:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/63708441/lili-cheng.0.1536811325.0.jpg)
Less than a week ago, Microsoft researcher Lili Cheng was in San Francisco at the company’s Build conference, showing off six months of work her team had done on bots, conversational computer systems combining artificial intelligence with vast databases and tons of computing power.
Just a few days earlier, she had been in Redmond, Wash., reeling as Microsoft worked to repair the PR damage done as the company’s chatbot, Tay, turned racist in less than a day.
“Obviously we were caught off guard,” Cheng told Re/code.
Then, just yesterday, she was in New York, being honored by her alma mater, New York University’s Tisch School of the Arts, as part of its 50th anniversary gala. It’s all been a bit surreal, Cheng agreed.
For Cheng, Tay going bad was a big blow. But, she says, the lesson of Tay isn’t that bots are bad; it’s just that more thought needs to be given to the humans who use them.
“I think it’s actually harder than doing the technology,” said Cheng, who has headed up Microsoft’s Future Social Experience lab for several years, after previously doing human interface work on Windows Vista, among other projects. “A huge part is about human conversation and what works, social norms.”
Establishing human trust of robots will be a key hurdle, as will privacy and security issues, as well as new conversational cues, such as knowing when a bot has entered or left a conversation. “I think we need to experiment a whole lot,” she said. “I feel like we really need to get this right.”
It’s not just about more testing with humans, though that is important. Cheng said it is also important for Microsoft to work with others.
She hopes to convene a gathering that will bring together others doing important work in the area, including folks from academia and people from companies like Slack and Facebook. The worst thing would be if developers were forced to write separate code to interface with each bot put out by a tech giant.
“If Microsoft can lead, that would be great,” she said.
Microsoft is hoping that with bots, it can once again be at the forefront of a trend, rather than struggling to catch up, as has been the case with mobile.
The company is open sourcing the technology needed to build bots — the Bot Framework it unveiled at Build.
That said, Microsoft does hope to make money eventually. One key place would be from Cortana, the personal assistant that debuted in Windows Phone but that is now migrating to all manner of devices and onto the Web.
Microsoft’s plan is to let various bots interact with Cortana — at the right time. In one demo, Microsoft brought in a Westin hotel bot, and in another example, showed how one might order a pizza with a Domino’s bot. It also showed another bot approach that was specific to its Skype calling and messaging product.
The real win for the company would be if a lot of search activity shifts from general Web queries — where Microsoft is a distant second to Google — and onto bots, potentially becoming a significant source of digital advertising dollars. There’s also potentially significant business in shifting various bookings from websites and mobile apps into conversations.
That’s still a big if, of course, and even if bots do take off, Microsoft is likely to face stiff competition. On the consumer side, Google and Facebook seem likely rivals, and potentially Amazon and Apple as well, while business rivals could include Slack, Salesforce and IBM.
This article originally appeared on Recode.net.
