#conversational agents

Bot platform Sequel allows its users to build bots of themselves.

A bot platform called Sequel today launched a way for you to build bots with personas, including ones that mimic your own personality.

The Sequel software, which had been in beta testing for months, joins several competitors in the market that have launched over the past year. However, Sequel’s bots are designed to have personality, unlike most bots, which are built to do discreet jobs, like helping you order pizza or order flowers ...

Google is also expected to get involved later this year, by offering a bot platform that is competitive to Sequel’s. Google Director of engineering Ray Kurzweil, speaking recently at a Singularity Conference, said Google is building a way for anyone to create their own unique chatbot by feeding it a large sample of your writing, for example by letting it ingest your blog. This would allow the bot to adopt your “style, personality, and ideas.” (Kurzweil’s comments, recorded in a video posted by Verge last week, may not have been ready for public release, because the video was suddenly made private last night.) He said elements of the bot project would be ready this year, but didn’t specify details.

Meantime, Sequel has beat any such Google project to the punch. For now, it caters mainly to brands and creators in the entertainment and media industries, helping them build bots that converse with their followers.

A creator without a technical background typically needs about 15-20 minutes to learn the basics of creating a bot conversational flow with Sequel’s software, according to Omar Siddiqui, CEO of Kiwi, the company that built Sequel. After the creator’s initial onboarding, its then up to them to write up the conversational creative they want their bot to have. The software offers drag-and-drop tools that let creators use text, photos, videos, audio, GIFs, emojis, quizzes and polls.

Sequel designs the flow so that the bot can respond to any question, and to have answers (ideally witty) in case the bot doesn’t understand a particular question.

Sequel uses natural language processing, so that the creator’s bot understands as many questions as possible posed by users. It also lets the creator integrate links to web services through APIs. So, for example, if you are a chef, and want to create bot to share recipes with followers, you can integrate a recipe API, but that does take some technical acumen. Sequel is working to automate these processes for non-technical users.

A Facebook bot of Werner Herzog answers to questions with recommended reading.

  "Why aren't you reading?"

This is the first thing WernerBot -- a whimsical Facebook message bot that claims to be the "best and only way to chat with Werner Herzog over the Internet" -- will say to you when first engage it with a simple greeting, "Hello Werner."

WernerBot is not affiliated with the famed director, but it responds in a way that one can't help hearing Herzog's excited voice.

Write a response about any book you're reading (or about anything in general) and WernerBot will answer with, "Stop looking at this Face Book and read a real book."

Of course, you could be reading Shakespeare or the latest James Patterson mystery novel and it won't matter to WernerBot because it already has a couple of book suggestions at the ready.

"You should read "The Catalogue of Dwarfs" from the Icelandic Poetic Edda, an extraordinary collection of Old Norse poems compiled in the fourteenth century and probably handed down, in the oral tradition, a thousand years earlier," WernerBot answers when I try to impress it by talking at great length about how much I loved rereading Neil Gaiman's book "American Gods."

Thanks Werner, but as big of a Thor fan that I am, Old Norse poems aren't really my thing.

"Sitting at home on your own, in front of a screen, is a very different experience from being in the communal spaces of the world, those centers of collective dreaming," WernerBot responded.

Bots are to evolve with Natural Language Generation (NLG) technology.

  The playing field for linguistic software isn’t singular. Microsoft isn’t the only athlete coming out of the block, only the most well-known in the public’s eye.

Arria NLG to name one, offers a sophisticated software tool that combines cutting-edge techniques in data analytics and computational linguistics. These capabilities allow the engine to convert large and diverse datasets into meaningful natural language narratives, helping solve the problems of dealing with Big Data.

In short, Arria has made the process of converting data into narrative easy. We have built an AI that does what your human, natural intelligence does.

The dawn of natural language

Natural Language Generation is all about taking data and turning it into written or spoken language. This is indisputably the best way to provide actionable analytics. It’s how we as a species have communicated information and calls-to-action for tens of thousands of years. It’s a finely honed device that evolves as we evolve, and it marks us out from other species. It’s the best means by which we answer the requests “Tell me what I need to know” and “Tell me what I should do”.

Matt Gould, CSO & Co-Founder of Arria NLG feels the resulting problem with Tay could have been prevented with different technology.

“Right now we’re still playing question and answer with chatbots. But there’s still another level until the Internet can communicate with you naturally, until it can come back with a more general inference about the intent behind your question, and NLG is the solution that embodies the required expertise. To understand, conversely, is to be able to infer something from information received– analyzing and extracting insights from diverse data sets.”

Giving the internet a voice

Gould sees Natural Language Progression as a logical step in the progression of the bot marketplace. Comments Gould,

“The issue with today’s bots is their very limited areas of knowledge – these bots must fall back on search engines for anything outside of their narrow capabilities. The Internet is incredibly smart, but bots today are technologically equivalent to a two year-old just learning to speak for the first time. But bots can evolve with Natural Language Generation (NLG) technology.”

Gould feels that current NLG technologies can enable someone to have a free-flowing conversation with a bot, but none of the big tech companies have picked up on the intricacies involved. Not yet anyway.

NLG, explains Gould, “Combines state-of-the-art data science and analytics with cutting-edge computational linguistics techniques, delivering information that is hidden within data and conveying it through natural language.”

In other words, NLG gives the internet a voice without making it sound like a robot.

It provides an automated solution that can take diverse data sources, aggregate their information content, and deliver that content in a way that is immediately accessible. As a result of advancements in NLG, bots in the marketplace will become more sophisticated, useful, and friendly.

Microsoft is building a virtual assistant bot for Bing.

  Conversational bots, which can understand complex voice commands and complete tasks for you, are quickly becoming the rage. Earlier this month, the inventors of Siri showed off their creation Viv, which ordered flowers and booked a hotel in under a minute. Google Assistant, revealed this week, promises similar capabilities.

Microsoft has now let slip that it’s also working on a bot to help get things done, reports ZDNet off of a job posting from the company that’s now been removed.

The Bing Concierge Bot will be able to respond to users over platforms like Skype, Messenger, SMS, WhatsApp and Telegram. The job listing for an engineer to join the team behind this project describes the bot as follows:

The agent does what a human assistant would do: it runs errands on behalf of the user, by automatically completing tasks for the user. The users talk to the agent in natural language, and the agent responds in natural language to collect all the information; once ready, it automatically performs the task for the user by connecting to service providers.

For example, the user might ask ‘make me a reservation at an Italian place tonight’, and the agent will respond with ‘for how many people?’; after several such back-and-forth turns it will confirm and book the restaurant that the user picked.

The bot revolution is about to get weird. Get ready for your own personal 'me bot'

  Unless you're new to the planet, you know that soon you'll be chatting away with artificially intelligent bots.

But the bot revolution will also usher in something strange: It will give us a bot to talk for us, as us. I call it a "me bot."

A developer named Irene Chang (a.k.a. Irene Lion) created a "me bot" at the recent TechCrunch Disrupt hackathon in New York. The software is called the Chat Bot Club. It learns how you chat, then interacts with your friends as if it were you.

Chang used IBM's Watson chatbot software to build the bot to work on the Cisco Spark platform. The software "learns" your style and creates a database of phrases and responses that you frequently use. It then participates in group chats as you.

Chang is porting her "me bot" to Facebook Messenger, WhatsApp, WeChat, Kik and other platforms.

While Chang's Chat Bot Club is the first attempt I'm aware of to create a "me bot" for messaging, the general idea isn't new.

A beta social network called ETER9 does something similar for social networking interaction. (Bear with me here, because this is going to get weird.)

ETER9 lets you do Facebook-like social networking in a part of the site called the "Bridge." All the while, the site captures what you say and do in something called the "Cortex."

When you're not logged in, a virtual, A.I. version of you called the "Counterpart" continues to do social networking on your behalf based on the data in your "Cortex" -- commenting, liking, chatting. Disturbingly, your "Counterpart" continues to interact as you even if you stop using the service, even after death.

That post-mortem aspect of ETER9 reminds me of a category of services that claim to enable a virtual you to continue interacting on social networks after death, including Eternime and Liveson. (Liveson's slogan: "When your heart stops beating, you'll keep tweeting.")

These "like-after-death" or "poke-mortem" services strike me as parlor tricks -- specifically, funeral parlor tricks. After all, if they don't work well, it's not as if the customer can complain.

Another project in this general space is far-reaching and maybe a little creepy. Lifenaut, which is a service offered by the Terasem Movement Foundation, wants to collect information about you in order to construct a virtual you that can live after death.

Lifenaut gets this data from a personality test and also by harvesting social networking activity. The result is what they call a "Mindfile." The service is free for now, but may cost something when it's time to build the virtual you.

Lifenaut is agnostic about the application of this "Mindfile." They suggest that "perhaps in the next 20 or 30 years technology will be developed to upload these files, together with futuristic software into a body of some sort -- perhaps cellular, perhaps holographic, perhaps robotic."

It's easy to dismiss all this as tinfoil-hat nonsense. But the woman behind it, Martine Rothblatt, is the founder of Sirius Radio and the highest paid female and transgender CEO in the U.S. (at the biotech firm United Therapeutics).

Hanson Robotics even built a prototype robot for Terasem Movement of Rothblatt's wife, whose name is Bina. (Martine Rothblatt is president of the Terasem Movement Foundation and Bina is vice president.)

Here's a video of Bina talking to the Bina48 robot. If anything, the video reveals that our robot clones are not ready for prime time.

Long before our robot clones are ready, we're going to be using "me bots."

Why you'll join the Chat Bot Club

Some media outlets are reporting on Chang's Chat Bot Club as a gimmick for sociopaths. But I believe you'll use something like it, and so will I. Bots that interact as us and on our behalf are coming, and we will love them.

How do I know that? Because we're already trying to automate our conversations. What "me bots" represent is simply the automation of communication in the age of bots and A.I.

Organizations big and small, as well as some individuals, have used email auto-reply and mail-merge to blindly send messages to people as if they were individually crafted.

With auto-reply (as when you're on vacation), you don't know who you're sending to or what email you're auto-responding to, but the recipients usually know it's an automated message when they get it. With mail merge, we know who we're sending to, but the recipients often don't know the message is automated. Either way, we're automating communication in a desirable and socially acceptable way.

I've also mentioned in this space an A.I. chatbot called Amy, x.ai's email-based virtual assistant. Amy interacts via email and arranges meetings. The interaction is so good and natural that many or most of the people interacting with Amy don't know she's a bot. And they don't care. Amy's existence as a bot matters only when she gets flowers or asked out on dates because of her winning personality and professional competence.

Google recently rolled out a feature called Smart Reply for its Inbox email service. Smart Reply uses recurrent neural networks to produce three possible replies to most email messages. You can ignore them, or use one by simply clicking on the button. The fact that Smart Reply is a popular feature shows how appealing the idea of automated communication based on artificial intelligence is.

"Me bots" will also fill two roles with existing technology. Voice mail serves the purpose of enabling people to communicate with us when we're not available. It uses a recorded version of our voice to enable a psychological connection for the caller in real time, but a delayed engagement with the caller for us. Voicemail makes conversations "live" for the caller but asynchronous for us. "Me bots" will do this for us on messaging platforms.

Social networks enable people to maintain a much larger number of relationships than they otherwise could. If you maintained regular contact of some kind before Facebook with 50 people, you can do so now with hundreds. "Me bots" will do the same thing on chat-based platforms. (In fact, Chang created the Chat Bot Club because she couldn't keep up with all the group conversations going on). As messaging continues to grow and replace social networking, people will be overwhelmed and will happily share the load with a "me bot."

Amazon’s digital assistant Alexa swears, because we do.

  A few months ago, a pair of grown (but not necessarily grown-up) men discovered that Amazon’s digital assistant Alexa swears. You can use her Simon Says command to make her say “crap.” They also found that if you try to make her say “shit,” she’ll do it but bleep herself out.

This is remarkable when you think about it. People at Amazon took time to discuss this. They came up with possible solutions. They even made a decision. At one point, a highly-paid professional probably stood up and said, “The C-word and the F-word are obviously off the table, and ‘crap’ is ok, but how do we feel about ‘bitch?’ And where are we on the question of what Alexa says if someone asks her for a blowjob?”

I’m being slightly crude here because we need to understand the level to which you have to take your thinking with something like this. If you try to make intelligent digital technology, a major mark of its intelligence is how well it reflects and anticipates our baser instincts. If you put a digital assistant in front of people, sooner or later, they’re going to ask it to perform sex acts. It’s just the way it is, and you have to be ready for that.

And sex is far from the only thing you need to worry about. Alexa has to be ready for a wide range of inquiries that have almost nothing to do with her stated purpose. For example, if you channel 2001: A Space Odyssey and ask her to open the pod bay doors, she’ll tell you she can’t because you’re not in space, and she’s not Hal. She has hundreds of such Easter eggs ...

You might think that Alexa’s team was wonderfully clever in anticipating all these questions. In point of fact, they were nothing of the sort. If you enter Alexa’s Easter eggs into Siri, you’ll find that she too has her answers to almost all of them. Believe it or not, Alexa, Hound, Siri, and Google Now all have answers to the Monty Python question “What is the airspeed of an unladen swallow?”

In other words, these are not really Easter eggs. They are a common and necessary feature of these products. They are there because people will ask them, and if your assistant is truly intelligent, she should have answers to them.

I’ve worked on projects like this and know they have two features that almost no one would guess. First, they’re very easy to do. Second, they tend to attract a very high level attention from corporate management.

Let’s start with the easy part. Contrary to what you might think, you don’t have to consult reams of data or spend weeks in a conference to come up with Easter egg questions. Lists of them are available on Reddit and have been floating around since the early search engine days. The writing is done typically by a team, whose work goes fast, because the jokes don’t even have to be funny. They win by showing up.

Now for the executive-level review. These things always go up the ladder. In Amazon’s case, Jeff Bezos himself probably looked at them. You might think he had better things to do, but that’s not necessarily the case. This is important stuff, perhaps even more so than things like Alexa’s ability to order an Uber car.

The reason is that when we build something like Alexa, we do not merely need to create a bot that can hail a cab. We also have to reflect our cultural baggage, questions about the nature of the universe, and our nervousness about the language of sex. All of what is human must be put into the machine ... Alexa swears because we do. It’s that simple.