I call them chatbots, but apparently the proper term is chatterbot. You may have seen this type of software on a web site, purporting to be “Jenny, your helpful assistant.” It’s a trick: That’s not a human, but rather carefully crafted AI software. Is this a concern?

Chatbots (I’m still going to use that term) might be a concern in the future for several reasons, but they’re not anything new. Anyone remember Clippy?

Of course you do! If you suffered through Office 2000, then the annoying, overly-helpful pest constantly bothered you. In Figure 1 you see a screencap from my book Word 2000 For Dummies.

Figure 1. The most loathed chatbot in computer history.

Clippy was officially known as the Office Assistant. I preferred the dog, but several models were available; people just didn’t know that you could switch the avatar, so they were stuck with Clippy. That’s why he has the reputation.

The Office Assistant had its roots in another program, Microsoft Bob. Historically, however, the old ELIZA program was the first attempt at a chatbot. ELIZA would interactively querry you to the point where many users would believe that the software was a real person.

ELIZA was developed in the mid-1960s. The technology today is far more sophisticated.

For example, suppose you need tech support and vie for the chat option. Right now, you type at what you assume is another person, but it’s probably a chatbot. It’s difficult to tell. A chatbot interjects human elements — even typos — to lead you into believing that a person who once had a job hasn’t been replaced by software.

I just went to the Lenovo website and started browsing laptops. (It’s an ordeal at that site, as Lenovo has far too many variations to make the experience pleasant.) After dallying for a few nanoseconds, a prompt appeared asking whether I want to chat with a Lenovo specialist, shown in Figure 2. Is that a real person?

Figure 2. Would this conversation be with a human, a chatbot, a ghost?

When I’ve used the chat feature in the past, I generally ask some human questions, such as “Where are you?” or “What’s the weather like?” The answers make it difficult to determine whether you’re typing at a real person or a chatbot, but one trick you can try is to ask the same question twice. For example, you ask how they person is doing, the type some more questions, then re-type the “How are you doing?” question. If you see the same response, such as, “GREAT!” then you probably are dealing with a chatbot.

How is all this bad? I’ll pontificate upon that topic in Wednesday’s blog post.