BBC_Knowledge_Asia_Edition_-_May_2016_

&

PHOTOS: GETTY, DREAMSTIME ILLUSTRATION: SAM FALCONER

1. LAVA FLOWS
   Hawaiian and Icelandic volcanoes produce
   slow-moving lava. Lava temperature is
   700-1,200°C, so it melts or ignites most
   things. A house on stilts of titanium or
   tungsten might survive, if the stilts were
   strong enough to withstand the lava
   pushing against them.


2. AIRBORNE ASH
   Violent volcanoes, such as Vesuvius and
   Mount St Helens, tend to explode and
   throw up several cubic kilometres of ash
   and rock. A 30cm-thick ash layer can be
   heavy enough to cause roofs to collapse, so
   you’ll need a reinforced roof with a steep
   pitch to stop the ash building up too much.


3. POISONOUS GAS
   After an eruption, pyroclastic flows can
   engulf a town in superheated steam and
   poisonous sulphur dioxide or asphyxiating
   carbon dioxide. To escape this, you’ll need
   an airtight home with an air supply –
   preferably underground. But ensure your
   access hatch doesn’t get blocked!

THE THOUGHT EXPERIMENT

Could I build a house that would survive a volcanic eruption?

To make

we use sof a e a

simulates neurons

connected in networks like in

a brain. These networks are

trained with data until they

can learn patterns or make

predictions about what data

of UK people actually have

coeliac disease, yet 7 per

cent believe they have a

gluten intolerance

1

per cent

was a T. rex’s top speed –

even a fairly fit human

could outrun it at full pelt

11

km/h

IN NUMBERS

g

lik h h l

understand speech or

recognise car number plates,

so in this respect computers

can learn a little bit like

humans. But humans are still

much better – we can learn

l d st

u be of d ffe e deas

As we still don’t fully

understand how brains work,

computers are unlikely to be

as good at learning as

humans for hundreds of

years. PB

e computers learn,

software that

might come next. Methods

like these help computers

complex concepts and a va

number of different ideas.

Can computers

learn like

humans?

Robots would definitely

love BBC Magazines