New Scientist - 21.09.2019

6 | New Scientist | 21 September 2019

TRY as we might, we can’t prove

Albert Einstein wrong. One

prediction of his general theory

of relativity is that black holes are

simple objects. Now listening to

them “ring” suggests this is true.

According to general relativity,

any black hole can be described

by three properties: its mass, spin

and electrical charge. In practice,

this boils down to the first two,

because we don’t expect black

holes to accumulate charge.

All other information about

a black hole – like the properties

of objects that have fallen in – can’t

be observed from beyond the

event horizon. This information

is called “hair” and so the idea is

known as the no-hair theorem.

Observations of black holes

have all been consistent with

this idea. But Maximiliano Isi

at the Massachusetts Institute

of Technology and his colleagues

wanted to test it in a different way:

using the ripples in space-time

called gravitational waves.

We know that when a pair of

black holes merge, the leftover

hole should ring like a bell,

emitting gravitational waves in

several frequencies. Just as the

note of a bell is determined by its

shape, the frequencies of waves

produced are determined by the

black hole’s mass and spin.

“These frequencies and their

lifetimes are inextricably tied

to the shape of the bell, so you

can listen to the ringing and learn

about its structure,” says Isi.

The longest-lasting frequency

is called the fundamental. There

are also shorter-lived notes called

overtones. “The fundamental

rings like a high-quality wine glass,

and the overtones are more like

a thud,” says Leo Stein at the

University of Mississippi.

Isi and his team found an

overtone in a signal detected

by the Laser Interferometer

Gravitational-Wave Observatory

(LIGO) in 2015. This is the first time

anyone has found more than one

tone in a gravitational wave.

The mass and spin of the black

hole had already been calculated

by the LIGO team based on all the

information in the signal. Isi and

his colleagues have now used just

overtone frequency to estimate

the mass and spin (Physical Review

Letters, doi.org/gf799b).

They calculate that the black

hole is about 68 times the mass

of the sun and spinning some

100 times a second. That is a

good match with the previously

calculated value. What’s more,

because Isi’s estimate is based on

the no-hair theorem, this suggests

that the theorem is correct. In

other words, Einstein is still right.

The result isn’t very precise.

The black hole’s properties could

still deviate from those predicted

by relativity by up to 20 per cent.

But there is a good chance that the

test can be repeated. “The result

was from the loudest binary black

hole signal we’ve had so far, but

there are more signals that haven’t

been analysed yet,” says Katerina

Chatziioannou at the Flatiron

Institute in New York. Those

measurements should allow

physicists to nail down whether

SCIENCE PHOTO LIBRARY black holes have hair. ❚

News

Machine learning

Bots defy laws

of physics to win

at hide-and-seek

NEVER play games with a bot – it

will find a way to cheat. A team from

OpenAI, an artificial intelligence

lab in San Francisco co-founded by

Elon Musk, has developed artificially

intelligent bots that learned to

cooperate by playing hide-and-

seek. The bots also learned how to

use basic tools and that defying the

laws of physics could help them win.

In April, a team of bots known as

the OpenAI Five beat the human

world champions at team-based

video game DOTA 2. Bowen Baker

at OpenAI and his colleagues

wanted to see if the team dynamics

of the OpenAI Five could be used

to generate skills that could one day

be useful to humans.

The hide-and-seek bots use

similar principles to learn but the

simpler game allows for more

inventive play. The team set the

bots loose in a simulation filled with

fixed walls and movable boxes and

left them to play millions of team

games of hide-and-seek.

At first, the hiders simply ran

away. But, they soon worked out

that the quickest way to stump

the seekers was to find objects

in the environment to hide

themselves, using them like a

sort of tool. For example, they

learned that boxes could be

used to block doorways and

build simple hideouts.

The real surprise came when the

bots started exploiting glitches

in their environment. Seekers

found that if they pushed a ramp

towards a wall, they could launch

themselves into the air and spot

hiders from above. Hiders found

that they could get rid of the ramps

for good by shoving them through

exterior walls at a certain angle.

Such tricks show that AIs

are able to find solutions that

humans miss, says Baker. Maybe

the bots will even be able to solve

problems that humans don’t yet

know how to, he says. ❚

Douglas Heaven

Astronomy

Leah Crane

The sound of a black hole

Cosmic monsters seem to “ring” in just the way Einstein predicted

Black holes produce

characteristic notes

as they merge

“ Maybe the bots will even

be able to solve problems

that humans don’t yet

know how to”