When the bot is you
- 16 May, 2016 20:00
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."
How your 'me bot' will work
Here's how I believe "me bots" will usually work. First, you'll probably have a "white list" of contacts who will not get the "me bot" treatment. Spouses, BFFs and bosses may be spared.
"Me bots" will use a variety of tricks to answer interactions. For example, it will recognize when conversations are trivial and will respond in kind. If someone chats, "Hey, Mike. What's up?" Your "me bot" might choose to respond with: "Not much. What's up with you?" (Let's face it: Mike doesn't really need to be involved in the conversation at this point.)
Like Google Now or other applications, your "me bot" will harvest data from various sources. If someone chats: "What are you up to?" it might check your calendar and respond to: "Not much. Just going out to dinner with my parents later."
Likewise, our "me bots" will detect non-trivial conversations. If someone texts you about a promotion or some life-changing event, it will alert you so you can take over the conversation.
Crucially, your "me bot" will provide you with a daily summary of who talked to "you," and also what "you" said. Of course, this all gets pretty dumb when two bots have these conversations with each other.
Regardless, different "me bot" services will work in different ways. But I think most of them will generally handle the trivial for you, then alert you to the substantive conversations where you can step in and really "be yourself."
The bottom line is that the core idea behind Irene Chang's Chat Bot Club is fascinating, profound, inevitable and coming to a messaging app much sooner than you think.