Transcript
Hybrid Intelligence The New Paradigm
in Artificial Intelligence
@clarecorthell
Clare Corthell
thinkingmachin.es/events
Luminant Data is a Machine Intelligence Consultancy
building more intelligent businesses with data science strategy and artificial intelligence technology
clare@luminantdata.com
based in San Francisco
Clare Corthell, Founder
datasciencemasters.org
The Open Source Data Science MastersThe most popular curriculum for
learning data science
Artificial Intelligence, “AI” has massive potential to change
how we live and work
@clarecorthell
Two visions of the future
@clarecorthell
The Vision of the Interneta huge catalog
highly structured standardized
firstly,
@clarecorthell
the data format
the dream
@clarecorthell
The Vision of the InternetThe Web was designed as an information space, with the goal that it should be useful not only for human-human communication, but also that machines would be able to participate and help. One of the major obstacles to this has been the fact that most information on the Web is designed for human consumption, and even if it was derived from a database with well defined meanings (in at least some terms) for its columns, that the structure of the data is not evident to a robot browsing the Web.
Tim Berners-Lee, 2001On the “Semantic Web”
@clarecorthell
The Vision of the Interneta huge catalog
highly structured standardized
@clarecorthell
The Vision of the Interneta huge catalog
highly structured standardized
Reality
a mess
@clarecorthell
The Vision of a world with Machinessecondly,
1. Machines take over all tasks 2. Humans no longer required 3. Humans Prosper.
@clarecorthell
1. Machines take over all tasks 2. Humans no longer required 3. Humans Prosper
some stuff
The Vision of a world with MachinesReality
@clarecorthell
So in a world where the internet is a mess and machines can’t do everything for us
What do we do?
@clarecorthell
Artificial Intelligence is a collection of technologies that emulate intelligence
two specific technologies Machine Learning uses statistics to make educated guesses Natural Language Processing helps computers understand human language
@clarecorthell
Broadly stated, Machine Learning lets us take patterns from the past and project them going forward, sometimes allowing us to infer what might happen, including guess what a human might think or do
@clarecorthell
The Computer’s Limits of Understanding
those r cats! where is the cat plzcomputercomputer
@clarecorthell
Machine Learning is about 75% accurate, or maybe 90% in certain circumstances.
That’s not good enough for most products.
@clarecorthell
Computers and Humans have very different talents
Humans Creativity Intuition Choosing arbitrarily Learning new things Being empathic
Computers Math Paring down options Repeating forever Being fast
@clarecorthell
Neither the Computer nor the Human is objectively better —
they just have different strengths
@clarecorthell
Given this truth about humans and computers,
@clarecorthell
In building businesses and technologies, we must anticipate a different future.
@clarecorthell
A New Paradigm
@clarecorthell
The Hybrid Intelligence Paradigm(not neural nets or the superintelligence)
@clarecorthell
Leverages what humans do best and what computers do best.
Asks humans to make judgments when the computer is less confident —
resulting in the most accurate, trustworthy system.
The Hybrid Intelligence Paradigm
@clarecorthell
Dev
elopm
ent: T
rain
ing
& T
est
mac
hine
learn
ing a
lgorit
hm
newinput prediction
Active Learning Pattern Credit: Lukas Biewald, Crowdflower
Hybrid Intelligence Design Pattern:Active Learning
human predictionco
nfide
ntnot confident
output
human labelled “feedback” data added back into training data
Production: Prediction Production: Active Learning
@clarecorthell
Where are we seeing Hybrid AI in applications?
@clarecorthell
Consumer Applications Artificial Assistant
@clarecorthell
It’s simple, right?
@clarecorthell
Dev
elopm
ent: T
rain
ing
& T
est
decis
ion
tree
algor
ithm
newrequest
suggested response
Hybrid Intelligence Design Pattern:Hybrid Interaction
human validation
sendresponse
human labelled “feedback” data added back into training data
send response
don’t send response
extra
ct e
ntitie
s & c
aptu
re in
tent structured
data
compose new response
“I want to take the next flight to Cebu this afternoon.”
“I have booked you for the 5pm flight. Have a nice trip!”
execute task
book flight
existing context
current time current location credit card
@clarecorthell
Hybrid Intelligent Agent Systems
Hybrid Distributed Humans
Automated
urge
ncy
of r
espo
nse
slow
from fully automated to fully humanfas
t “Hybrid Interaction”
“Active Learning”
SiriDuer
Cortana
Customer Service Centers
@clarecorthell
25%75%
Human AI
The Reality for Artificial Assistants
humans deal with everything that isn’t
an easy, known case
@clarecorthell
90%10%
Human AI
The Dream for Artificial Assistants
humans deal with contextually tough
or new cases
@clarecorthell
Artificial Assistants don’t work very well yet.
But there’s one important thing to pay attention to —
and it has to do with design.
@clarecorthell
Hi , How are you?
@clarecorthell
Anthropomorph Design
chat interaction establishes an interaction paradigm
human language establishes empathy and trust
@clarecorthell
The Hybrid AI Paradigm will change everything — it’s already happening.
In the next customer service revolution that leverages Hybrid Interaction,
BPOs must be ready for the boom.
@clarecorthell
Business Process Outsourcing
Humans will never be irrelevant.
@clarecorthell
The Most Important Point Here:
Humans are the only part of the equation endowed with empathy
and the creativity to imagine the future.
@clarecorthell
top related