chatbots, personal assistants and the future of artificial intelligence
TRANSCRIPT
WHERE WE ARE NOW, WHERE WE ARE HEADING AND HOW TO GET THERE SAFELY
CHATBOTS AND THE PATH TO ARTIFICIAL INTELLIGENCE
Yisela Alvarez Trentini | MRM McCann, August 2016
WHAT IS A BOT
A bot is any application that you communicate with, via speech or text,
in order to execute commands.
2
INTRO The bot revolution
can be compared to the mobile revolution
Although it was a progression, bots are possible now due to massive amounts of data and hardware
It’s more of a tech-driven vision, not a response to concrete user demandsYisela Alvarez Trentini | MRM McCann, August 2016 3
Facebook‘s M
Microsoft Win + Azure
THREE TRENDS DRIVING SMART BOTS
• Messaging-as-OS: Messaging as a new platform
• The app problem: People are reluctant to install apps, or apps are becoming redundant
• The “conversational interface”: A new model for interacting with online services
Yisela Alvarez Trentini | MRM McCann, August 2016 4
Yisela Alvarez Trentini | MRM McCann, August 2016 5
1974: Command line 2016: A Slack channel
“Command lines were notoriously intimidating and difficult to get the hang of. Slack is the exact opposite—it’s
charming, fun, and easy to understand—yet it runs off the same principle.”
Archana Madhavan
BACK TO A MINIMAL INTERFACE
The future: ?
1. The Evolution of BotsThree cases
6
ELIZA 1966
HOWDY 2015
SAMANTHA-
STATELESS SEMI-STATEFUL STATEFUL + INTELLIGENT
Shallow tracking of conversations
Can read, comprehend and reason
No memory or storage
Yisela Alvarez Trentini | MRM McCann, August 2016
1. Stateless Bots2. Semi-Stateful Bots3. Stateful Bots
8
STATELESS BOTS
• Each request that a program processes disappears from the server’s memory.
• Bots are stateless by default. Apps receive requests, bts receive messages. If the web server were to keep track of the requests it had processed, it would soon collapse under its own weight.
• Each message is considered a new interaction. Yisela Alvarez Trentini | MRM McCann, August 2016 9
Created by Joseph Weizenbaum (MIT) in 1966.
Acts like a non-directional psychotherapist in an initial phsyachriatic interview. Open, introspective questions.http://www.masswerk.at/elizabot/
Yisela Alvarez Trentini | MRM McCann, August 2016 10
THE FIRST BOT: ELIZA
Simple pattern matching techniques (parsing and substitution of key words)
Yisela Alvarez Trentini | MRM McCann, August 2016 11
ELIZA’S INNER WORKINGS
OTHER STATELESS BOTS
Yisela Alvarez Trentini | MRM McCann, August 2016 12
A bot that twits images from museum collections
A bot that replaces the word „boy“ with „bot“A bot that posts random art assignments
1 2
3
1. Stateless Bots2. Semi-Stateful Bots3. Stateful Bots
13
Made by XOXCO in 2016 and integrated with Slack.
Easy to install, works out of the box. A „digital coworker“ to automate:
Asynchronus communication Plan meetings Collect lunch orders
Highly effective for a limited scope of tasks.
Yisela Alvarez Trentini | MRM McCann, August 2016 14
Howdyhttps://howdy.ai/
WHAT SEMI-STATEFUL BOTS ARE GOOD AT TODAY
In the workplace: Automating small routine tasks, run surveys, act as a bridge between gaps
Healthcare: Follow-ups and reminders (Sense.ly), answer questions (Your.MD), health coaching
Onboarding: Guiding through materials Used for evil: Social network spamming, profile
clonning
Yisela Alvarez Trentini | MRM McCann, August 2016 15
Yisela Alvarez Trentini | MRM McCann, August 2016 16
November 1996Clippy Too
anthropomorphic but not human.
“Optimized for first use”. Always.
Enabled by default. Every time.
USEFUL VS FUN
Yisela Alvarez Trentini | MRM McCann, August 2016 17
Why people hate the paperclip:Labels, appearance, behavior and
Social responses to user interface agentsLuke Swartz
Stanford University 2003
FITNESS ASSISTANTIn 2008, Cory Kidd completed a study with a robot intended to aid in fitness and weight loss goals, by providing a social presence with which study participants tracked their routines.
Yisela Alvarez Trentini | MRM McCann, August 2016 18
Most used “he” or “she” when talking about their robot
1 never returned the robot
Group 1Pen & Paper
Group 2Touchscreen
Group 1Touchscreen + Robot
THE PERCEPTION OF BOTS
We instinctively treat computers like people and use the same standards of politeness, gender stereotypes, teamwork and reciprocity.
Many said that Eliza helped them, and some asked the people conducting the test to leave them alone with her so they could discuss things in private. It was even considered a low-cost way to handle people with mild psychological problems.
Yisela Alvarez Trentini | MRM McCann, August 2016 19
THE UNCANNY VALLEY
However, making making machines more humanlike is good up to a point, after which they become discomforting or creepy.
Yisela Alvarez Trentini | MRM McCann, August 2016 20
WHAT DOES THIS MEAN FOR CHATBOTS?
Project Name | Date 21
PERSONALITY
Most users will build a relationship with their bots.
In conversational UIs, personality is the new UX. The entire app experience is reduced to a few lines of text. Microcopy is now king.
Writers and comedians collaborate with UX to create engaging bots.
Yisela Alvarez Trentini | MRM McCann, August 2016 22
CHAT BOTS WORKING ALONGSIDE HUMANS
Supervised A.I: In Facebook M, bots gather information for an eventual interaction with a human rep.
Yisela Alvarez Trentini | MRM McCann, August 2016 23
While the rep answers, the machine learns.
How?
1. Stateless Bots2. Semi-Stateful Bots3. Stateful Bots
24
A.I. POWERED BOTS: MACHINE LEARNING
For a machine to recognise cats, a person must first provide it with thousands of photos of cats and not cats.
Yisela Alvarez Trentini | MRM McCann, August 2016 25
CAT!
Feed the machine thousandsof photos of cats (and noncats)
Show the machine a photo of a cat,and it should recognise it as such.
?
cat
cat cat
cat catNot cat
Not cat
A.I. POWERED BOTS: DEEP NEURAL NETWORKSNeural network are used to simulate densely interconnected brain cells. The computer can learn things, recognize patterns, and make decisions in a humanlike way.
Yisela Alvarez Trentini | MRM McCann, August 2016 26
NOT CAT!
CAT!
color edgescolorblobs
Samatha is an intelligent personal OS from the movie Her by Spike Jonze. Has impressive knowledge of the physical world. Can understand human emotion and show
empathy. Can reason and debate.
Yisela Alvarez Trentini | MRM McCann, August 2016 27
BOTS AND LANGUAGE
Our language is a compact and effective system, but relies on the assumption of intelligence and a
common social and physical world.
Yisela Alvarez Trentini | MRM McCann, August 2016 28
„The trophy will not fit in the brown suitcase because it was too big.“
What was too big? Answer 0: the trophy; Answer 1: the suitcase.
BOTS AND LANGUAGE
Yisela Alvarez Trentini | MRM McCann, August 2016 29
• Bots need to learn to speak (or think) like us, without a physical world or the time to learn about it like we have.
• Machines also need to understand how humans work on an emotional level: Detect and analyse emotions, extract concepts from dialog, showing empathy.
• Symbolic processing (humans) + machine learning (system)
• We both need a model of the other, as well of one of the world.
ONCE OUR BOTS CAN TALK...
• Do we want „always-aware“ systems?• Should our assistant bots interact with other people’s?• How can we teach the machine introspection? (the ability to
communicate the exact process that leads to their choices.) • Is it necessary to make machines human-like? Do they need
to converse? Do you talk to a bot, or use one?• What are the dangers of captology and creating bots that
can be truly persuasive?• Human Rights Watch is looking for an international treaty to
ban military robots with autonomous lethal firing power. How far are we from this?
Yisela Alvarez Trentini | MRM McCann, August 2016 30
IF YOU EVER MAKE A BOT...
1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
31
REFERENCES• Weizenbaum, Joseph "ELIZA – A Computer Program For the Study of Natural Language
Communication Between Man and Machine" in: Communications of the ACM; Volume 9 , Issue 1 (January 1966): p 36-45.
• About neural networks: http://www.explainthatstuff.com/introduction-to-neural-networks.html
• How Humans Respond to Robots: Building Public Policy through Good Design https://www.brookings.edu/research/how-humans-respond-to-robots-building-public-policy-through-good-design/
• "WHY PEOPLE HATE THE PAPERCLIP: LABELS, APPEARANCE, BEHAVIOR AND SOCIAL RESPONSES TO USER INTERFACE AGENTS", Luke Swartz, June 12, 2003, Honors Thesis for Symbolic Systems Program, Stanford University
• C. Kidd and C. Breazeal. A Robotic Weight Loss Coach. Twenty-Second Conference on Artificial Intelligence, 2007.
Yisela Alvarez Trentini | MRM McCann, August 2016 32