OPINION PIECE | THE MACHINE CAN RAP. SO WHAT MAKES AN ARTIST HUMAN?

There was a time when suc­cess­fully imper­son­at­ing a rap­per required, at min­im­um, a base­ball cap, sev­er­al regret­table hand ges­tures and the con­fid­ence of a drunk man com­mand­eer­ing the aux cable at 2am.

Tech­no­logy has ‘improved’ mat­ters considerably.

Arti­fi­cial intel­li­gence can now clone a voice, approx­im­ate a flow, man­u­fac­ture a beat, write a verse, con­struct an artist and, increas­ingly, depos­it the res­ult­ing cre­ation onto the same stream­ing plat­forms occu­pied by people who have endured the con­sid­er­ably more incon­veni­ent busi­ness of actu­ally being alive.

Which leaves hip-hop with an unusu­ally exist­en­tial question.

If a machine can rap, what exactly have we been listen­ing for all this time?

AI has already wandered through hip-hop’s front door. Syn­thet­ic voices increas­ingly resemble recog­nis­able artists closely enough to cre­ate the pecu­li­ar exper­i­ence of hear­ing some­body who sounds uncan­nily like a rap­per you know say­ing words that rap­per nev­er spoke.

The tech­no­logy is mov­ing faster than the etiquette around it.

And per­haps that mat­ters more in hip-hop than almost any­where else.

Because rap has always placed enorm­ous weight on the voice not merely as an instru­ment, but as evidence.

Who said it matters.

Where they came from matters.

Why they said it matters.

The his­tory behind the voice matters.

This is also why hip-hop is per­haps the worst genre from which to mount a simplist­ic argu­ment that tech­no­logy cor­rupts authen­t­ic creativity.

This is a cul­ture built, in part, by tak­ing machines designed to do one thing and mak­ing them do some­thing they were nev­er inten­ded to do.

Turntables became instru­ments. Breaks were exten­ded. Records were sampled, chopped, reversed and reas­sembled. Drum machines pro­duced sounds that ini­tially struck tra­di­tion­al musi­cians as arti­fi­cial. Pro­du­cers learned to bend samplers bey­ond their sup­posed lim­it­a­tions until tech­no­lo­gic­al restric­tion itself became an aesthetic.

Hip-hop has nev­er been frightened of artificiality.

What it has tra­di­tion­ally cared about is who is doing the manip­u­lat­ing, and why.

Sampling James Brown is not the same cre­at­ive act as pre­tend­ing to be James Brown. Flip­ping a soul record into some­thing its cre­at­ors could nev­er have ima­gined is fun­da­ment­ally dif­fer­ent from man­u­fac­tur­ing a syn­thet­ic voice and ask­ing it to per­form on command.

One trans­forms a recording.

The oth­er starts inter­fer­ing with a person.

And hip-hop has always under­stood that dis­tinc­tion bet­ter than out­siders some­times assume, because the genre’s obses­sion with authen­ti­city has nev­er really meant every lyr­ic must func­tion as sworn testimony.

Rap­pers exag­ger­ate. They con­struct char­ac­ters. They myth­o­lo­gise themselves.

MF DOOM per­formed behind a met­al mask while build­ing one of hip-hop’s most intric­ate fic­tion­al uni­verses around the per­sona of a com­ic-book villain.

Ghost­face Kil­lah can move from street report­age to sur­real­ist imagery so quickly that real­ism almost ceases to be a use­ful category.

Kool Keith has cre­ated enough altern­ate iden­tit­ies to pop­u­late a mod­er­ately sized psy­chi­at­ric wait­ing room.

Hip-hop has always under­stood performance.

What audi­ences want is not neces­sar­ily fac­tu­al truth.

They want human truth.

That is con­sid­er­ably harder to manufacture.

A great rap voice car­ries inform­a­tion bey­ond its lyr­ics: geo­graphy, his­tory, insec­ur­ity, bravado, humour, class, age, ambi­tion, trauma, absurdity. Some­times all with­in four bars.

Con­sider how much you know before cer­tain rap­pers have actu­ally said anything.

Nas can make a street corner feel sim­ul­tan­eously like memory and mythology.

Black Thought attacks a verse with the slightly unnerv­ing pre­ci­sion of someone who appears to have mis­placed the concept of run­ning out of breath.

Ka could make a line barely louder than con­ver­sa­tion feel heav­ier than anoth­er rapper’s entire album.

billy woods often sounds as if some­body has handed you a clas­si­fied doc­u­ment, removed three pages and refused to explain why.

Amy True can move from intro­spec­tion to social obser­va­tion with the unmis­tak­able sense that the polit­ics are not dec­or­at­ive addi­tions to the music but part of the reas­on the music exists in the first place.

ODB occa­sion­ally soun­ded sur­prised that record­ing equip­ment had been invented.

These are not simply voices.

They are accu­mu­lated lives.

That dis­tinc­tion becomes clear­er when you listen to artists whose writ­ing is insep­ar­able from specificity.

Take Nas.

Ill­mat­ic would be tech­nic­ally impress­ive even if you removed every bio­graph­ic­al detail from it. The intern­al rhyme, imagery and eco­nomy remain extraordin­ary. But nobody loves that record simply because the syl­lables are arranged efficiently.

You can hear Queens­bridge inside it.

Hall­ways. Pro­ject win­dows. Stair­wells. Friends. Police. Radi­os. Child­hood memor­ies already harden­ing into adult understanding.

A machine could ana­lyse every rhyme scheme on the album.

It could learn how fre­quently Nas places intern­al rhymes, how long his lines tend to be, what vocab­u­lary appears repeatedly and how his cadence sits against a par­tic­u­lar drum pattern.

Even­tu­ally it could prob­ably pro­duce some­thing that sounds dis­con­cert­ingly plausible.

But plaus­ib­il­ity is not Ill­mat­ic.

The streets described on that record exis­ted before some­body con­ver­ted them into data.

And the same prin­ciple travels.

It is there in Ka’s Browns­ville. In Little Simz’s North Lon­don. In Earl Sweatshirt’s interi­or land­scapes. In the strange, frag­men­ted worlds of MIKE and billy woods.

And it is there in Amy True’s work too.

True has nev­er soun­ded par­tic­u­larly inter­ested in sep­ar­at­ing the MC from the cit­izen. Her music moves through hip-hop, soul and spoken word, but what holds it togeth­er is inten­tion: ques­tions of com­munity, power, per­son­al respons­ib­il­ity, res­ist­ance and the occa­sion­ally unfash­ion­able idea that a rap­per might actu­ally have some­thing they would like soci­ety to think about.

That mat­ters here.

Because you could train a mod­el to recog­nise the son­ic prop­er­ties of polit­ic­al rap remark­ably quickly.

Feed it enough con­scious hip-hop and it will learn the vocabulary.

Oppres­sion.

Sys­tem.

Free­dom.

Power.

Revolu­tion.

Throw in Babylon if the con­fid­ence threshold is high enough.

Tech­nic­ally, you might end up with some­thing that resembles protest music.

But polit­ics assembled as vocab­u­lary is not the same thing as conviction.

Amy True’s Ques­tion Author­ity is not inter­est­ing because an algorithm could identi­fy a stat­ist­ic­ally unusu­al con­cen­tra­tion of anti-estab­lish­ment ter­min­o­logy. Her work exists with­in years of grass­roots cul­ture, per­form­ance, col­lab­or­a­tion, com­munity work and the wider eco­sys­tem of UK con­scious hip-hop from which those ideas emerged. Her debut album Elev­en sim­il­arly moved between hip-hop, jazz and soul rather than treat­ing “con­scious­ness” as a pre­set that could simply be applied to a beat.

That is the dis­tinc­tion AI keeps for­cing us back towards.

Con­text.

A machine can repro­duce the vis­ible res­ult of belief.

Believ­ing some­thing is rather more inconvenient.

The same is true across the best rap music.

These artists are not inter­est­ing because they con­form per­fectly to the tech­nic­al defin­i­tion of rap.

They are inter­est­ing because some­body lived long enough to devel­op a way of seeing.

AI can repro­duce the aud­ible char­ac­ter­ist­ics of that world­view extraordin­ar­ily well.

What it can­not pos­sess is the his­tory that made the world­view necessary.

At least, not yet.

And that might be the point at which the increas­ingly tedi­ous debate about wheth­er AI music is “real art” becomes more useful.

Because tech­nic­al com­pet­ence was nev­er the scarce resource.

There are thou­sands of people cap­able of rap­ping extremely well.

Plenty can rhyme multisyllabically.

Plenty can stay per­fectly in the pocket.

Plenty can con­struct immacu­late six­teen-bar verses which you admire enorm­ously while they are hap­pen­ing and for­get com­pletely before the next bus arrives.

Skill mat­ters.

But skill alone has nev­er explained why one rap­per becomes indispensable.

The mys­ter­i­ous bit has always been why this person?

Why does Yasi­in Bey sound like Yasi­in Bey before you’ve even pro­cessed the words?

Why can Roc Mar­ciano turn under­state­ment into a worldview?

Why does Earl Sweat­shirt some­times leave more emo­tion­al inform­a­tion in the space around a line than anoth­er rap­per puts into the line itself?

Why can Ghost­face start a verse sound­ing furi­ous, become hil­ari­ous halfway through and some­how end some­where close to heartbreak?

Why does Amy True’s social com­ment­ary feel dif­fer­ent from simply ask­ing a machine to write “a con­scious rap song”?

Because art is not merely inform­a­tion trans­mit­ted elegantly.

It is inform­a­tion attached to a human being.

And nowhere is that attach­ment more obvi­ous than in hip-hop.

Rap has spent half a cen­tury turn­ing bio­graphy into mythology.

Neigh­bour­hoods matter.

Crews mat­ter.

Accents mat­ter.

Rival­ries matter.

Clothes mat­ter.

Inter­views matter.

Record shops, pir­ate sta­tions, hous­ing estates, sub­way lines and the par­tic­u­lar sofa some­body spent six months sleep­ing on all even­tu­ally find their way into the mythology.

We do not encounter a great rap­per solely through songs. We build a con­tinu­ing rela­tion­ship with a char­ac­ter whose music keeps acquir­ing mean­ing as their life unfolds.

Even fail­ure becomes material.

Espe­cially failure.

Which cre­ates an obvi­ous prob­lem for the AI rapper.

What exactly can it have failed at?

It has nev­er been booed.

It has nev­er spent six years mak­ing music nobody cared about.

It has nev­er pressed 500 cop­ies of a record and dis­covered that approx­im­ately 487 of them now live per­man­ently under­neath somebody’s bed.

It has nev­er fallen out with its pro­du­cer over £80.

It has nev­er per­formed to four­teen people above a pub while insist­ing to its friends after­wards that the crowd was “actu­ally really receptive”.

It has nev­er watched anoth­er artist it privately con­sidered less tal­en­ted become suc­cess­ful first.

It has nev­er become suc­cess­ful itself and dis­covered that suc­cess does not repair whatever it ima­gined suc­cess would repair.

There is no humi­li­ation in its archive.

And humi­li­ation, incon­veni­ently, is one of humanity’s great cre­at­ive resources.

The tech­no­logy com­pan­ies might answer that none of this matters.

Listen­ers care about the fin­ished product. If an AI-gen­er­ated song cre­ates the same neur­o­lo­gic­al response as a human-made one, why should its ori­gin dimin­ish the pleasure?

This argu­ment is more per­suas­ive than artists might like.

Most of us already con­sume cul­ture with sur­pris­ingly little con­cern for how it was assembled. Hip-hop records have long involved pro­du­cers, engin­eers, sample sources, fea­tured musi­cians and col­lab­or­at­ive writ­ing pro­cesses of vari­ous kinds.

The romantic image of one wounded geni­us trans­lat­ing their soul dir­ectly into a micro­phone has always involved a cer­tain amount of excel­lent PR.

But AI intro­duces a more pro­found separation.

It allows us to remove the incon­veni­ent human entirely.

And once you do that, some­thing strange happens.

Music stops being com­mu­nic­a­tion and becomes content.

There is a difference.

Con­tent exists to occupy space.

It accom­pan­ies workouts, adverts, shops, scrolling, gam­ing, com­mut­ing. It is per­fectly cap­able of being pleas­ant. Increas­ingly, it may be machine-generated.

Music, at its best, cre­ates the dis­con­cert­ing sen­sa­tion that anoth­er per­son has felt some­thing you thought belonged only to you.

Hip-hop made an entire glob­al cul­ture out of that recognition.

A teen­ager thou­sands of miles from Queens­bridge could hear Nas describe the view from his win­dow and some­how recog­nise their own.

Someone who had nev­er set foot in Brook­lyn could hear Yasi­in Bey rap about the city and under­stand that geo­graphy can become psychology.

A young listen­er in Bri­tain could hear an artist like Amy True artic­u­late ques­tions about power, soci­ety or per­son­al struggle and recog­nise that hip-hop did not have to arrive wear­ing some­body else’s accent, geo­graphy or politics.

They could encounter The Roots, Com­pany Flow, Gang Starr or De La Soul and dis­cov­er an entire cul­tur­al fam­ily tree lead­ing back­wards and side­ways into jazz, soul, funk, poetry and resistance.

That trans­mis­sion requires some­thing more than son­ic accuracy.

It requires some­body on the oth­er end.

This does not make AI use­less to musicians.

Quite the opposite.

Artists will use it as they have used every pre­vi­ous tech­no­lo­gic­al devel­op­ment: irre­spons­ibly at first, bril­liantly eventually.

They will build new instru­ments from it. Pro­du­cers will cre­ate sounds that cur­rently seem impossible. Inde­pend­ent artists will gain cap­ab­il­it­ies once avail­able only to wealthy studios.

Some genu­inely strange record will even­tu­ally arrive that could not have exis­ted without AI, and every­body cur­rently com­pos­ing furi­ous Linked­In posts about “pro­tect­ing cre­ativ­ity” will pre­tend they always knew this was possible.

That is gen­er­ally how these things work.

The most inter­est­ing artists, though, will not ask AI to replace imagination.

They will use it to extend imagination.

And AI may inad­vert­ently make one unfash­ion­able artist­ic qual­ity con­sid­er­ably more valuable.

Spe­cificity.

If machines can cre­ate com­pet­ent music infin­itely, com­pet­ence becomes almost worthless.

The thing worth listen­ing for becomes the detail a machine would nev­er think to invent because it nev­er happened to it.

The smell of your grandmother’s hallway.

The bus route home.

The par­tic­u­lar takeaway every­one went to after the club shut.

The friend who stopped call­ing when the money arrived.

The jack­et some­body stole from you in 2009 that you remain, against all reas­on­able adult judge­ment, furi­ous about.

The polit­ic­al argu­ment that changed your mind.

The com­munity that shaped your under­stand­ing of the world before any­one was inter­ested in stream­ing it.

The oddly shaped frag­ments from which actu­al iden­tity is made.

Listen to Little Simz and Lon­don is not merely a loc­a­tion prin­ted under­neath a biography.

Listen to Ka and Browns­ville does not func­tion as branding.

Listen to Nas and Queens­bridge is not scenery.

Listen to Amy True and con­scious­ness is not an aesthetic.

Place becomes part of the thought process.

Memory determ­ines rhythm.

Exper­i­ence determ­ines vocabulary.

Belief determ­ines what is worth say­ing in the first place.

Per­haps the arrival of arti­fi­cial artists will not make human­ity obsolete.

Per­haps it will simply expose how much gen­er­ic human-made music was behav­ing like arti­fi­cial intel­li­gence already.

The inter­change­able flexes.

The algorith­mic­ally approved vulnerability.

The chor­uses engin­eered for twelve seconds of social video.

The pro­duc­tion choices copied from whatever per­formed well last quarter.

The appar­ently “con­scious” song assembled from a shop­ping list of approved polit­ic­al nouns.

The songs that appear to have been writ­ten by some­body who has stud­ied every suc­cess­ful record of the pre­vi­ous six months while some­how fail­ing to have a single ori­gin­al exper­i­ence dur­ing that period.

AI might become extremely good at mak­ing that.

In which case, good luck com­pet­ing with it.

The altern­at­ive is to become more incon­veni­ently, embar­rass­ingly and spe­cific­ally human.

That may ulti­mately be why artists such as DOOM, Nas, Ka, Black Thought, Little Simz, Amy True, Earl Sweat­shirt or billy woods remain such use­ful ref­er­ence points in this conversation.

You could feed a machine their entire catalogue.

Teach it the rhyme schemes.

Map the cadences.

Cata­logue the vocabulary.

Meas­ure every pause.

Even­tu­ally you might gen­er­ate some­thing that resembles them closely enough to make some­body stop scrolling.

But imit­a­tion runs into a problem.

DOOM had to become DOOM.

Nas had to stand at those windows.

Ka had to know Brownsville.

Simz had to live the life from which those obser­va­tions emerged.

Amy True had to devel­op the con­vic­tions behind the words, not simply the vocab­u­lary in which those con­vic­tions are expressed.

billy woods had to acquire whatever col­lec­tion of exper­i­ences has res­ul­ted in him sound­ing per­man­ently as if West­ern civil­isa­tion has dis­ap­poin­ted him personally.

A mod­el receives the arte­fact after the fact.

The artist had to accu­mu­late the life that pro­duced it.

That dif­fer­ence may become increas­ingly dif­fi­cult to hear.

It does not there­fore become meaningless.

A machine can learn your cadence.

It can study your rhyme patterns.

It can repro­duce the rasp in your throat and cal­cu­late the tempo at which your audi­ence most reli­ably nods its head.

It can learn every word Amy True has ever recor­ded without under­stand­ing why some­body might feel com­pelled to ques­tion author­ity in the first place.

What it can­not do is explain why your voice mattered before some­body taught it how to copy it.

And for a cul­ture built on the rad­ic­al pro­pos­i­tion that an over­looked per­son with a micro­phone could make the world listen to their story, that dis­tinc­tion feels worth defending.

AI can rap.

The more inter­est­ing ques­tion is wheth­er it has any­thing to say.