The First Truly Awesome Chatbot Is A Talking T. Rex

National Geographic uses a virtual Tyrannosaur to teach kids about dinosaurs–and succeeds where other chatbots fail.

The First Truly Awesome Chatbot Is A Talking T. Rex
[Image: Flickr user Philippe Put]

As some have declared chatbots to be the “next webpage,” brands have scrambled to develop their own talkative bots, letting you do everything from order a pizza to rewrite your resume. The truth is, though, that a lot of these chatbots are actually quite stupid, and tend to have a hard time understanding natural human language. Sooner or later, users get frustrated bashing their heads up against the wall of a dim-witted bot’s AI.


So how do you design around a chatbot’s walnut-sized brain? If you’re National Geographic Kids UK, you set your chatbot to the task of pretending to be a Tyrannosaurus rex, a Cretaceous-era apex predator that really had a walnut-sized brain (at least comparatively speaking).

She’s called Tina the T. rex, and by making it fun to learn about dinosaurs, she suggests that educatioN–rather than advertising or shopping–might be the real calling of chatbots.

Like a lot of chatbots, Tina the T-Rex runs on Facebook Messenger. Chatting her up is easy. All you do is go to the Facebook page and hit “Message.” When I did so, the toothy reptile messaged me back a few seconds later:

Hello, human! I’m Tina the T. rex. My friends at National Geographic Kids have brought me back from extinction to help you learn about my life. I’m still learning about your world, so I may not have all the answers. To get started, why not ask me about my appearance?

So I did, asking Tina what she looked like. Tina’s response:

I was huge! I was longer than a bus and had a massive jaw to crush my dinner with. I stood on two legs and had two small arms with two claws on each. I think you’d have found me very scary!

I know that some dinosaurs are supposed to have feathers, so that was the next question I asked Tina: “Did you have feathers?” Her reply was pretty informative!

Humans still haven’t found a fossil to crack the answer to that one. Some experts believe I was scaly like a crocodile. Other experts think I had feathers, but they wouldn’t look anything like the ones you see on birds today.

And when I followed up by asking: “Well, if they didn’t look like birds feathers, what did they look like?” Tina said:


I’ve been extinct for 66 million years, so there are a lot of things I don’t know about, and that’s one of them. But keep asking me questions and you can help me learn. Why not ask me how many bones I had?

So far, this is all Chatbot Design 101: Design a likable personality, admit when you don’t know what a user is trying to ask, and try to lead them back to areas where your programming is robust.

But what I thought was really interesting about Tina the T. rex was how forgiving I was of her AI failures. At one point, after asking Tina what she ate (“You could be pretty tasty!” she told me), I went down a scatological path, asking Tina: “How big were your poops?” Tina didn’t understand the questions, talking instead about the size of her brain. So I followed up with the question: “How big was your excrement?” Tina’s reply:

Everyone poos and dinosaurs did too. We know that because experts have found plenty of coprolites all over the world–that’s fossilized poo! How exactly did we go to the loo? Well, we may never have the true answer to that.

Not exactly the answer I was looking for, but I still found the exchange delightful, because of course, what I was doing was something very childish: trying to trick an AI to talk about poop. And while I might have been frustrated if, say, a Domino’s chatbot didn’t understand my order for a pepperoni pizza on the first try, I was quite happy to keep plugging along, rephrasing my question until Tina, the modest T. rex, finally gets it right.

What works here is the fact that when you’re interacting with a brand, you don’t want to have to repeat yourself: You want them to do all the heavy lifting, and meet you 90% of the way, because you’re the customer. Think about how quickly frustrated most people get when they call a customer service line, only to have a representative not immediately understand what they’re asking about. Brand chatbots have the same problem: Our expectation is that they should be bending over backwards to figure out what we want, making their failures to do so all the more alienating.

Flickr user Julian Fong

In education, though, our expectations are different. No one feels like a teacher should do all the work: We all know that learning is a give-and-take process of listening, asking questions, experimenting, and rephrasing questions when we don’t get the answers we’re trying to find. Couple that with a kid’s endless love of asking goofy questions, and their passion for dinosaurs in particular, and you’ll understand why Tina the T. rex might be the first legitimately awesome chatbot. She’s not necessarily more powerful or even better designed than other conversational interfaces, but by meeting kids in the metaphorical classroom to teach them about a subject they love, she succeeds where others fail, despite her limitations.

So maybe chatbots aren’t the future of brands at all. Maybe they’re the future of education. Certainly, if you’ve ever been bludgeoned by a child’s incessant questioning, there’s something about setting them down in front of a chatbot tied into Wikipedia that makes a lot of sense, in a way buying a dress from an AI over Facebook never will.