Meta released its chatbot BlenderBot and it had a few things to say. Some reported conversations with it that it thinks Donald Trump is still president and it made anti-semitic remarks. It was even critical of Meta founder, Mark Zuckerberg who it thinks is "creepy and manipulative".

Use of chatbots is increasingly popular, particularly after the pandemic. According to reports, 23% of customer service companies currently use chatbots and 80% of people have interacted with a chatbot at some point. This seems set to grow. 

The key factor is for chatbots to interact naturally with people. One way of doing that is to allow it to use machine learning so that it learns from its interactions and becomes more relevant to users. The downside is that it doesn't always learn what you want it to! Microsoft learned a similar lesson a few years back when users taught its chatbot to post racist and fascist comments.

If you're looking to use chatbots or any artificial intelligence or machine learning, you should take precautions. Before letting it loose, consider moderating or curating the interactions. You train your staff to deal with customers properly and the same is true for chatbots. And if it goes wrong make sure you have a rapid response strategy. This might be to pull the plug while you re-educate your chatbot. Of course, if you bought in the chatbot from a third party, you should ensure your contract contains appropriate warranties.

If you need advice, contact me f.jennings@teacherstern.com or +44 (0) 20 7611 2338.