Meetings with Amy - and other bots
In more than 30 years in the BBC I have never had a personal assistant, or anyone to help fix my busy schedule (extraordinary, I know). But this week I took on a PA with a specific brief to organise my meetings. She's called Amy Ingram, and I regret to say her work so far leaves a little to be desired. But then again she is a virtual assistant, or bot.
Amy is the product of two years of work by a New York based artificial intelligence start-up called x.ai. Its co-founder and chief executive Dennis Mortensen tells Tech Tent in an interview for this afternoon's programme that the idea was to prove that artificial intelligence could now handle a single useful task. "We've reached this inflection point where AI can do very specific jobs," he explains. "They can't do everything but if you have one very well-defined job you want done you can give it to the agent and they can do that for you,"
So how did it go? First I handed Amy control of my diary, and gave her a few parameters about where and when I like to have meetings (though the truth is I prefer to avoid them altogether). Then I set about emailing colleagues to suggest appointments. The idea is that Amy knows my calendar and when people come back and suggest different times can work out if that fits with my schedule.
But so far, it isn't working too well. First, she's rather a slow worker. It seems to take hours for her to respond to meeting requests. This morning for example I asked her to shift the time of a meeting with Tech Tent producer Jat Gill back an hour because something had come up. At the time of writing - well past our original meeting time - she still hasn't responded.
But the other problem was my colleagues who seemed to be in a mood to tease my virtual PA. Here's an email conversation between me, Amy and my colleague Leo:
On Apr 13, 2016, Amy Ingram wrote:
Leo has asked that this meeting be at Kennedy Space Center, Cape Canaveral. Does that work for you?
Rory Cellan-Jones wrote:
No it doesn't.
Tell him he can buy me dinner at the Langham Hotel.
Amy Ingram wrote:
Right now I can't pass requests like this to your guests. I think your message would have a stronger impact if you sent it directly to Leo.
Amy was also left confused when my colleague Lakhvir asked about the arrangements for our meeting over coffee: "Will there be cake? I've been let down before."
One male contact also asked, quite inappropriately, for Amy to send a picture of herself so that he would recognise her at the meeting. This time her response seemed eminently reasonable:
Hi, Because I'm an artificial intelligence personal assistant, I'm unable to join you in person. Have a good meeting! Amy
Dennis Mortensen says this may say more about humans than about the capabilities of bots:
"If I had a human PA your first thought would not be how do I make her cry - because then you're just evil." Instead he says we should think: "How do I speak to her so clearly that I end up having a coffee with Dennis? Somehow when we know it's a machine, the worst comes up with us".
He says that we soon tire of trying to manipulate the bots and then can find Amy and her ilk very useful. He is also clear that Amy is not designed to replace a human PA but to provide a service for millions of office workers who will never have that kind of help.
This week Facebook announced that bots were going to be the new way for business to interact with customers via its Messenger service. I've tried a couple of them - the Wall Street Journal bot and a weather assistant called Hi Poncho.
Both invite you to have a text-based conversation to seek out business news or weather forecasts, the idea being this will be more simple and natural than going to a website or app. But in the WSJ's case it just seems to involve a clunkier process than going direct to its site, and the weather bot is plain annoying., wanting a "humorous" chat before handing over a forecast.
I am sure bots - and Amy in particular - have a bright future. But first we and they have to learn how to get along together. That may mean that the humans have to stop teasing, and the bots have to stop pretending they are more than just clever pieces of software.