New Scientist - USA (2019-07-13)

(Antfer) #1
12 | New Scientist | 13 July 2019

THE race for hypersonic
weapons is heating up.
China, Russia and the US
are all attempting to create
weapons that travel at more
than five times the speed of
sound. The technology could
escalate tensions around
the world.
Hypersonic weapons
move incredibly fast, but
what differentiates them from
traditional ballistic missiles is
that all of their journey is done
within the atmosphere, rather

than through space. They
can also be manoeuvred
during flight, making them
harder to defend against.
Last month, US aerospace
giant Raytheon announced
it was preparing to test a
hypersonic scramjet, a jet-
powered vehicle that moves
at rocket-like speed and gets
oxygen from the atmosphere.
The test was part of the US

Pentagon’s three-pronged effort
to obtain hypersonic weapons, a
top priority that attracts billions
of dollars in funding.
The other two approaches
are the Air-Launched Rapid
Response Weapon, which uses
new but not publicly known
technology, and the Hypersonic
Conventional Strike Weapon,
which repurposes existing
technology. All three weapons
will be launched from aircraft
and carry conventional
explosives. They could be
in service in the early 2020s.
Russia is also progressing
with two hypersonic weapons:
the Kinzhal (Dagger), launched
from a plane, and the Avangard,
a hypersonic vehicle launched
from an upgraded 1960s
ballistic missile. Avangard is
“invulnerable to intercept by
any existing and prospective
missile defence”, said Russian
president Vladimir Putin,
after witnessing a test in 2018.
Avangard may be ready by the
end of the year and Kinzhal
could already be in use.
China has carried out more
hypersonic tests than the US

in recent years, such as for
the Xingkong-2 missile and
Jiageng-1, which is believed
to be a scramjet.
“Developing hypersonic
weapons has become an
end in itself, first as a kind of
competitive sport, then as an
arms race,” says Mark Gubrud
at the University of North
Carolina. It is part of a trend
towards faster warfare, he
says. A report by US think
tank RAND Corporation warned
that hypersonics will compress
reaction times and lead to “hair
trigger” responses. The result
could be increased instability,
says Justin Bronk at defence
think tank Royal United
Services Institute.
“When nuclear missiles
replaced bombers, it was highly
destabilising because they gave
you minutes rather than hours,”
says Bronk. He says that was
why nuclear powers had
ways to mitigate the reduced
response time, like the Moscow-
Washington hotline, a means of
direct communication between
US and Russian leaders.
Recently, US president
Donald Trump said he changed
his mind about launching
conventional strikes on Iran
with minutes to spare. The
faster pace of hypersonic
operations might reduce
time for such second thoughts.
Gubrud says a ban
on hypersonic test flights,
which has been discussed by the
United Nations, would stop the
arms race. “US national security
would hardly suffer if plans
were put on hold at least long
enough to see if China and
Russia would reciprocate.”  ❚

A Russian fighter
jet carries a Kinzhal
hypersonic missile

Military technology

David Hambling

AL
EX


EI^ N


IKO


LS
KY
/GE


TT
Y


News


1.7 km/s
Hypersonic weapons
travel at least this fast

China, Russia and the US


in hypersonic arms race


MANY people have been inspired
by football’s 2019 Women’s
World Cup in France. Now artificial
intelligences are learning the
game too.
Karol Kurach and his colleagues
at Google Research in Zurich,
Switzerland, have made a virtual
football training pitch that AIs can
use to understand how to play.
Software simulates a
standard football game, with
features including goals, corner
kicks, offsides and penalties. An
AI controls one or all 11 players
on a team and tries to defeat
another AI opponent.
Google has tested its own
AI on it and another from sister
company DeepMind, but anyone
can download the program to
train their own AIs.
Because football requires both
short-term control and high-level
tactics, it is challenging for AIs to
grasp, says Kurach.
To master football, an AI must
learn to deal with unpredictability –
for example, when a player kicks
the ball, it may land in different
locations or be intercepted by

the opposing team. “Unlike games
such as chess or Go, there is not a
set model of moves,” says Kurach.
The approach differs from the
computer opponents used in games
such as the FIFA series, which are
manually programmed by game
designers to use specific rules and
strategies. “Such bots can do only
what they were programmed for
and always follow the same
pattern,” says Kurach.  ❚

For machines to
master football,
they need to
balance short-
term control and
high-level tactics

Machine learning

Donna Lu

GO

OG

LE
RE

SE
AR

CH
,^ ZÜ

RIC

H

Google has made a
virtual football pitch
for training AIs
Free download pdf