AIthropology Lab
ESENTWI

Anthropology ne artificial intelligence

AI a ɛda so ara nnim sɛnea ɛbɛtie

Pan African AI & Innovation Summit 2026 ho dawurubɔ krataa a ɛreka “The $2.9 Trillion Question: Will Ghana Shape Africa’s AI Decade or Be Shaped by It?” ho asɛm wɔ Kempinski, Accra.
Pan African AI & Innovation Summit, Accra, 22–23 September 2026. · Summit no hyehyɛfoɔ na wɔde mfonini no maeɛ.

Accra Pan African AI & Innovation Summit no mu no, nkɔmmɔ no yii asɛm foforo adi wɔ Afrika kasa ahorow ne AI ho. Ɛnyɛ kasa bɛn na AI te ase nko, na mmom nkitahodi bɛn na wɔasiesie no sɛ ɛnhwɛ kwan.

·Martín González Senosiain

AI a ɛda so ara nnim sɛnea ɛbɛtie

Accra Pan African AI & Innovation Summit no mu no, nkɔmmɔ no yii asɛm foforo adi wɔ Afrika kasa ahorow ne artificial intelligence ho. Ɛnyɛ kasa bɛn na AI te ase nko. Asɛm no nso fa nkitahodi kwan bɛn na wɔhwɛ kwan sɛ nnipa de bɛba ho. AI pii a yɛde di dwuma nnɛ no susuw sɛ onipa no tumi kenkan, kyerɛw na ɔwɔ internet daa. Nanso mmeae pii no, kasa ne nkitahodi yɛ soronko.

Pan African AI & Innovation Summit 2026 da a ɛto so abien wɔ Accra no mu no, ɔkasa bi maa asɛm bi a mpɛn pii no ɛyera wɔ model, computing capacity ne digital sovereignty ho nkɔmmɔ mu no da adi pefee. Saa da no, wɔde wɔn adwene sii talent ne impact so, na wɔkaa skills, Afrika mu solutions, akwahosan, kuayɛ ne climate ho dwumadi ahorow ho asɛm.

“NLP in Ghana is not a text problem. It is a voice problem.”

Adwene a ɛbaa akyi no yɛ mmerɛw sɛ wobɛte ase. Sɛ wɔyɛ language model wɔ kurom kasa mu mpo a, ɛrenyɛ mfaso kɛse sɛ nnipa a wɔpɛ sɛ wɔde kɔma wɔn no mfa saa kasa no nkyerɛw pii. Nnipa binom a technology yi betumi aboa wɔn paa no tumi tie na wɔka kasa, nanso ebia wɔnkenkan. Sɛ yɛhwɛ saa kwan yi so a, ade a ɛsɛ sɛ yɛfi ase nyɛ sɛ yɛde kasa foforo bɛka text-based system ho. Ɛsɛ sɛ yɛfi ase wɔ voice ho.

Saa asɛm yi betumi ayɛ te sɛ ade ketewa kosi sɛ yɛbɛhwɛ tebea ahorow a artificial intelligence a ɛwɔ hɔ nnɛ gyina so. Sɛ wode prompt kyerɛw wɔ box mu a, ɛkyerɛ sɛ wutumi kenkan na wokyerɛw, wowɔ screen ne keyboard anaa touch interface, wunim sɛnea wobisa system no asɛm, na mpɛn pii no wowɔ internet a ɛyɛ den sɛ wubetumi ne remote servers akasa. Saa tebea yi abɛyɛ ade a wɔn a wɔyɛ digital systems binom sua ho araa ma ɛtumi yera wɔ wɔn ani so.

Interface no nyɛ ade a ɛnni afa biara. Ɛwɔ adwene bi wɔ onipa a wɔhwɛ kwan sɛ ɔde bedi dwuma no ho.

Data ba model anim

Saa presentation no ara kyerɛɛ sɛ linguistic inequality no fi ase ansa na model no reba. Sɛnea ɔkasafo no kae no, Whisper, speech-recognition systems a nnipa nim paa no mu baako, de audio bɛyɛ nnɔnhwerew 680,000 na wɔde tete no. Local project a na wɔrekyerɛ no de nnɔnhwerew 16 pɛ na na ɛwɔ.

“How do you compete?”, ɔkasafo no bisae.

Mmuae no nyɛ sɛ wɔbɛtwɛn akosi sɛ wɔn nso benya corpus a ɛte saa. Project no afi ase reyɛ audio, text ne translation pairs na ɛrekyɛ no wɔ open kwan so sɛnea teams afoforo betumi de adi dwuma. Dɔnhwerew biara a wɔde to gua no yɛ dɔnhwerew bi a team a ɛdi hɔ no renkɔyɛ bio fi mfiase.

Nsonsonoe a ɛda nnɔnhwerew 680,000 ne 16 ntam no ma asɛmfua scarcity nya ntease foforo. Kasa bi betumi ayɛ den wɔ nkɔmmɔ, gua so, fie, radio, afahyɛ ne abusua mu na, bere koro no ara, machine no ahu no sɛ data kakraa bi na ɛwɔ hɔ. Kasa a ɛwɔ asetena mu ne data a computer tumi nya no nyɛ ade koro.

Sɛ voice bɛdan training material a, ɛsɛ sɛ wɔrecord no. Afei ebia ɛho behia transcription, labelling, classification, storage ne permission a wɔde bɛsan adi dwuma. Nea ennam saa kwan no so no pii mmra language ho statistical representation no mu.

Awia akyi nkɔmmɔ no de ade foforo kaa ho. Presentation bi a ɛfa sɛnea wɔsi technology company wɔ Ghana ho no mu no, wɔkaa sɛ data pii a AI de sua fi internet so, na wɔkaa internet no ho sɛ English na ɛdɔɔso wom na West nso na ɛdɔɔso wɔ mu. Ɛno akyi pɛɛ, ɔkasafo no kɔɔ everyday multilingualism so. Onipa betumi de kasa ahorow pii adi dwuma da koro mu, na ɔsesa ne kasa ne ne tone nso sɛnea beae a ɔwɔ ne onipa a ɔne no rekasa te. Sɛ wɔdan saa nkitahodi no nyinaa kɔ English mu na wɔsan dan no ba kasa foforo mu a, ebia information no bi bɛtena hɔ, nanso nea na ɛma nkɔmmɔ no nya social meaning no betumi ayera.

Joke bi renyɛ joke koro no ara bio. Sɛnea wokasa kyerɛ obi no betumi ahwere abusua mu abusuabɔ, ntam, obu, irony anaa intimacy. Kasa nyɛ ade a wɔde meaning hyɛ mu na wɔtumi sesa no ara kwa.

Data nso wɔ context

Saa ɔhaw yi ka anthropology of artificial intelligence no mu ade titiriw bi. Large models tumi hu regularities yiye paa. Nanso nnipa nkasa denam grammar regularities nkutoo so.

Wɔsesa register. Wɔsesa kasa. Wɔtwa sentence bi tiawa efisɛ onipa foforo no nim abakɔsɛm a ɛwɔ akyi. Wɔde expression bi di dwuma a ne ntease gyina kurow fa, awo ntoatoaso, abusua ne osuahu a wɔakyɛ so. Wɔkasa ɔkwan foforo wɔ adwumam, fie anaa bere a wɔretɔ ade wɔ gua so.

Sɛ wɔdan saa dwumadi no nyinaa kɔ normalised text mu a, social structure a na ɛka nsɛmfua no ho no bi betumi ayera.

Ɛno nti, sɛ yɛma kasa dodow a system bi tumi de di dwuma no yɛ bebree a, ɛno nko ara nni ɔhaw no so. Interface betumi de kasa ɔha ama, nanso ɛda so ara asusuw sɛ usage no bɛyɛ ɔkwan koro. Onipa no kyerɛw, machine no yɛ processing, na text foforo san ba. Multilingualism bɛyɛ system no property, nanso ebia ɛrenyɛ property a ɛfa asetena a system no pɛ sɛ ɛkyerɛ no ho.

Accra kyerɛɛ kwan foforo.

Offline ansa na “universal”

Hack-AI-Thon no mu no, Haska AI de application bi bae a wɔayɛ no sɛ ɛbɛka wɔn a wɔwɔ cocoa-industry waste ne wɔn a wobetumi de adi dwuma anaa atɔ no bom. Presentation no mu no, wɔkaa sɛ models no yɛ adwuma fully offline, na application no wɔ local languages pii mu. Nkyerɛkyerɛ no yɛ practical koraa. Onipa a ɔwɔ baabi a ɛwɔ akyirikyiri no nsɛ sɛ ɔhia internet anaasɛ wɔhyɛ no sɛ ɔmfa English mode ansa na watumi de tool no adi dwuma.

Ɛnyɛ saa project no nkutoo na offline yɛ design no fã. Proposal foforo a ɛfa medical laboratories ho nso kae sɛ wɔn models betumi ayɛ adwuma a internet nni hɔ, titiriw wɔ facilities a resources sua mu. Saa ade no san bae mpɛn pii. Offline annyɛ feature ketewa bi a wɔde kaa product no ho akyiri. Ɛyɛ design conditions no mu ade.

Saa nsakrae no ho hia. Technology bi betumi aka sɛ ɛyɛ universal efisɛ ɛyɛ adwuma wɔ browser biara mu, nanso bere koro no ara, ebia egyina tebea ahorow a ɛnyɛ universal koraa so.

Judges no nkɔmmɔ no mu no, wɔde criterion no too gua wɔ ɔkwan a ɛbɛn anthropology ho. Bere a wɔrehwɛ proposals no, wobisaa sɛ solution no betumi akɔ nnipa asetena mu a wɔrenhia sɛ wɔsesa wɔn da biara practices nyinaa na ama wɔatumi de AI adi dwuma anaa. Wɔde mobile money ne text messages yɛɛ nhwɛso. Technology bi yɛ mmerɛw sɛ nnipa begye atom sɛ ɛfa behaviour bi a wɔn asua dedaw so ba.

Nsonsonoe no betumi ayɛ ketewa, nanso ɛdannan kwan a yɛtaa fa adaptation ho. Sɛ́ anka yɛbɛka akyerɛ nnipa sɛ wonsua sɛnea artificial intelligence pɛ sɛ wɔde no di dwuma no, system no fi ase sua sɛnea nnipa a wɔpɛ sɛ wɔde ma wɔn no tena ase dedaw.

Machine a etie

Mpɛn pii no, sɛ wɔka African AI ho a, wɔka local models, data centres, computing capacity, investment ne technological sovereignty ho. Saa nneɛma yi nyinaa puee Accra, na ne nyinaa ho bɛhia. Nanso nkɔmmɔ a ɛkɔɔ so da a ɛto so abien no mu no, ade bi a ɛbɛn da biara asetena nso puei.

Hena na otumi kasa kyerɛ machine.

Kasa bɛn mu.

Ɔkwan bɛn so.

Internet tebea bɛn mu.

Na meaning no dodow ahe na ɛka hɔ bere a voice no adan data.

Sɛ yɛyɛ voice-first systems a wobetumi ayɛ adwuma offline na wɔayɛ no ama multilingual contexts a, ɛnkyerɛ sɛ yɛreyɛ dominant interfaces no version ketewa. Ebetumi akyerɛ sɛ yɛrefi reality ho nkyerɛkyerɛ foforo mu.

Artificial intelligence a yenim no asua sɛ ɛbɛtwɛn written prompt. Mmeae pii no, sɛ ɛbɛsua tie no, ebetumi ayɛ innovation a emu dɔ sen saa koraa.

Nkyerɛase ne nsɛm a wɔde gyinaa so

Temas