Bullets and Roses: To Bot or not to Bot

Pratima H

INDIA: Like pixies jumping out of nowhere and skedaddling out of their ground holes, bots have suddenly started dotting the world hitherto inhabited by devices and ruled by apps. They may be small, may look inconsequential; but make no mistake – they are here with wings, purpose and speed.

If you haven’t heard of words like Slack, WeChat, Kore, Facebook Messenger etc. you would certainly get a whiff somewhere soon. That ‘somewhere’ may not be a tech portal though. It could be your next boardroom meeting, your marketing manager’s next presentation or even the lunch-time gossip table.

Believe it or not, bots have hustled their way in faster than scientists or AI enthusiasts would have imagined. The final nail seems to be in the air, and not for the post-phone-operator world only, but for the post-email, post-device, and yes, even the post-app world.

Who could have thought that apps would find their rivals or successors in something akin to robots!

Yet, bots, specially messaging chatbots are around- promising to help with workplace teamwork, with travel bookings, with cab hailing, with doctor’s diagnosis, restaurant dining, finding sales leads, hiring people, training people, ordering pizza and what not.

It is in theory, a mere chat-based interface that augments apps to help with routine, process-oriented tasks or mundane chores but in practice, the bot is being rooted for the Holy Grail already – i.e. for chatting with consumers.

Speaking into an app is being replaced with speaking ‘with’ a brilliant, self-learning, self-evolving, NLP (Natural Language Processing) - enabled piece of code.

Perhaps one of the reasons that by 2018, as much as 30 percent of our interactions are predicted to be swapped with technology through 'conversations' with smart machines (as Gartner augurs).

What might transpire in 2018 started decades back when bots breathed in our world in form of Internet Relay Chat (IRC)’s automated services that replied in channel or in 1964 when Joseph Weizenbaum wrote ELIZA, a computer program to simulate conversation with a Rogerian psychotherapist; and later in labs working towards new levels of AI like IBM’s Watson, or Google’s DeepMind.

But their existence has morphed into something louder now that companies like Apple, Facebook, Google, Microsoft and some newbies are putting their good weight behind these disruptors.

Even when emails started calendar notifications and suggestions automatically, we were already squeezing an NLP form only – Information Extraction.

Recently the arrival of bots became more palpable when the chat world, messaging systems and app-extensions of many shades started tapping bots.

And then, with explosive promise, followed exponential skepticism. Bots had to meet their chaperone in the practical, commercial arena of technology industry – Doubt!

Bulls in a China Shop?

Even if you have not heard of a solid bot yet, you must have caught the name Tay somewhere in the air some time back. Tay was a Microsoft baby – a bot that was to emulate conversations but one such chat with with a 19-year-old woman over Twitter made it find its way back behind Microsoft’s doors because the unexpected outcomes of racist comments and Twitter abuse popped up due to a user's influence.

This may sound a bit like what poor Siri went through with the anti-abortion programming criticism.

So how does this make a bot look culpable? Isn’t it just a program that is still raw and in progress, learning the human language and context one step at a time, and mainly- from humans only?

The question swings again to AI and Machine Learning – the weak, and true AI zones. Bots, that way, hugely depend on the data or interactions that loop into them and the algorithms that guide them.

So until NLP covers the evolutionary road it has to cover, survival of the dumbest hardly looks like the outcome.

Machine Translation runs a significant track here when we try to understand the strengths and deficiencies of NLP. From the days of the Turing Machines and the Pascaline, machines have gradually shaped into beings that can understand text, translate language and provide answers.

A corollary of it moved forth as NLP started exploring various dimensions of human language and communication- from phonology, morphology, syntax, semantics to pragmatics and Deep Learning.

NLP applications forked into the symbolic and statistical realms – the former working on a set of programmed or learnt rules; while the latter leveraging machine learning algorithms for learning language aspects. The very frontier of NLP has been dealing with the advantages and cons of moving from symbolic to statistical ways and has been trying to strike a balance in the hybrid approach too.

As this is happening, NLP is also trying to accomplish the right conflation of computational linguistics, computer science and AI.

So for a chat-bot or a conversational assistant to actually have a cogent, clear, articulate, outcome-oriented and coherent conversation with a human, it would have to traverse the same path that NLP has been navigating.

There are issues that bots have to innately tackle, like - nuances of language, design flaws, parsing of colloquial language, making sense of complex conversations and interpreting multi-step requests etc.

These are, in turn, issues that actually bot-makers have to handle. Because unless they do it, as some experts wisely observe into the crystal ball already, the big factor of building digital trust between humans and machines would be the real problem.

Can bots be reliable? At the same time, can they be fast, easy, contextual and devoid of friction when a brand or a business employs them to walk the distance towards a customer? Can they be quick and big enough to assimilate and understand data so as not to sound too dumb?

For instance, despite all the hype and excitement around Facebook’s Messenger bots, would users be tolerant if the bot started spamming them with odd-hour or every-hour pings?

Tech users are already frustrated with mail spam and intrusive push notifications so a bot-beep spam would be the last thing they would accommodate.

Even if that part is taken care of, can businesses count on bots offsetting the big loss of human interaction value by bringing in a good axe on costs and a grip on efficiency?

This part needs to be assessed well for the apparent advantages of bots are too tempting to ignore even at this point.

Bots may find favor over apps given the burden on consumers that apps are causing

Apps tapping out?

Bots must have some special muscle that is making them a favourite so soon and so far with so many pioneers.

For every Tay, there is a We-Chat or a XiaoIce (that has, as some guess, clocked over 40 million conversations and) free of any incidents or accidents.

WeChat is known for helping people in Asia, specially in China with everything from shopping for clothes to looking for jobs. Slack, the messaging startup, has been successful too with automated services that answer questions and toss up notifications for colleagues working for a project or on a meeting.

As Julie Ask from Forrester reveals, bots find a ready room in displacing apps because the latter have been putting a huge burden on consumers. “The app ecosystem forces consumers to orchestrate getting the content and services that they need -- sometimes in a single app, most times through a composition of many. And this doesn’t even address individual app quality -- too many of them are simply awful. We're forced through processes translated from online that make no sense on the go or on our mobile phones.” She underscored in a post.

There’s more. They are not only ready-assistant like beings for us but they also make it easy for all the data that apps wasted to be used in a better, deeper way.

Interactions lend a different air of engagement altogether and when peppered with context, they can be a stronger way of understanding and servicing customers than brands would have only fantasized about when wielding apps.

If we were to borrow Alan Turing’s words here - A computer would deserve to be called intelligent if it could deceive a human into believing that it was human.

But, all said and not done, the expectations meter has to be set well when it comes to bots.

They will take time to grow up. Because not growing up is a luxury that only pixies enjoy – not bots.