In author Frederik Pohl's epic Hugo- and Nebula-award winning Heechee saga, protagonist Robinette Broadhead runs his multi-planetary industrial empire — and his equally complicated life — with the help of three AI-powered chatbots.
A Siri-like assistant manages day-to-day business and personal activities; “Einstein” collaborates with Robin to invent innovative solutions to complex business and personal problems, while managing his family’s health; and “Sigfrid von Shrink” helps Robin come to terms with guilt over a traumatic event from his distant past that ends up playing a role in the story’s conclusion (time gets weird, of course, when you cross a black hole’s event horizon).
These stories shaped my expectation that, in my lifetime, AIs would collaborate with humans to improve our lives. They would chat with us in natural language, backed by brains the size of a planet (to paraphrase Douglas Adams).
And feeding that brain would be all the information in the universe (including continuous real-time news flows), plus environmental sensors providing nonstop telemetry on all aspects of our physiological condition. I dreamed of the day I’d chat with my own Einstein. Like "M*A*S*H’"s Radar O’Reilly, he’d anticipate my every need and serve it up at the precise moment I realize I need it.
I’m still dreaming. The current generation of AI-powered chatbots exciting brand marketers today are not blowing my mind.
While Microsoft justifiably blames Twitter trolls for talking its Tay chatbot into becoming a racist, sexist holocaust denier, Sigfrid von Shrink would have run rings around them. But Tay was bleeding-edge for 2016 — maybe the first-ever full-blown attempt to loose an unconstrained AI “general intelligence” on the English-speaking world (for an explanation of the different types of AIs, check out Sarah Fay’s post).
Microsoft previously tested Tay’s technology in China (where the government bans many “offensive” words on the Internet) under the name XiaoIce. It racked up 40 million sufficiently polite users within 72 hours.
By comparison to Tay, most chatbots available for marketers’ use right now are plain stupid. That, of course, makes them safer to use; they’re unlikely to end up slandering your customers and damaging your brand.
The problem is, what they do is mundane. It’s a big important step in the right direction that Twitter, Facebook, Google and a host of other companies and start-ups make AI-powered chatbots available for marketers to engage audiences.
But so far, they’re not really doing anything that you couldn’t do — probably better — with old-fashioned software, a database and some business rules. For example, Domino’s pizza-ordering AI chatbot on Facebook Messenger seems to only understand one word, “pizza,” by which it assumes you want a regular pie. No special orders. Most anything else you ask is likely to return “Sorry I don’t understand.”
These uses of AI- and machine-learning-powered chatbots are a far cry from “the operating system for society” that KIK founder Ted Livingston envisions chatbots will become.
Meanwhile, I met an IBMer earlier this week at a conference put on by the Business Marketing Association’s New York City chapter. He pointed out that because Watson has one of those really big brains (not quite planet-size yet, though), it actually “understands” that 60 degrees in Florida is very different from 60 degrees in Minnesota. In Florida, they’re wearing their heavy coats or heating up the Campbell’s soup (another user of Watson-powered “cognitive ads”), and in Minnesota they’re walking around in T-shirts and shorts.
So at the moment, chatbots powered by IBM Watson do seem to show the most promise for the future. I played with the one at North Face, which asks some basic questions and recommends a jacket for you, your planned activity and your environment. The problem, still, is that on the surface it seems like it’s not a lot better than a rules-based interactive database tool.
But play with it a bit; try a few different locations, activities and other preferences. You realize that maybe there’s a real brain back there somewhere doing the work, and maybe it’s using more rules than would be cost-effective to program. Plus, it has access to a lot of data, like The Weather Company data. All that data adds potential.
Now, think about scaling that. This is where the magic comes in with AI and machine learning. As chatbots learn and grow, they will get smarter and do more things. Ultimately, brand marketers will be able to engage audiences in deep, intelligent ways, as if they had millions of human experts on hand to talk one-on-one with every individual customer about just what’s most important to him or her.
But to get from here to there, you have to start with these relatively lobotomized constructs and their tiny little brains. You have to learn by doing — and so do your chatbots. The great chatbots I dream about will stand on the shoulders of pioneers like Microsoft, Google, Facebook, Twitter, KIK, Slack, North Face, Domino’s, Campbell’s, Unilever, Toyota and so many others.
Today’s chatbots are the absolutely necessary first baby steps on the road toward Einstein and Sigfrid von Shrink.