Wishesh Magazine February 2020 Wishesh Magazine February 2020 | Page 52

DESI NEWS
Incredible AI That ‘Understands’ Is Being Built
By A Desi Scientist
For advancement in the field of AI, it will be great to have artificial
intelligence that can even understand humans. This is going to be a
real task to accomplish in the field of Artificial Intelligence. For now,
Alexa and Siri follow all the instructions such as booking tickets or
reading a message for you.
52
The Defence Advanced
Research Projects Agency
in the US is giving funds to
researchers to make agents
that can even understand
human emotions. Further,
they want to see the AI
agents be teammates
for humans.
An Indian Scientist
and a team of
researchers from
the University of
Arizona are working
on this. To make this
possible they have
been given a grant of
$7.5 million. Further,
they want to create
an agent who has the
ability to understand
social cues and make
independent rational
decisions. The Desi Scientist
is a 30-year-old Adarsh
Pyarelal. He is the Principal
investigator for the DARPA-
Funded project.
This entire concept has been
WWW.WISHESH.NET | FEBRUARY 2020
termed as ‘Theory of Mind’
and the idea is to be able to
predict human interactions,
actions, and non-verbal signs.
What is their plan? The team
is going to partner with human
players to complete custom-
with human teammates. So
they will rely on AI agents
to respond back and give
their suggestions to help the
teammates.
A professor from the
University of West of
England is skeptical about
whether this concept will
work as he thinks that
using video games to
teach AI social intelligence
has disadvantages as it
might not be suitable to
situations in real life.
Pyarelal is from Kerala
and he says that studying
human interactions in the
real world is vast when
compared to studying
human-machine teaming
in a virtual testbed.
designed Minecraft missions.
Taking help from digital and
physical sensors, AI will collect
information about individual
players and their interactions
This entire concept will
capture the video and audio
of a person so that the agent
is used to identifying humans
by their voice and face,
instead of just their face.