What does a coffee shop have to do with facial recognition powered by artificial intelligence? In the US, nothing at all. Starbucks stores are everywhere, but computerized identification of individuals’ faces is a huge no-no. Just ask Amazon, which is facing objections from both employees and outsiders as it rolls out its own facial recognition software to US law enforcement.
In China, the story is very different indeed with startup restaurant chain Luckin Coffee recently teaming up with Hong Kong AI company Sensetime to offer customized menu recommendations based on facial recognition. A few points of explanation:
- Luckin is a great story in its own right, even though the business is so new it doesn’t even have an English language Wikipedia entry yet. Since its launch at the end of last year, it has opened over 1,000 stores across 21 Chinese cities.
- It’s hook: customers order exclusively via an online app whether they are in-store or remote. Yes, Starbucks has a similar set up with its own app, but Luckin’s mobile-only approach frees up store associates to produce beverages rather than making some of them stand behind a traditional register. Clever, and a great use of technology to increase asset utilization and lower customer turnaround time.
- No cash changes hands at a Luckin location. They only accept WeChat (a payment system owned by Tencent) or their own wallet, similar to the balance on a Starbucks card.
- Facial recognition adds one more way Luckin can suggest menu items when a customer walks into a location before they even open their smartphone to place an order. There is apparently also an entertainment feature: Sensetime’s algorithm can also suggest what famous person the customer most closely resembles and display an image in-store. Perfect for selfies, we assume…
All this neatly (if cutely) explains why we call Sensetime “the most important company most western investors have never heard of”. Coffee house applications don’t really pay the bills for this AI-powered unicorn; rather, it is facial and object recognition software that makes this company both valuable ($4.5 billion at the last round in May) and disruptive. Every time we write about Sensetime, the portfolio of it core technologies grows. In a nutshell, here’s what it does:
- Facial recognition. Its software offers “millisecond-level positioning of 21,106,40 facial feature key points”. It can identify individuals against a database of known persons, or track an unknown subject in real time with low-resolution video cameras. Sensetime reports a 96% accuracy rate.
- Positive identification of objects such as cars, clothing, personal electronics and other products by manufacturer and model.
- Automatically tag videos by content type.
- Clarify still images that may be clouded by haze, obscured by nighttime, or out of focus.
- Translate 2D images into a 3D model.
- Aid robots with object recognition and manipulation in real time.
- Enable autonomous driving (Honda has a tie up here)
You might be thinking “OK, but plenty of companies do that, including many US Tech heavyweights… What’s the big deal?” The answer goes back to the top of this section and the coffee shop example.
Remember that artificial intelligence (such as what powers all those services listed above) needs data to get better, just as a human child needs teachers and parents to fully develop its mental capacity. Sensetime not only has the Chinese government as a major customer (naturally), but it can pick and choose from a wide array of forward-thinking private companies like Luckin. That’s its real edge, and it is as much societal as technological.
More data equals better AI, everything else equal. And Sensetime’s unique position in AI powered facial/object recognition in China means it is more equal than others…