural versions of chatbots can hold down brief conversations or write an email indistinguishable from that of a normal human being, other chatbots are programmed to stay firmly ‘on the rails’. While this cuts the costs of their production, ‘click-through’ chatbots are often strictly limited in how they can respond, what they can tell the user about, and the subjects they can discuss.

Users may experience frustration or glitches if the chatbot isn’t primed well enough to respond to the subjects that they want to discuss with it. Negative customer feedback for truly ‘robotic’ chatbots often focuses on their lack of warmth, customer and region-specific knowledge, lack of authenticity, and predictable, stock responses that might have been better served by a dedicated FAQ page on the company website.

Nevertheless, is it really that important if your chatbot can’t pass for a human being?

Usually not. Some tasks and industries benefit from the exact opposite – a no-nonsense chatbot that can respond quickly to a ‘set menu’ flowchart of choices and record data rapidly. Sectors such as taxi brokering and fast food takeaways that rely on routine tasks have benefited greatly in terms of speed and productivity from chatbots that explicitly limit the subjects a customer can discuss with them.

What Level Of Natural Language Ability Does My Chatbot Need?

It really depends what you intend to use the chatbot for. When commissioning a chatbot from Brandon Cross, we’ll need to talk through what you’re intending the chatbot to do before we start development.

In a business setting, chatbots that will be used to respond to formal business emails, customer messages, or manage business-to-business communication, answer a vast range of queries, or provide troubleshooting answers, should be as sophisticated and lifelike as is possible on your budget. A decent ‘natural language’ chatbot can ease customer interaction by always providing smooth, helpful responses that can deviate from a fixed script.

The type of customer interaction that you’re aiming for should be your guide here. A Turing-friendly bot will undoubtedly get better responses in emails and messengers aimed at selling the customer a product or service. Dynamic processing allows for a spontaneous conversation to develop and flow with a convincingly human chatbot.

Chatbots & Personal Data

It’s not recommended that you use chatbots to process sensitive information such as tax information, however lifelike you think the software is. Automated responses can come across as callous and insensitive or even prove illegal.

The more complicated and unpredictable the tasks the chatbot will be set, the more human it needs to be. The inverse is also true. In businesses where the customers constantly ask predictable questions or a quick turnaround is needed on a lot of smaller orders, a distraction-free, mechanical chatbot might prove more productive.

In these cases, it can be better to commission a chatbot that pushes information about your products to the front of the queue and presents the fixed options without any parsed language. Do customers really need to have a ten-minute, casual conversation about the weather with a snack vending machine or an equipment catalogue before they make their choice?

Chatbots From Brandon Cross

If you need any type of chatbot for your business, talk to Brandon Cross. Regardless of the level of sophistication that your bot can reach, we’ll create an AI that matches your needs, customer base, and USP outline exactly. Call or email us today.
Image source: Pexels