Skip to Content
Artificial intelligence

I tested an AI sales assistant that’s trained to push you into spending more

December 3, 2018

Even the persuasive skills of a sales clerk are not beyond the creep of automation. TwentyBN, a Canadian AI startup, is touting a virtual salesperson trained to recognize what people are doing, and interact with them in ways that will drive up sales.

First impressions: I tried out the company’s avatar, called Millie, at NeurIPS, a major AI conference in Montreal. Millie suggested I try on one pair of sunglasses, then another. Her sales pitch was nothing if not direct: “Are you a model?” the software gushed.

Action tracking: For the past couple of years, TwentyBN has been developing a computer vision system that goes beyond simply recognizing objects in images to understand physical actions. “Millie is just like a game character, but with the difference that it can see you and respond,” says Roland Memisevic, CEO of TwentyBN.

Watchful eyes: Though limited, the software shows how computer vision is invading the retail world. Amazon is already testing stores where vision software alleviates the need for checkout. Millie highlights how interactions with customers could be the next part of the retail experience that's automated. At the same event, Alibaba demoed software that takes spoken orders from customers in Chinese KFC restaurants. It might not be long before that system talks you into an extra Coke.

Human factor: Memisevic admits that the technology might perhaps unsettle those who work in the retail world, but he suggests it would mostly augment human workers. “This comes with challenges, obviously,” he says. “But the things this character does are very different. It’s a promotion system right now.”

Hard sell: To be honest, I wasn’t entirely sold. The experience isn’t quite polished enough yet, and Millie’s communication skills are a bit crude and basic. Memisevic sees the rough edges as a plus. “It’s imperfect, but that makes it much more humanlike,” he says.

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

What’s next for generative video

OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.