Will human creativity survive automation & AI? | Viputheshwar Sitaraman | TEDxBend

The rise of mobile-web and social media ushered in an era of content creation like never before, putting the power of creatorship at the literal fingertips of almost every digital citizen. Yet just as quickly, for the first time, technology is encroaching from the realm of efficiency, into that of creativity. Take, for instance, news publishing: robot writers & algorithmically curated news feeds have automated the entire process from ideation to distribution. While automation's implications for the economy & job markets are well-explored, the ethical dilemma & broader consequences of removing the human element to creativity are often overlooked. Today, it's in the hands of technologists & engineers to guide the future of creative fields.

Vip Sitaraman is the creator of digital products & platforms which have reached over 50 million users online and the youngest American to raise venture capital. He is a creative consultant to organizations ranging from leading Silicon Valley startups to Fortune 500 & Inc. 1000 companies.

This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx

I want you to think back

to your favorite scene
from a movie or a novel.

What makes that so great?

The job of a creator,

from the first prehistoric man
to paint on a cave

to Andy Warhol,

has been to construct human experiences.

What defines the geniuses among us

are those whose creations
grab your attention,

the kind of wonderment and captivation

that keeps a child awake with a book
and a flashlight under the sheets.

We can feel and relate
to the lonely solitude of Edgar Allan Poe,

the metaphysical delirium of Dali,

and even the comedic frustrations
of everyday life in Seinfeld.

In that sense,

creative works fundamentally
foster an unspoken dialogue

between the creator
and his or her audience.

So I couldn't believe it

when I read that The Washington Post
had published over 850 articles

in the pilot year
of their robot writer Heliograph.

I couldn't fathom

that an unthinking, unfeeling machine

could create something
to truly engage real human readers,

so I decided to search further into it.

So it turns out the majority
of the articles written by Heliograph

concerned the 2016 election cycle,

and the rest covered sports
and the Olympics.

But perhaps what's more interesting

is that the post is not alone
in its automation endeavors.

Giants like Associated Press and Forbes

have invested in methods
known as natural language processing,

which generate reports
and stories automatically

on data-heavy topics
like sports and finance.

Long story short,

we are unknowingly reading
thousands of stories written by bots.

This begs the question:

Have we reached a point where machines
can entirely replace human creativity?

Before we answer that,

let's first define what creativity means.

There's a four-step process

that unifies a life cycle
of almost every creative concept:

first, learning - the process
of gathering information and knowledge;

then phase two, ideation -
the origin of a new creative concept;

followed by three, production -
going from concept to reality;

and finally four, bringing creativity
to the masses - distribution.

Historically, technology has accelerated
the production and distribution phases

of the creative lifecycle.

In the 14th century,
during the Renaissance,

perhaps the most notable creative
technology of all time

came in the form of the printing press,

which enabled the first widely
distributed newspapers and books.

The birth of modern fashion

came with the 18th-and-19th-century
industrial revolutions,

which created the textile loom,
cotton gin and the first synthetic dyes.

And most recently,

the digital era ushered in
new mediums for communication

replacing the telephone and radio
with screens - our first televisions -

and computer technologies

such as the digital camera, software,
the Internet and social media.

New technologies have always preceded
periods of unprecedented creative novelty

in fields ranging from design and fashion
to literature and marketing,

and almost universally,

innovation has increased
access for creativity

for both consumers and creators.

However, the next revolution,
which I call the age of automation,

threatens to extinguish
the last flame of human creativity.

Mechanization has already started
to sweep across creative endeavors.

Ingenious advertising campaigns,

once brought to life by shrewd creative
directors in Madison Avenue agencies,

have been replaced by digital platforms

which make microsecond decisions
choosing from hundreds of bidders

based on data on each
user's likes and interests.

Big-data, automated decision-making
has lobotomized the industry.

Brands now focus on competing
how well they can target users

while the creative art of memorable
storytelling has been all but lost.

In news media, writers
aren't the only ones suffering.

Editors have lost their clout
to social feeds,

where content is personalized
and curated at scale.

We're seeing the effects in the industry:

5-10% haircuts in staff across the board.

Just consider ESPN's layoffs in 2017

that included prominent
on-air personalities,

or just earlier this year

when 2100 reporters, journalists
and editors were laid off

from Huffington Post, BuzzFeed,

Vice and the newspaper
conglomerate Gannett.

Machines have even made
a foray into fine art.

AI trained on the works
of 14th-20th-century French painters

produced a portrait

which went on to sell at Christie's
for half a million dollars.

All of this marks a reversal in a trend
that has spanned millennia,

and we have to consider two implications
in the wake of this dramatic change:

first, a culture of consumers;

and second, the challenge
of human engagement.

A society dominated
by consumers of creative content

with few or no original creators

means the death of diversity of opinion
and originality of thought.

We used to just have to end up
in a waiting room or an airport gate

to find a broadcast or magazine
with views that defer from our own,

but today, social feats
insulate us in bubbles,

echo chambers of our
own thoughts and ideas.

This has dramatic effects
on the political and ideological climate,

widening gaps and re-entrenching
dichotomies of opinion.

Moreover, the immediate assumption

would be that moving
humans from the equation

would decrease bias in media.

But in reality, third-party datasets
being pulled by robot writers

leaves a vacuum of accountability.

This is how social media virtually
manufactured the fake-news scandal

by bringing factually incorrect
but socially controversial headlines

to the top of our Facebook news feeds.

However, the drawbacks of automation

aren't merely ethical
or philosophical dilemmas;

there's a real often overlooked
risk reward in business.

Several startups and large companies,

such as Spotify Creator Lab Technologies,

Google Magenta Nsynth and IBM Watson Beat,

attempt to use AI in identifying
or creating hit music.

Similarly, publishers motivated by
the breakout growth of Facebook and Reddit

have sought to use automation strategies
on sites like Huffington Post and Yahoo.

Automation sounds like
a promising approach

to reduce decision-making costs
and increase creative throughput,

and in the short term it can be.

But as publishers
have learned the hard way,

the risk is losing human engagement
and revenue in the long term.

Treating AI as a panacea

for all operational inefficiencies
and decision-making costs

has long-term consequences,

and continuing to overinvest in it
in such approaches

can easily become tomorrow's
dotcom or Bitcoin bubble.

The limitations of AI can be understood

by exploring the dominant form
of deep learning today,

known as supervised learning.

In this model,

a machine learns a mapping function
from an input X to an output Y

by examining a wealth of input data.

But in that sense,

today's creative automation technologies

do not create truly novel
original concepts;

rather they work very well

within the variables and model
they're trained for

but cannot consider what Rumsfeld
called unknown unknowns.

In the words of Scientific American:

"For AI to get creative, first it must
learn the rules, then how to break them."

This shortcoming has been evident

since a 1996 study
of a joke bot called Jape,

which generates puns and riddles,

uses the same technology
as Forbes and Associated Press do

to create their stories,

and all of its results
were taxonomically correct,

but it's comedic success
was significantly improved

by pairing with the human creator.

For instance, here's an example
of something it came up with on its own,

left to its own devices:

"What do you call a lenient shelter?

A lax deduction."

It's a play on words
but it isn't particularly funny.

On the other hand,

here's something that a human
combined with the bot came up with:

"How's a nice girl
like a chocolate birdie?

They're both sweet chicks."

Cute and silly.
Definitely funnier.

Much better.

At this point, it should
be abundantly clear

that human decision-making complemented
with computer-generated suggestions

is far more effective than either alone.

In the words of Doug Engelbart,
creator of the computer mouse:

"Technology should not aim
to replace humans,

rather to amplify human capabilities

and to end automation risks

either having no engagement at all
or have the wrong impact altogether."

But at the same time,

I'm not saying that augmentation
entails no automation -

merely implementing it in the right
aspects of the creative process.

Building human experiences
requires human discretion.

And there are examples

of how creative distribution
and production have been accelerated

by automation and AI technologies.

Newsrooms have used automation
for image captioning, sourcing,

and aggregating data
from various libraries.

Adobe has put out
a variety of AI projects

that accelerate
time-consuming design tests,

such as stitching scenes
and masking skies,

or removing discrepancies from video.

IBM Watson's news Explorer

uses natural language processing
and sentiment analysis

to extract meaning
and connotation from the news -

aggregating global trends -

and Google's Tensorflow library

accelerates tedious video
processing activities,

recognizing objects and even human poses

with applications ranging from
video-game design to self-driving cars.

The burden of technologists

is to guide the forward progress
of humanity with innovation,

just take, for example,

the incredible potential
of atomic fission,

the same concept behind
the world's deadliest weapon

as well as what is arguably one of our
most renewable energy sources.

Similarly, AI has incredible,
unimaginable potential

to change creative possibilities,

but finding its correct implementations
will be the difference

in preserving the incredible potential
that only humans have for creativity.

But for those of us who aren't
creating the latest AI technology,

we are far from powerless.

It will take actionable,
self-inspired action

to identify what creative
content we consume,

rather than allowing technology
to dominate the entire process,

in order to give back power
to original creators

in an increasingly mechanized market.

The challenge to pioneers
of this latest technological revolution

will be to find the right place
for automation in the creative process,

and doing so will mean preserving
the incredible and infinite potential

that humans have for creativity.

Thank you.

(Applause)