MAY 2019. DISCOVER 63
VIS
UA
L^ G
EN
ER
ATI
ON
/SH
UT
TER
STO
CK
“I’m alive-ish,” Cortana replied.
In a similar vein, Alexa said, “I’m not really
alive, but I can be lively sometimes.”
The Google Assistant was clear cut on the
matter. “Well, you are made up of cells and I
am made up of code,” it said.
Siri, meanwhile, was the vaguest. “I’m not
sure that matters,” she answered.
Foster says that while the writers don’t want
Cortana to masquerade as human, they also
don’t want her to come across as an intimidat-
ing machine. It’s a tricky balance. “She’s not
trying to be better than humans,” Foster says.
“That’s a creative stake we put in the ground.”
I tested Cortana’s humility by asking, “How
smart are you?”
“I’d probably beat your average toaster in
a math quiz,” she replied. “But then again, I
can’t make toast.”
THE FUTURE IS CUSTOMIZATION
Some developers dream of abandoning uni-
formity and instead customizing voice AIs.
One reason that this hasn’t already happened,
though, is personas require intensive manual
effort to create. While machine learning now
powers many aspects of voice AIs, their char-
acters are currently rigged using manually
authored, rules-based approaches.
Some researchers have begun to explore
ways that computers could use machine learn-
ing to automatically mimic different perso-
nas. Personality customization, taken to the
logical extreme, would result in a different AI
for each user. While that sounds impractical,
intense tailoring is something that computer
scientists are considering. Witness U.S. Patent
No. 8,996,429 B1 — “Methods and Systems
for Robot Personality Development.” With a
mix of dull legalese and what reads like 1950s
pulp fiction, the document describes a vision
for bespoke AIs.
The hypothetical technology described in
the patent is able to customize how it talks and
behaves by learning everything it can about the
user it serves. The robot looks at the user’s cal-
endar, emails, text messages, computer docu-
ments, social networks, television viewing,
photos and more. Armed with all this informa-
tion, the robot then builds a profile detailing
the “user’s personality, lifestyle, preferences
and/or predispositions,” according to the pat-
ent. It would also be able to make inferences
about the user’s emotional state and desires at
any given moment. The ultimate purpose for
all of the above would be so the bot can present
the best possible personality to any given user,
one that is “unique or even idiosyncratic to
that robot.”
The document could be dismissed as an
entertaining curiosity if not for a couple of
key factors. It was written by two respected
computer scientists, Thor Lewis and Anthony
Francis. And the patent assignee is Google.
The technology they describe is far from
reality. But we’ve now seen how computer
scientists can teach voice AIs to understand
speech and produce it themselves and do so
with verve and personality. All of this makes
our interactions with AIs more efficient and
enjoyable as we task them with little chores
throughout the day.
But similar to how eating one potato chip
makes you crave the whole bag, the first tastes
of personable interaction have made some
technologists hungry for a whole lot more.^ D
Excerpted from TA LK TO
ME: How Voice Computing
Will Transform The Way
We Live, Work, And Think
© 2019 by
James Vlahos.
Reproduced by
permission of
Houghton Mifflin
Harcourt. All
rights reserved.