[devit]

Can anyone explain how a chat-based interface can possibly be generally successful except briefly as a novelty?

Surely an UI with controls specific to the task at hand is much better than having the user guess what the bot responds to and what it can do?

In the restaurant use case, for instance, a QR code + text URL that points to a website where you can advertise the restaurant with whatever graphics and UI you want seems far better than some sort of text interface (which also requires them to be Kik users, know they can scan the code, etc.).

[potatolicious]

It can be, but with a bunch of caveats that I'm not sure people are fully aware of. There's a fair bit of over-enthusiasm about chat interfaces and IMO not a lot of realization of its limitations.

The usefulness of chat interfaces is proportional to how complex the task is and how poorly the user understands the scope and structure of the task.

To use a contrived example, if you have no idea what's left in your fridge, saying "Computer, restock all my essential groceries" is easier than going into the fridge, figuring out what you need, and ordering it via a traditional interface.

Or think of a customer service hotline, where the caller has no idea what options are even available to them, nor do they understand the hellish tree-structure of the touchtone menus. In these instances natural language interaction can remove the need to understand the structure of the task and get them directly to the thing they need. Many customer service lines are already doing this in a simplistic keyworded way, and they can be better.

On the other hand we're seeing a lot of people try to apply chat/natural language interfaces to tasks that are both simple and well understood by their users, and I think these will be doomed to fail. Ultimately voice/text is harder and more annoying for the user than punching buttons, the tradeoff being that they can be valuable when the task is complex/opaque. Most of these applications don't hit a positive tradeoff for the user - and end up being a complicated/annoyingly unstructured mode of interaction nobody wants to use more than once to show off to their friends.

Simple/well understood things like "call an Uber" I think will be pretty DOA if you try to shoehorn a chatbot into the middle - users will prefer the actual UI over it any day of the week.

[geon]

- Computer, restock all my essential groceries.

- Sorry, I didn't quite get that.

- I want you to restock.

- Sorry, I don't know how to "Youtube Rostock". Do you want me to google that?

- Well, it's... Uh...

- I don't know what you mean by "Whale that's a".

[davemel37]

> Simple/well understood things like "call an Uber" I think will be pretty DOA if you try to shoehorn a chatbot into the middle - users will prefer the actual UI over it any day of the week.

I completely disagree when it comes to texting (voice is dif.). I have no interest in waiting for my uber app to load, get me a quote and than order me a ride... I would much rather use a chatbot to look up quick and easy things.

If it's a one off piece of information that I can ask in 1-2 queries/interactions with a bot...I would much rather that over opening an app. Especially, if I can look something up quickly without leaving the current UI i'm using.

I also think you are ignoring all the many use cases that are made possible that were never before possible (i.e. the ability for chatbots to augment humans to achieve access at scale), and the many business models that can shave off 20-30% of their labor costs with well executed chat bots.

I also think it's important to think about the bots calling other bots use case, which might become even more important. Just because you can't envision a human preferring a bot over an app with simple/well understood things...doesn't mean there wont be many use cases where a current bot you are using talks to other bots on your behalf, so the UI choice is never made by you. Finally, and most importantly, if chat bots start replacing parts of discovery, search and research, it will become crucially important to make sure your brands are found and used by bots and people over your competitors brands and bots.

Think of doing SEO in 2002. If you got it right, you banked tens of millions with minimal effort.

I can't predict the future, but viewing chat bots as a UI alternative limits its actual scope.

[rmason]

In Waterloo there's a restaurant where you can text your order. Personally I'd rather text bring me a 38 with a diet coke and hold the mustard. The number of times a waitress or drive thru person gets the order wrong is much higher than simply texting it.

http://venturebeat.com/2016/03/01/why-bots-not-a-i-are-the-f...

[kevando]

Kik CEO actually has a pretty good narrative on this

https://medium.com/@tedlivingston/the-future-of-chat-isn-t-a...

[munificent]

> Surely an UI with controls specific to the task at hand is much better than having the user guess what the bot responds to and what it can do?

A well-designed and well-implemented UI is better, yes. But I think where this could be useful is for the many customers who don't have the cash to make a decent app. A stripped down chat interaction could end up more usable than what they'd build from scratch.

[lfowles]

The value is in making a legitimate platform for the lowly CLI app then. I like it :)

[Jordrok]

I honestly have no idea. Chat bot mania sure seems to be in full swing though. Call me cynical, but all I can see is a lot of tech and development hours desperately in search of a problem to solve.

[WorldMaker]

Ultimately as with so many trends the underlying thing here seems to be "Go where the users are". Users are using lots of chats and chat apps, so now we all need bots to meet them there.

That's why this fascination with bots and conversational UIs seems particularly cyclical, as various chat rooms/chat apps come and go in popularity. Just the other day, I was pointing a colleague to The Jack Principles, a book/presentation/best practices compendium whose lineage traces back to the 90s (and I think now is a great time for Jellyvision to sell us all a new edition in bookstores).

The reason this is particularly exciting this time coming back around in the zeitgeist is that we have better tools than ever (NLP has come a long way, machine learning and deep learning are doing exciting things, the "platform" level concierge/assistants of Cortana, Siri, Alexa). Maybe this time it might be more than a passing novelty.

[Aleman360]

It's a command line where you don't have to be precise.

[ewzimm]

Although bots have been around for a long time, most of them have required precise commands. You're right that imprecision is why it can be successful, and the key to making this work is having manager interfaces like Siri, Cortana, Alexa, or Hound which can add layers of abstraction over simpler single-purpose bots. As we standardize communication protocols between bots, it will be easy to send a header describing abilities and responses so that the user interface can both parse the capabilities for the user and send feedback about how to be more responsive to the bot and its programmers. In the end, if this works, we should have a massive amount of machine learning taking place as bots improve each other.

[hammock]

>Surely an UI with controls specific to the task at hand is much better than having the user guess what the bot responds to and what it can do?

The challenge of syntax discoverability can be mitigated in a text interface with smart autocomplete to guide people towards what they can say. It becomes harder with voice. Also one messaging app, Telegram, allows chatbots to present customized keyboards that change based on the context of the conversation, e.g. thumbs up/down or star rating system.

[dpflan]

Perhaps it is the illusion of communication with a human (if the bots can handle variable expressions for the same concept - e.g. people can say the same thing many different ways) that makes this appealing. I don't have to click, which would mean I construct a concept in mind and I then convert that to the correct expression provided by the service. Integration of simple bot support is a first step as AI/ML + NLP evolves to support greater complexity.

[hkmurakami]

What about CLI though?

[fooey]

At least they're internet famous among developers for being the guys who blew up NPM, so maybe now everyone will be excited to start working with them?