ADVANCES
18 Scientific American, May 2019
Illustration by Thomas Fuchs
TURJOY CHOWDHURY
Getty Images
SEISMOLOGY
Missing
Earthquakes
Found
Geologists fill in mysterious gap
in Nepal’s quake record
Like a slow-motion car crash, the Indian
subcontinent is colliding with Eurasia. This
impact, along a fault known as the Main
Himalayan Thrust, is the force driving the
rise of the Himalayas. In April 2015 it trig-
gered Nepal’s magnitude 7.8 Gorkha
earthquake, which destroyed villages and
parts of Kathmandu, killing thousands.
Nepal is no stranger to such temblors—
but in the western part of the country, no
significant earthquakes have been recorded
since 1505. Such a “seismic gap” could be
bad news: if the region’s faults are not rel
easing their pentup stress every so often,
one or more very large and potentially cata-
strophic earthquakes could result. “With
more than 500 years of waiting, the stored
energy due to the convergence of India
[and Eurasia] could be considerable,” says
Zakaria Ghazoui, a geologist at the Institute
of Earth Sciences in Grenoble, France. The
sudden release of this energy could devas-
tate nearby places such as Pokhara, one of
Nepal’s largest cities.
To determine whether this seismic
gap actually exists, Ghazoui and a team
of researchers ventured onto Rara Lake in
western Nepal’s Himalayas and retrieved
cores of sedimentary layers from the bot-
tom. They suspected the cores might con-
tain records of when any past earthquakes
occurred because quakes can cause under-
water avalanches that leave behind layers of
what Ghazoui calls “disorganized” sediment.
The team found evidence of at least
eight avalanches since 1505, correspond-
ing to moderate to large quakes. “We were
hoping to find the trace of the 1505 earth-
quake, but the discovery of the other
earthquaketriggered avalanches was a
real surprise,” he says. This means that the
area’s faults may not be storing as much
stress as has been thought. But it also
NEURAL ENGINEERING
Decoding
Speech
New approach is a step
toward translating thoughts
into machinespoken words
Neurological conditions that can cause
paralysis, such as amyotrophic lateral scle-
rosis (ALS) and strokes in the brain stem,
also rob many patients of their ability to
speak. Assistive technologies enable key-
board control for some of these individuals
(like the famed late physicist Stephen
Hawking), and braincomputer interfaces
make it possible for others to control
machines directly with their thoughts. But
both types of devices are slow and imprac-
tical for people with lockedin syndrome
and other communication impairments.
Now researchers are developing tools to
eavesdrop on speechrelated brain activity,
decode it and convert it into words spoken
by a machine. A recent study used stateof
the art machine learning and speechsyn-
thesis technology to yield some of the most
impressive results to date.
Electrical engineer Nima Mesgarani of
Columbia University’s Zuckerman Institute
and his colleagues studied five epilepsy
patients who had electrodes implanted in
or on their brain as part of their treatment.
The electrodes covered regions involved in
processing speech sounds. The patients
listened to stories being read aloud as
their brain activity was recorded. The team
trained a “deep learning” neural network to
match this activity with the corresponding
audio. The test was then whether, given
neural data it had not seen before, the sys-
tem could reproduce the original speech.
When the patients heard the digits zero
through nine spoken four times each, the
system transformed the neural data into
values needed to drive a vocoder, a special
kind of speech synthesizer. A separate
group of participants heard the synthe-
sized words and identified them correctly
75 percent of the time, according to the
study, published in January in Scientific
Reports. Most previous efforts have not
measured how well such reconstructed
speech can be understood. “We show that
it’s intelligible,” Mesgarani says.
Researchers already knew it was possi-
ble to reconstruct speech from brain activ-
Aftermath of the 2015 Nepal earthquake in Sankhu.
© 2019 Scientific American © 2019 Scientific American