You are using an older browser version. Please use a supported version for the best MSN experience.

A few words on chatbots

TechCrunch TechCrunch 5/05/2016 Natasha Lomas

Lo and behold! A new wave of technology is preparing to crash down upon the unsuspecting consumer. Chatbots are surging towards our conversations at breakneck speed!

No, not the kind of chatbots that have been around for donkeys’ years, adding robotic interjections to irreverent chatrooms when the Internet was a niche playground just for nerds. This new brainy breed of algorithms will, we’re told, be leveraging the power of advanced AI to engage and delight humans with unprecedented conversational smarts. People will be taking to robots like never before! They’ll want to chat with bots like they chat with their buddies, says Mark Zuckerberg. Behold the looming chatbot singularity!

Ok, so much for the hype. Here’s my take: this latest wave of chatbots will not live up to their paradigm shifting billing. And those chatbots that do impress will not really be doing much chatting. Or else will turn out to be Mechanical Turks — ie they will have actual people intelligently pulling the strings behind the scenes.

Witness, for instance, Facebook employing humans as the connective tissue behind its ‘virtual’ assistant M — to plug what are evidently still very sizable gaps in its algorithmic intelligence. Yes, even Mark Zuckerberg, the tech world’s poster boy developer-turned-platform-king himself, isn’t confident enough to let his advanced AI try his users’ patience on its own. Move fast and break things clearly has limits when it comes to sparkling conversation.

Another example: Apple’s Siri squarely failing to dent the universe. The best you can say of Siri is it sometimes serves up some pre-programmed sass. Pre-programmed by humans of course.

Search tech is of course wildly popular, and with good reason: it’s efficient. Or it can be. Any new search technology that works well — whether you can type questions, or show it a picture to find similar shots, or ask for stuff verbally (a la Amazon’s Alexa) — has the power to win friends and influence people. (And when I say ‘works well’, I mean performs your search speedily/effectively and does not waste your time with dumb responses.)

But are chatbots lining up to be an efficient wrapper for search? Early reviews suggest not — especially if companies delude themselves that chatty robots are a golden opportunity to spam spam spam their users like they’ve always dreamed they can can can.

Messaging app Telegram, for example, which has had its chatbot platform live since last June, is already trying to remove friction from the chat process by reducing how much people have to type in order to interact with these robots. So already its pushing for less-chatty-chatbots. It is also trying to incentivise devs to build “useful” chatbots. So it’s not exactly sounding like an emergent AI intelligence is about to spring forth and transform human-machine interactions quite yet. I for one am not holding my breath to better hear the wit of these robot overlords.

Nor does it sound like people are falling over themselves to spend lots of time conversing with robots. Most of the bots on Telegram’s platform don’t have many users yet. And the early interest looks to be sparked by developers rushing in — because they’re being told bots might be the next big thing… That’s hype doing its devilish work, not evidence of consumer demand.

Sure, if you want to do something super simple like, say, find a YouTube URL (so you can share it with another human in an existing chat) then an embedded algorithm that lets you @it in the same messaging thread — to trigger a search and deliver the results back to you — might be useful. Like a keyboard shortcut is useful to the people who bother to figure out which keys they need to press. Although it’s arguably also a bit weird to start chatting with a robot in the middle of talking to an/other human being/s.

But utilitarian bots are not the kind of champion conversationalists we’re being billed to expect. That sort of ‘chatbot’ is really just a search box clad in a little messaging fancy dress.

Hello world… ?

In my view, chatbots are best thought of as an attempt to bolt-on additional utility to their host platform — as a way to keep users on said platform for longer. And to keep the conversation/content flowing (see: chatbots on Slack, for instance — work productivity? Pah! Here’s a new catfact you can discuss with your colleagues instead…).

If a user can book a restaurant in the same messaging thread where they are discussing going out for dinner with their friends then, in theory, there’s no need for the humans to break off their conversation to go figure the organizational stuff out. Although you can’t expect to keep people locked inside your walled garden forever or things start to get pretty dystopic.

But perhaps the greater incentive for platforms when it comes to chatbots is they hold out the promise of being able to generate more conversation than might have otherwise occurred between the humans. To act as conversation starters/primers. And/or conversational maintainers.

This notion meshes well with Facebook’s needs to encourage people to share more personal stuff, for example, given that rates of personal sharing are in apparent decline. While the primer scenario could work on a dating platform, say, with bots delivering potted ice-breakers to get conversation flowing between strangers. Point is, conversation is already stilted and artificial on a dating platform so a little extra robot isn’t going to seem any more awkward.

Really, then, chatbots’ mission statement, if they had one, should be something along the lines of: ‘Keep the humans talking to each other for longer’.

Hate or appreciate

Consider the following two lists…

1) Some technologies humans usually hate:

  • Automated phone systems that make it really hard to talk to actual humans
  • Hard to navigate websites that make it really hard to find the phone number where you can talk to an actual human
  • Voicemail systems that make it really hard to retrieve the messages sent to you by actual humans
  • Any overly complex interface that requires a lot of attention to be expended in order to finally achieve the desired result

2) Some technologies humans generally appreciate:

  • Modern smartphones which make it super simple to stay connected to other humans
  • Basic mobile phones which make it super simple to talk to other humans
    • Feeless cash machines sited in convenient, never too busy locations; card goes in, money gets spat out
    • Faultless, well stocked vending machines; money goes in, treats get spat out

So to which list do the impending wave of chatbots belong?

Well, that depends. Purely human-powered ‘chatbots’ (if I can stretch the term for a moment) which make it super simple for a human to have a question/complaint thoughtfully addressed by another human are certainly going to be appreciated.

And in case it’s not clear, I’m talking about Twitter here, where a short form text-based platform is utilized as a very efficient layer for fielding and managing customer service complaints (yes, far better than websites, email, even phone calls given the typical waits involved — at least it is for now, with (presumably) a lower volume of queries being addressed by lots of enthusiastic, human employees. (Albeit, if Twitter manages to grow its user base it might actually result in less responsive customer service via Twitter… )

Twitter triumphs in customer service (for now) because it’s just about the the fastest way a human can to talk to another human who works in a faceless corporate’s customer service department. In short Twitter humanizes companies.

But of course humans being paid to talk to other humans is not a technology that scales massively like software scales, and so the big push for chatbots…

Searching for something

And then there are basic algorithms doing what algorithms do best (and do better than humans): which is to say sifting large volumes of data and quickly retrieving specific results. Aka search.

These algorithms are of course going to continue to be useful. And if you want to describe a search box that can be addressed inline within a messaging thread as a ‘chatbot’ then sure, such ‘chatbots’ may well thrive — although really it’s just another way to perform a search.

There is also an educational challenge worth noting here, given that people won’t necessarily know how to command search in this more chatty guise. So even here, it’s not necessarily plain sailing for these info-retrieval ‘chatbots’.

There’s another risk too; that the companies putting micro search engines inside chatbot interfaces will be far too tempted to lard them with extraneous nonsense, thereby reducing the effectiveness of the core search in the process. Ergo, back to the spam problem.

But a search alternative is not really how chatbots are being billed. Perhaps it’s just overly effusive marketing at this nascent stage of the latest chatbot wave but we’re being told to expect sassy and useful conversationalists. Powerful, prescient capabilities. Search with a personality. Data retrieval machines that know what we like better than we do. Algorithms with a quasi-human heart. Robots you’ll want to chat to in the same way you talk to your friends.

And mostly that sounds like pure fantasy.

Thing is, humans want meaningful connections with other humans. From machines our chief desire is utility — whether to link us to the data we need or the people we desire. And unless or until a general AI can be developed that is, in mental and emotional terms, indistinguishable from human intelligence those needs are not going to change. So why on earth would we want to be chatted up by artificial algorithms?

For the foreseeable future it seems pretty clear robots will remain plenty robotic. Still largely running along their pre-programmed rails, even if they can learn to change gears a bit more often on their own. And that means we won’t want them trying to chat us up. Rather we’ll want them to listen and obey. To understand our specific request — order me a cab to X, book me a table at Y, bring me Z etc etc — get it done hyper efficiently, and then get the hell out of our way. Meanwhile our tolerance of algorithmic tardiness and stupidity is only going to decrease. Any AI gains will quickly be assimilated and taken for granted. The quality bar keeps rising when your baseline comparison is the elastic intelligence of the human brain.

Efficient, function-focused chatbots might make sticky extra appendages for messaging platforms — to keep users engaged for a little longer. But people won’t be staying because they love talking to the robots, it will be because the algorithms have saved them time to spend talking to other humans. Make no mistake chatbots are not going to win our hearts and minds. The best they can hope for is to please us by performing a task so quietly and quickly we don’t even notice they’re there.

More from TechCrunch

image beaconimage beaconimage beacon