New Scientist - 19.10.2019

10 | New Scientist | 19 October 2019

Technology

NEW SCIENTIST readers are more

knowledgeable than the general

public and experts on certain issues,

but don’t score much better than

monkeys would on some questions.

“To score worse than monkeys

requires misconceptions,” Ola

Rosling, author of Factfulness, told

attendees of New Scientist Live on

10 October. Most people are not

only ignorant of some basic facts

about the world, they don’t even

realise they are ignorant, he said.

For example, about 88 per cent

of children around the world are

now vaccinated against at least one

disease, but most people think the

figure is much lower. When asked if

this proportion is around 20, 50 or

80 per cent, only about 15 per cent

of people in countries such as the US

and the UK get the answer right in

Rosling’s surveys. At a recent world

health summit, only 27 per cent

of attendees got it right. Nobel

laureates and medical scientists

would be outsmarted by monkeys

randomly picking answers, he said.

New Scientist readers do a bit

better, though. In an online survey,

46 per cent of New Scientist

readers correctly answered the

vaccination question. On climate,

they excelled. When asked what

climate experts believe will happen

to global temperatures over the

next 100 years – grow warmer,

cooler or stay the same – 99 per

cent opted for the right answer.

In other surveys, the proportion

answering correctly has ranged

from 94 per cent in Hungary to

just 76 per cent in Japan. In the US,

81 per cent get it right and 87 per

cent do so in the UK.

But on questions about

endangered species and world

population trends, New Scientist

readers fared worse. Overall, they

got 3.9 out of 12 questions right.

“That’s on par with monkeys,”

said Rosling. The average score

is just 2.2.  ❚

AUTOMATED accounts on social

media platforms have been

accused of many abuses, including

large-scale political manipulation,

but they aren’t all bad. Speaking

at New Scientist Live on 12 October,

Tony Veale from University

College Dublin, Ireland, and

Mike Cook from Queen Mary

University of London argued that

bots can also be forces for good.

For example, many Twitter

bots created for fun or art don’t

pretend to be human, unlike

those with more serious or

sinister aims. These bots

often follow in the footsteps

of surrealist artists, working

to make the everyday seem new

and fascinating, said Veale.

“When a bot strives to be

creative, people respond

creatively,” he said. People choose

to follow bots to see how they

subvert our expectations by

building language algorithmically

rather than intuitively, he said.

Cook shared a number of

examples, including a bot that

tweeted every word from the

dictionary in order, provoking

startling engagement with certain

words. Notably, “butt” received far

more retweets than the average.

“Lots of people in our field see

this as the future of creativity,”

said Cook. Anyone with a basic

knowledge of coding can build a

bot, and its success relies on how

others interact with it. “Some of

the best bots are incredibly simple.

It’s really about how it works

in the space that it’s in,” he said.

Cook also demonstrated three

Twitter bots that he had created

for the event. All three use a set

of 45,000 New Scientist headlines

that have appeared online since

2003 to build new headlines, but

each bot uses a different method.

The first bot (@NewerScientist1)

uses a neural network that learns

from the original headlines

to generate strange new

ones, including gems such as

“Self-destruction of the brain

power” and “Inside the most

powerful thing”.

The second (@NewerScientist2)

uses a Markov chain, which works

by learning what words generally

go together in New Scientist

headlines and combining them

in common orders. Its output

includes “Global warming: Will

the anaconda or the ‘Garbage of

Eden’?” and “A total solar eclipse

with non-addictive cigarettes by

alien worlds without words”.

The final bot (@NewerScientist3)

uses a simpler protocol similar to

cut-and-paste methods used in

songwriting by artists like David

Bowie. Cook divided the headlines

into lists of topics; the bot replaces

words from a headline under one

topic with words from another.

Its output included headlines

such as “There may already be

crows on Mars”, “Tiny pebbles may

be the reason most politicians

spin in the same direction” and

“Leopards that live in memes are

protecting people from rabies”. ❚

“ One dictionary-tweeting

bot provoked startling

engagement when it

posted the word ‘butt’”

Knowledge

A bot-created headline

that splices together bits of

previous New Scientist ones

Leah Crane

News

The surrealist art of Twitter bots

Three bots that generate New Scientist-inspired headlines unveiled

Congratulations: you

are (a bit) smarter

than average

Michael Le Page

Nobel laureates

would be

outsmarted

by monkeys on

some questions

about the world

SEREGRAFF/GETTY IMAGES

2019