The EconomistJune 1st 2019 Science & technology 732
1Once the Local Bubble was established,
any cosmic rays created by a supernova
within it would have kept bouncing off this
magnetic wall and back into the bubble.
They would thus have strafed every object
within it, including Earth, for tens, or even
hundreds of thousands of years after the
explosion that created them.
Some of these rays were the nuclei of a
radioactive isotope of iron that is created
almost exclusively in supernovas. These
unstable nuclei, together with their decay
products, have been found in the ocean
floor on Earth and in rock samples brought
from the Moon—another reason to believe
the supernovas happened. Those isotopes
found on Earth can be dated from the sedi-
ment they are in. The strongest signal is
from 2.5m years ago, indicating that this
explosion was the closest.
A geological feature that coincides with
the period when Local Bubble supernovas
were going off is an increase in traces of
charcoal in oceanic sediment. That is evi-
dence of wildfires on land. This increase
starts about 7m years ago and in turn coin-
cides with a period when much of Earth’s
vegetation shifted from forests to grass-
lands. The fires recorded by the oceanic
charcoal could explain this vegetational
shift, because grass is more resilient to fire
than trees are. What explains the fires,
though, remains mysterious.
Dr Thomas and Dr Melott propose that
the culprit is cosmic rays from the local su-
pernovas. The main arsonist of wildfires is
lightning. The hammering of atmospheric
molecules these rays handed out, they sug-
gest, caused more lightning. The rays
would knock such molecules apart, liberat-
ing electrons from their atoms. These liber-
ated electrons would in turn knock loose
other electrons, creating cascades that
would make the air electrically conductive.
This would encourage lightning strikes.
Observations made recently on a moun-
tain in Armenia, of electron cascades
caused by normal cosmic rays, showed that
many of these did indeed end in a lightning
flash, so the idea is plausible. Encouraged
by this, Dr Thomas and Dr Melott calculat-
ed the effect that the cosmic rays of the ex-
plosion of 2.5m years ago would have had
on the number of cascades. They conclude
that the cascade rate would have increased
50-fold.
The replacement of forests by grassland
is thought by some anthropologists to have
encouraged the evolution of humanity’s
ancestors away from tree-climbing and to-
wards the bipedalism. It was this change in
locomotion that freed human hands to get
up to all the mischief which distinguishes
people from other species. Human beings,
in all their manipulative glory, are thus, if
the chain of events Dr Thomas and Dr Me-
lott are suggesting is correct, the children
of dying stars. 7H
uman beingscan pick up and manip-
ulate objects and tools with hardly a
thought. This seemingly simple task, how-
ever, requires the precise, co-ordinated
movement of individual fingers and
thumbs, each applying the correct amount
of pressure, at exactly the right places, to
the object handled.
That people can do this successfully is
thanks to special nerve endings, called
mechanoreceptors, found in their skin.
These provide instant tactile feedback to
the brain of the shape, feel and weight of
whatever is being grasped. With time and
experience, people learn to vary their grip
instinctively when lifting a golf ball, for ex-
ample, as opposed to an egg.
Replicating that dexterity in robots is
hard. A machine usually picks things up by,
first, identifying the object via a camera
and appropriate software, and then using a
preprogrammed grasping strategy appro-
priate to what it thinks it has seen. This ap-
proach has improved greatly in recent
years, thanks to advances in machine
learning and vision. Further improvement
will, however, be best served by a more pre-
cise understanding of the mechanics of
how people themselves manipulate ob-
jects. A new “smart” glove, from computer
scientists at the Massachusetts Institute of
Technology, promises to do just that.Writing in this week’s Nature, Subrama-
nian Sundaram and his team describe a
way to measure, quickly and easily, the
forces a human hand exerts as it grasps and
manipulates things. Their invention is a
sensory sleeve that fits over a knitted glove.
Attached to the sleeve’s palm, and running
up the fingers and thumb of the glove, are
pieces of a film that generates electricity in
response to pressure. This film has electri-
cally conducting threads running through
it to form a set of grids (see picture). Each of
the 548 places on the grids where two
threads overlap acts as a localised pressure
sensor—the equivalent of a mechano-
receptor in the skin of a hand. The signals
from the threads are fed to a computer for
storage and interpretation.
In their experiment, Dr Sundaram and
his colleagues asked people to put on one
of these gloves and use their now-gloved
hand to pick up and manipulate 26 every-
day objects—a mug, a pair of scissors, a
spoon, a pen and so on—one at a time for a
few minutes each. The system then record-
ed the signals from the threads seven times
a second as every object was held and
moved in its turn.
The trick was to take these recordings
and train a machine-learning program,
called a neural network, to interpret them.
Since many of the best neural networks
available are designed to learn and inter-
pret images, Dr Sundaram decided to pre-
sent the team’s data to the network visual-
ly, by making each of the seven-a-second
samples into an image in which the colour
of the grid points represented the pressure
applied there in shades of grey from low
(white) to high (black). Once the network
had been trained on these images it could
then identify any of the 26 test objects from
new pressure maps recorded by the glove.
Measuring in this way just how a hu-
man hand exerts force will, as originally in-
tended, be useful in programming robots
to mimic people more closely when they
pick objects up. But Dr Sundaram’s experi-
ments also provide insights into how the
different parts of the hand work together
when grasping things—how often, for ex-
ample, the first finger is used at the same
time as the thumb or the second finger.
Those data, he reckons, could assist de-
signers of prosthetic limbs in the perfec-
tion of their devices.
Dr Sundaram’s invention is clever, then.
But one of the cleverest things of all about it
is that it is also cheap, for the glove costs
only around $10 to make. This, he hopes,
will encourage others to create their own
versions. Building comprehensive tactile
maps of how people employ their hands to
manipulate the world will require huge
sets of data—ideally derived from many
thousands of individuals handling thou-
sands or millions of objects. To gather data
on that scale requires cheap tools.Improving robots’ grasp requires a new
way to measure it in peopleRoboticsHand in glove