Computer Shopper - UK (2020-03)

ISSUE385|COMPUTERSHOPPER|MARCH2020 107

computers, but instead uses a
technique called quantum annealing.
More akin to analogue than digital
computing, this model of computing
has been used foroptimisation
problems and image analysis.

Manual methods
Mathematician Lewis FryRichardson’s
contribution to computing pre-dated
the first digital computers, but it was
remarkably forward thinking, pre-
empting as it did the methods used
todayinweather forecasting, and even
the concept of massively parallel
computing. Where it differs from all
the other models of computing
presented here is that it relied on the
human brain instead of any artificial
method of processing. There’s no way
it has afuture,but that makes it no
less fascinating on our whistle-stop
tour of computing paradigms.
Richardson’s idea, which has been
dubbed his weather factory and was
published in 1922, used people,
equipped only with slide rules and
other basic calculating devices, to
produce anumerical weather forecast.
Each human calculator sat inside a

huge globe that showed amap of the

world, solving the differential equations

that related to the part of the map

where theysat. The results were

displayed so that people working in

neighbouring areas could read them,

so theycould be used in their own

calculations. Apulpit, on topofapillar

in the centre of the globe,was home

to aconductor,touse an orchestral

metaphor,who maintained uniform

speed of operation throughout the hall.

The moleculeoflife

The idea of computing by mixing

chemicals in atest tube might sound

odd in the extreme,but it’s agood

description of the early work in the

field of computing with DNA, the

double helix molecule that has been

called the chemical of life.

DNA molecules represent

information, which is encoded in the

sequence of so-called bases, or

chemical groupings, along their

backbone,anorder that can be

engineered artificially.What’s more,

that sequence of bases dictates which

DNA strands will react chemically with

each other,therefore suggesting their

potential forcomputation. But the

real advantage is that just afew grams

of DNAcontain avast number of

molecules, therebyopening up the

possibility of massively parallel

computing. This concept was first

demonstrated 25 years ago.Itwas used

Neuralnetworks

Animalbrainsworkinaverydifferentwaytodigital

computersand,forthisreason,theyexcelatsometasks,

suchaspatternrecognition,whereourcomputerstend

to struggle.What’smore,theydon’tactuallyrequire

programming,inthenormalsenseoftheword,asthey’re

taughtto carryoutataskthroughaprocessoflearning.

Forthisreason,artificialneuralnetworks,orANNs,have

beenstudiedfordecades.

Neuralnetworkscanbeconsideredastheultimatein

parallelcomputing.Eachprocessingelement,orneuron,

isextremelysimple,doinglittlemorethangeneratinga

weightedsumofitsinputs,generatinganoutputifthat

sumexceedsathreshold.Thepowercomesfromtheir

vastnumberandtheevengreaternumberofconnections

betweenthem,andthefunctionofaparticularnetworkis

definedbytheweightingfactorofeachneuron’sinputs.

Historically,mostofthisresearchhasprovedfruitless,

butrecentwork,forexampleatManchesterUniversity,

hasmadeconsiderableprogress.However,whilethe

underlyingprinciplesofthisworkarethesameasthose

thatoccurinthehumanbrain,theinherentlyanalogue

natureofbiologicalneuronshas,to date, beenemulated

digitally,albeitwithveryspecialisedhardware.

Intermsofthenumberofneuronsandtheir

interconnections,ANNsdon’tyetcomeanywhereclose

to thehumanbrain.Yetthereisnoshortageof

applicationswhereexpertsexpectthemto surpass

conventionalcomputers.Includedhereareself-driving

cars,autonomousdrones,intelligentsurveillancesystems

andevenhumanoidrobotsto actashomehelps.

RIGHT:Although today’s emphasis is on digital implementations,

the inherently analogue nature of mammalian neural networks

offers the benefits of massively parallel computing

ABOVE:The four

bases allow data to

be represented by

DNA modules, and

selective chemical

reactions permits

computation