AN EXPLOSION that baffled
astronomers in 2018 may
finally have an explanation.
Observations of X-rays from
the blast have revealed that
it was probably a massive star
that only partially blew up,
leaving behind either a dense
neutron star or a small black
hole – something we have long
suspected happens, but never
seen until now.
When astronomers spotted
AT2018cow – nicknamed “the
Cow” because of the chance
order of the letters in its official
designation – it took only days
to reach its peak brightness.
Most supernovae take weeks or
months to reach their peaks. It
was also 10 to 100 times brighter
than typical supernovae.
That made it difficult to
explain. “This sudden blip out
in space was really unusual,
and we don’t see many of these.
This is the nearest by and the
best studied,” says Kate Maguire
at Trinity College Dublin in
Ireland, part of the team that
discovered it.
It was so unusual that
astronomers around the world
immediately started observing
it. “We watch stars explode allthe time – supernovae are
very common now, and there
are lots of what are ultimately
fairly minor differences – but
it’s really rare that you look
at something and just say,
‘I have no idea what this is’,”
says Daniel Perley at Liverpool
John Moores University in the
UK. “There were so many
things that made it weird.”
Luckily, because the Cow
is relatively nearby, at about
195 million light years away,it was possible to use various
telescopes to observe it in
different wavelengths of
light. Dheeraj Pasham at the
Massachusetts Institute of
Technology and his colleagues
used the Neutron Star Interior
Composition Explorer (NICER)
space telescope to capture
the X-rays coming from the
explosion’s aftermath over
a period of about two months.
They found that the strength
of this radiation oscillated up
and down every 4.4 milliseconds.As a general rule, the speed of
oscillation of light coming from
an object in space is proportional
to its size, so the researchers
calculated that the object at
the centre of the Cow must be
no more than 1300 kilometres
across (Nature Astronomy, DOI:
10.1038/s41550-021-01524-8).
That means that it had to be
a neutron star or a relatively
small black hole, referred to
as compact objects. “People
have been suspecting that these
kinds of extreme explosions
could be the birth of black holes
or neutron stars, but this is a
final piece of evidence that I
think really settles the case,”
says Pasham.
The object at the centre of
the Cow could be either, but it
is harder to find a neutron star
model that fits as the oscillation
of light is steady. “In most
neutron star models, you
wouldn’t expect it to be steady,”
says Perley. “Other observations
about this object also point
towards it being a black hole.”
Combining this work with
other data from the Cow
suggests that the object
probably formed after a failed
supernova in which some
of the material exploded and
some fell back in on itself.
Regardless of which type
of compact object lies at the
heart of the Cow, the fact that
we spotted it days after the
explosion is a big deal, says
Pasham. “When you go to
school, you learn that when
stars explode, black holes and
neutron stars are born, but
nobody has actually detected
a compact object immediately
after a supernova,” he says. ❚10 | New Scientist | 18/25 December 2021
SL
OAN^ DIGI
TA
L^ SKY
SURV
EYAstronomyLeah Crane‘Space cow’ explosion was
probably a failed supernova
A 2018 blast was
labelled AT2018cow and
nicknamed “the Cow”195m
Distance in light years
to the Cow from EarthArtificial intelligenceChris Stokel-WalkerALMOST no one reads the long and
complicated terms and conditions
found on websites and apps, but
now an artificial intelligence can
pick out the important bits for you.
Jason Hong at Carnegie Mellon
University in Pennsylvania and
his colleagues trained a machine-
learning model to highlight
important clauses that users
may want to pay attention to.
To do this, the researchers
first took 1551 terms and
conditions (T&C) statements
from 27 shopping websites. They
then split the documents into more
than 200,000 pairs of sentences.
The team asked people to look
at a pair of sentences and rank
which was most important, using
the results to rank the statements
in the whole document in order of
importance. Participants picked out
terms defining consumer rights,
along with the ability to get refunds
or buy items on credit. Any fees
were also deemed useful to know.
The researchers then fed this
list to a machine-learning model,
training it to look for important
clauses. The AI was 92 per cent
accurate at highlighting important
statements in new T&C texts it was
given (arxiv.org/abs/2111.12182).
However, it tripped up on
sentences using the word
“not”, says Hong. “If they say
something like, ‘We will not
charge you for refunds’ versus,
‘We will charge you’, those tend
to often get classified the same,
unfortunately,” he says.
Lilian Edwards at Newcastle
University, UK, argues that making
T&C texts more readable doesn’t
solve the main issue: they are
so dictatorial that “meaningful
consent online in the consumer
context does not exist”, she says.
“Rather than trying to highlight
‘important’ phrases to be wary
of, we should be asking for better
rights for users.” ❚AI reads the boring
terms and conditions
so you don’t have toNews