5 Human Traits to Boost up Chatbots
How bots can defy the fate of widgets
In the aftermath and early demise of Tay , the Microsoft bot gone rogue, and the anti-climax following Facebook’s bot announcements at F8, it’s safe to declare that it’s high time for chatbots to mature from infancy into puberty, and not a moment too soon.
As the Bot Rush overwhelms Silicon Valley and bot miners are flocking into San Francisco, the future of AI and Bot productivity lies in the hands of engineers that are willing to take the hard road. We’re hoping that creative and well thought out research will win, rather than leaning on pundits’ advice suggesting that: “you can take a bot template, tweak it, and launch it- over a weekend.” I dare to guess that’s exactly what the Microsoft folks had in mind (perhaps more than a weekend in their case) when they let Tay roam Twitter unsupervised.
For Bots to become ‘Must Haves’, they’ll have to dance to the tune of a few fundamental human traits and qualities, currently ignored in the hype and rush to easy fortunes.
Why today’s Bots 1.0 are yesterday’s widgets
Remember the days of the browser widgets and gadgets? Yes, those catchy browser extensions. That wasn’t such a long time ago that we all had one for New Times news, one for our favorite baseball team and one for reminding us of the weather in three cities that we’ve never visited. It’s back to the future with Bots 1.0. Take for example Telegram’s Weatherman bot, which will gladly tell you the weather at the city of your choice, exactly the same as a weather widget. It’s not necessarily a bad thing, but it has nothing to do with productivity.
Are you still using those browser widgets or plug-ins? If you are, you might want to download a bucket of messenger bots covering various publications that may be of interest to you. I’m assuming that unless you’re a DeLorean fan, this back to the future mood is not a great omen, given the passing away of the widget industry.
Today’s bots, let’s call them Bots 1.0, are disconnected from any contextual data. They are basically one way silos. They can grab you a flower, a pizza, a cab, a cable. But you’d have to have three of them or four or twenty. They are mission specific and a bit shallow. Wouldn’t you expect one single bot to know all of your favorite publications and articles and have them proactively delivered to your mobile? Can the on-demand industry rely on a slew of single minded bots that you’d have to use to get some productivity going?
I can go on and on. Have you ever tried the team shopping bot on Slack? It’s actually pretty easy to shop for the stuff you need. If you’d like some android charging cables for your team all you need to do is type “I’d like to buy some android charging cables”, and there you have it. No need to browse out. Cool? Well, sort of. It won’t remind you that another member of your team has bought a pair of the same cables two hours ago. Not cool, but I guess we’ll survive the excess cables.
But things can quickly get worse, much worse. Think about asking your favorite mobile health-app equipped with a bot that can provide info regarding your medicine dose. It’s connected to your medical records and might respond correctly by answering the proper dose for your daily pill. However, that bot probably has no clue that you’ve just landed after a 12 hours non-stop flight and that specific dosage would probably be lethal after a long flight. Obviously, not what the doctor ordered.
Bots 2.0 : From ‘nice to have’ to ‘must haves’
If we’re to take bots to the next level of productivity and save bots from the fate of becoming widgets, then we’d have to change their status from ‘nice to have’ to ‘must haves’. The path is risky and will require a more profound level of innovation, more than recently suggested as some bot makers add location sharing.
A ‘Nice to have’ bot is about adding contacts to your Salesforce. For example: “Add Eddy to my list of buying influencers at P&G” and would respond: “Done”. A ‘must have’ bot on the other hand might respond with “May I point out that Eddie is no longer with that company. Eddie is now working for Ralston”. Now that would get things going.
Sometimes reading through claims from bot makers makes you scratch your head. A typical example from a recent post : “If you want to know how to get coffee stains out of a shirt, you could just go chat to a detergent company on Messenger and type ‘coffee,’” adding that you could also start chatting to a spices company and simply type “salmon” to get recipe ideas. Do we really need all that AI fire power in order to get a stain off?
What would Bots 2.0 look like to make them ‘must haves’? I’d like to propose the following human traits and qualities to become the foundations for bot frameworks or at least highly debated discussion points:
1. Social bots. No bot is an Island. Bots should be conversing with fellow bots in a secure way and taking in consideration other people’s events, time lines and habits when recommending to you. For example: you’re ordering a pizza with a Slackbot, which then reminds you to get a veggie pizza because your girlfriend is coming in for lunch (and she’s a vegetarian)
2. History and memory. Bots, like humans, should retain certain amounts of memory so that their actions are also based on your history and previous actions and are not content with the last chat or interaction
3. Temporal. Tracking activities over time ensures your bot engages in a meaningful way and is proactive and reactive around the clock to comply with your habits
4. Training. Chatbots are like babies and ample training data is fundamental to their know how. Google researchers recently used 60M lines of messages for basic training of conversational bots and that’s just for starters.
5. Domain lingo. Bots should be adaptive to the domain ‘language’ they serve, because naturally it’s a different conversation in gaming, dating, banking or insurance
Unfortunately for us, none of the incumbents, with the exception of Microsoft, have outlined a vision beyond the existing Bots 1.0.
AI and bots are fundamental to a new power shift that’s emerging in the information space. But bots without basic traits, as evident from the examples above are like people without a soul, they’re Zombies or Zombots. Using ample data, Tay the deceased bot, would have met the challenge of facing troll tormentors and could have rejected quite easily any attempt to use abusive language.
For bots to make an impact out of Silicon Valley, and make good on their promise, Bots 2.0 must boost user experience by ditching their silos, continuously crunching huge amounts of data for ample training, converse safely with other bots and provide meaningful insights and solutions to consumers across platforms.
The Baymax character is the property of The Walt Disney Company
Great insights from Ben Quazzo
I’m an entrepreneur and techie, passionate about taking cool products into the market, CEO @TomoboxHQ , a startup out of Tel-Aviv and the Bay Area