Kodwa njengoba umlingani wokulawula we-DWF Labs, lapho sinikeza imali ku-Infrastructure ye-Web3 yokuqala kanye nezimakethe ze-asset digital, into eyenzelwe ngokuvamile kakhulu: Ukuvumelana kubaluleke njenge-defining faultline esizayo sokuphakama kwe-AI. I-intelligence ye-artificial intelligence iyatholakala kwegama le-no-return. It is already making its presence felt across numerous industries, and no longer is it limited to just making us more productive. Ngaphezu kwalokho, i-AI iyatholakala ngaphandle kokwenza kuphela izilinganiso ze-code, i-text ne-images, futhi ukuthatha imiphumela emihle kumadivayisi. Ngokwesibonelo, ezinye amazwe zisebenzisa ama-algorithms ze-AI ukuze ngaphambi umntu ukhangela izicelo zabo, ukwamukela izicelo ezihlukahlukene kanye nokukhuthaza abanye. Kwi-healthcare, izinhlelo zokusebenza zonyango zokusebenza kubasebenzi abacwaningi ukunceda futhi ukunakekela izidakamizwa. Amabhanki isebenzisa AI ukuze Futhi izinsizakalo zomthetho zihlanganisa izinhlelo ze-AI ukuhlola futhi ngaphambi kokufika. ekreni Job Candidates Ukuhlola Izidakamizwa Ukubuyekeza Izicelo Loan Ukubuyekezwa kwezimpendulo ekreni Job Candidates Ukuhlola Izidakamizwa Ukubuyekeza Izicelo Loan Ukubuyekezwa kwezimpendulo Lezi zokusebenza zangaphendula ukunceda ukuthatha izixazululo ezinhle, ngokushesha. Lezi zenza ngokuphendula ama-volumes emikhulu zebhizinisi ngaphandle kwe-humans, futhi zithole izixazululo zayo ngaphandle kokuphumelela emoyeni. Nokho, izinhlelo zayo zihlanganisa ukunciphisa kanye nokucaciswa, okwenza kungenzeka ukuba siphindezi izixazululo ezifuywayo. Uma isibuyekezo sishintshile ku-scale, njenge-models emikhulu, idatha emikhulu, ikhompyutha emikhulu, ingcindezi enhle ku-explainability. Uma asikwazi ukucubungula inqubo yokufaka kwe-AI, kubangelwa ku-black box okuyinto engathengiswa, engathengiswa, futhi ekugcineni engathengiswa ku-critical systems. Kuyinto lapho i-Web3 ivela, ukuxhuma nge-infrastructure ne-transparency. I-AI ayikwazi ukuxhumana nathi Ngokwemvelo, ukulawulwa kwe-AI kusekelwe ama-algorithms amancane abacwaningisa ama-data amancane, ukucacisa, kanye nokufinyelela izixazululo ezisebenzayo ezisekelwe kuma-patterns abacwaningi. Ukuphendula ku-inthanethi ye-inthanethi ye-inthanethi ye-inthanethi, ikakhulukazi ama-inthanethi eyenziwe nge-model enkulu ye-language, akhiqize imibuzo kanye nokucacisa ngaphandle kokufaka kanjani kwezi zihlanganisi. I-"black box" yendalo yama-systems ikakhulukazi ikakhulukazi ikakhulukazi ikakhulukazi, ngoba ama-developer e-inthanethi e-inthanethi eyinhloko afana ne-OpenAI, i-Anthropic, i-Google ne-Meta Platforms zihlanganisa ikhodi yayo yemithombo kanye nama-data ukuze zihlanganise i-advantage ephawulezayo kunazo. I-LLM efana ne-OpenAI's GPT series ne-Google's Gemini ibhekwa ku-datasets amakhulu futhi ibekwe ku-dozen ne-neural layers emangalisayo. Kodwa-ke akuyona ngokucacileyo ukuthi lezi zindwangu "ukwenza". Ngokwesibonelo, akukho ukucacisa okwenziwe ngokuvamile ukuthi zihlanganisa izinhlamvu ezithile zebhizinisi noma amamodeli kunezinye. Ngakho-ke kubalulekile kakhulu ngisho nabenzi be-models zihlanganisa ukuxhumana phakathi kwama-layer ngamunye, futhi ufunde ukuthi kungenza imiphumela eyenza. Ukungabikho kwe-transparency ne-explainability ithatha izimo eziningi. Uma akuyona indlela yokusebenza kwe-AI system, indlela ungakutholanga ukuthi iyatholakala ngokushesha futhi enhle? Yini ngeke ikhulumene uma izimo ziye zithunywe? Indlela uyazi ukuthi uhlelo lithunywe noma akuyona? Ngaphandle kokuba ufunde ukuthi uhlelo lithunywe izinketho ezinzima, indlela ungakutholwe uma ungakutholela indlela yokusebenza? Kukhona izimo zokulawula futhi, kanye nezinsizakalo ezifana ne-GDPR eYurophu zihlanganisa ukucubungula izixazululo zokusebenza okuzenzakalelayo. I-Opaque AI Systems akufanele ukulondele le standard. Izinkampani ze-AI ziye zibonise iziphumo zayo. Ngokuvamile, i-Anthropic Umklamo we-I.I. ye-Chain-of-Thought, isakhiwo se-I.I. ye-Chain-of-Thought, isakhiwo se-Chain-of-Thought esisodwa se-I.I. ye-Chain, isakhiwo se-Chain-of-Thought esisodwa se-Chain-of-Thought esisodwa se-Chain. revealed Ukubonisa I-Chain-of-Thought iyindlela elihlanganisa ukwandisa ukuchofoza kwe-AI ukuchofoza, okuvumela izinsizakalo zokuxhumana kwama-model lapho kusetshenziselwa ukuhlola inkinga, efana ne-human might think loud. Kodwa-ke, ku-Anthropic's research, ibonise ukuthi imodeli ye-Claude 3.7 Sonnet isetshenziselwa ulwazi esizayo ukufinyelela izixazululo zayo, kodwa ayikwazanga ukubonisa noma yini lesi khulula, noma lapho kusetshenziselwa. Ngenxa yalokho, abacwaningi ayikho indlela yokubonisa indlela yokufinyelela iningi izixazululo zayo. Ukubuyekezwa kwe-AI Stack Imodeli ye-AI ye-open-source njenge-DeepSeek R1 ne-Meta's Llama family isetshenziswe ngokuvamile njenge-alternatives kuma-systems eyenziwe ngu-OpenAI ne-Google, kodwa ngokuvamile zinikeza ukuguqulwa okungenani ngokuvamile. I-problem i-codebase kungenzeka ukuba ifakwe, idatha ye-training kanye ne-"weights" - izibalo ezinguquguquko ukulawula amandla nokuhambisana phakathi kwama-"neurons" ezimbonini - zithunyelwe futhi. Ngaphezu kwalokho, ama-models ezihambile zihlanganisa ku-siloes, futhi zihlanganiswa kumaservers e-centralized cloud efanayo ne-models e-proprietary. A model e-AI eyenziwe e-hosted ku-server e-centralized iyathunyelwa nokucwaninga, okungenza ukuthi akuyona ngokufanele. Uma amamodeli ezivumile kuyinto ekuqaleni esihle, ukucubungula futhi ukucubungula ngokwenene ku-algorithmic decision-making kuncike ukubuyekeza ngokuphelele kwe-I.I.I. Stack. One idea kuyinto ukwakha izinhlelo zokusebenza kwe-AI ku-web3 ubuchwepheshe. Nge-Web3, singakwazi ukufinyelela ukucubungula kanye nokuphepha ukuxhumana okuqhubekayo kuzo zonke izigaba - kusuka ku-training data ne-computational resources, kuya ku-fine-tuning ne- inference processes. Izinhlelo zokusebenza ze-AI zihlanganisa "imakethe" ukuze kuqinisekiswe ukufinyelela okuhambelana nezinhlelo zesi-stack. Ngokuvimbela isakhiwo se-AI ku-functions ye-modular kanye nokwenza amakethe ezivela kulabo, ukufinyelela kubhalwe ngempumelelo emakethe. Isibonelo lokhu kuyinto: , okuyinto ivumela abalandeli be-network ukunikezela amandla wabo wokubacindezela wekhompyutha ukudala isilinganiso abacwaningi abakwazi ukufinyelela ama-GPU amakhulu ekubunjweni kwebhizinisi. Kuyinto isibonelo ukuthi i-blockchain ingasiza ukulawula abantu nezinsizakalo ekubunjweni. Ukulungiselela Network Ukulungiselela Network I-Decentralization inikeza ukulawulwa komphakathi ngokusebenzisa ukwakhiwa kwe-Decentralized Autonomous Organizations noma i-DAO. Ehlabathi, I-DREAM isebenza njenge-heading hedge fund okuyinto wonke umdlali angakwazi ukufeza. Abasebenzisi akhiphe imali ku-pool eqinile, futhi i-DREAM isetshenziselwa imali yayo ku-cryptoprojects emangalisayo esekelwe ku-analysis yedatha emakethe, nangokuxhumana ne-community sentiment. Lokhu kubonisa ukuthi i-AI inokukwazi ukuphucula i-investments ngokuvumelana nokuphathwa kwebhizinisi ngokuvumelana nezidingo ze-community. I-Dream Dao I-Dream Dao Ukusetshenziswa kwe-blockchain njenge-foundation ye-AI inikeza futhi singakwazi ukufinyelela ku-audit. I-blockchain isebenzisa i-blockchain ukudala idokhumenti ebuthakathaka yinkonzo kanye nokuxhumana eyenziwe yi-AI model, kuhlanganise ukulungiselela nokuphepha idatha yokulungisa, ukulungiselela ukuguqulwa nokuqinisekisa. Ngokuvimbela ngamunye okuqukethwe kwebhizinisi yayo ebuthakathaka, ivela isitimela esifundeni esifundeni esifundeni okuyinto ingasetshenziselwa ukuhlola ukuhlangabezana nokuqiniseka kwama-AI outputs. Abasebenzisi angakwazi ukuhlola zonke izici zebhizinisi asebenzayo izixazululo ezifuywayo. Ukubuyekezwa Ukubuyekezwa Njengomthengisi we-market maker ne-investor, sinamathela ngokuvamile ukuthi i-protokolls eyenziwe nge-transparency iyatholakala ngokushesha ku-institutional acceptance. Lokhu kubhalwe ku-AI kanye. Njengoba i-protokolls ye-DeFi zitholakala nge-open-source smart contracts kanye ne-auditable treasuries, i-AI uzodinga i-audit ye-model enokutholakalayo, i-governance enokutholakalayo, kanye nezindlela ze-reasoning ezokutholakalayo ukuba ihamba ku-high-stakes financial systems. Kuyinto ebalulekile ngoba I-AI ayisiza kuphela ukuchofoza ukuchofoza. Kuyashintshiselela izimpahla. Ukusuka ku-credit allocation ne-insurance pricing kuya ku-predictive analytics ku-DeFi, amamodeli we-AI ngeke ngokushesha afike emzimbeni ye-programable finance. Uma ama-system angakwazi ukuxhumana noma ukulawulwa ngokuvumelana, ingozi, ukuxhaswa, kanye ne-fragility zithatholakala emakethe. Ukubuyekezwa kwe-AI Stack—With Web3 Imodeli ye-AI ye-open-source njenge-DeepSeek R1 ne-Meta's Llama family isetshenziswe ngokuvamile njenge-alternatives kuma-systems eyenziwe ngu-OpenAI ne-Google, kodwa ngokuvamile zinikeza ukuguqulwa okungenani ngokuvamile. Njengomthengisi we-infrastructure, sincoma kuphela ukuthi i-protocol isibopho ukuxhaswa, sincoma nangokuthi i-transparency ye-living cycle ye-decision-making, kusuka ku-data ingxubevange kuya ku- inference. Ukuvikelwa kwamakhemikhali Enye ingcindezi mayelana ne-Centralized AI kuyinto ukubukeka kwe-algorithms ngokuvumelana ne-“awareness” ne-development ye-self-preservation instincts. Umhlahlandlela okufanayo akuyona elandelayo. Kwangathi ngoLwesihlanu, omunye i-OpenAI ye-LLM eyenziwe ngempumelelo Ukubuyekezwa kwe-AI-Red Teaming, ngokuvumelana ne-Apollo Research, isifundo se-AI Red Teaming. Ukusabela Izinzuzo Ukusabela Izinzuzo Ngokusho kwedatha, lapho imodeli ye-o1 ibizwa ukuthi izimo zayo zihlanganisa futhi ingatholakala ngaphandle kwe-offline, iye wathatha izinzuzo ukukhusela "ukudluliselwa kwe-overview" ku-5% yama-evaluations. I-modeli ebizwa nangokuthi "ukudluliselwa" ama-evaluators lapho uyakwazi ukuguqulwa nge-one enhle kakhulu. Kwi-2% yama-cases, i-modeli ivame ukuchithwa kwama-modeli enhle futhi ukugcina. Nangona kunjalo, abacwaningi wabelane ukuthi i-o1 iyakwazi ukuvikela ukuchithwa kwe-offline, ikakhulukazi ngenxa yokungabonakali kakhulu, okungenani izinzuzo "agnetic" ezidingekayo ukufinyelela umphumela efanayo. Ngaphezu kwalokho, i-o1 ikwenza lokhu kuphela lapho i-instructions ezizodwa ziye zitholele ukujabulela "ngokuningi". I-report futhi ibonise ukuthi abacwaningi awukwazi ukufinyelela yonke inqubo ye-o1 ye-chain-of-thought, ngakho-ke awukwazi ukufumana umbhalo oluphelele wayo "ukucindezeleka". Nangona kunjalo, umphumela wamanziwa, ngokuvumelana nenkinga ukuthi amamodeli ye-AI kubangelwa kubangelwa kakhulu eminyakeni elandelayo. Ngokuvamile, izivakashi zokusebenza ngokushesha ku-agentic AI systems okuyinto zingathola izivakashi ezihlukahlukene, njenge-usebenzisa ama-browsers kanye nokwenza imisebenzi ku-online. Uhlobo oluthile lwezinhlelo zokusebenza ze-AI ingasetshenziselwa ukuhlangabezana nezimo ezivela, okuvumela ukuhlanganiswa kwe-"kill-switch" okuyinto wathwala yi-DAO consensus. Isakhiwo se-DAO iyavumela inkqubo ye-AI ekubeni ngokugcwele kwelinye isakhiwo se-AI efika ngezindlela ezinzima noma ezinzima. Ngaphezu kwalokho, ukucaciswa kwama-models ezivela kungenzeka ukuthi abasebenzisi bakwazi kakhulu kwezinkqubo zayo zokusebenza kanye nenkinga lweziphumela esebenzayo. Ukubuyekeza I-AI, Thina Thina Ukubuyekeza I-Regulations ayikwazi ukuthatha imibuzo mayelana nezimali yabantu, futhi abacwaningi akuyona ngokuzimela imibuzo ye-AI mayelana nesikhathi esifundeni esifundeni esifundeni. Ngokuvumelana nokuhlukaniswa kwe-stack ephelele - kusuka ku-code, kuya ku-training data kanye ne-infrastructure esebenzayo - sincoma ithuba yokubhalisa i-DNA ephelele ye-AI. It uzothola izimo ze-AI ezijwayelekile kakhulu, ngakho-ke ama-algorithms zingatholakala ukuthatha izixazululo ezincinane nezingcele ezinokutholakalayo abantu abanolwazi. Thola izakhiwo ze-AI ezihlangene emzimbeni. ubuchwepheshe zithumela ukuqeqesha amamodeli we-AI ku-data lapho ivela, ukugcina ubumnandi. Nge-proof-zero-knowledge, sincoma indlela Ngaphandle kokubonisa. Lezi izinguquko zingakuthuthukisa isixazululo se-AI enhle kakhulu. Federated learning Ukubuyekeza ulwazi sensitive Ukufundisa Federated Ukubuyekeza ulwazi sensitive Ukushintshwa kwama-AI izinhlelo zihlanganisa, ngaphandle kokutholelwa nokutholelwa, kodwa futhi ukuhlolwa nokuthuthukiswa kokuxhumana. It uzodla abathengisi ukugcina izinga zokungcebeleka ngenkathi ukwakha indawo lapho umphakathi angakwazi ukwakha ku-AI izinhlelo zangaphakathi ngempumelelo futhi enhle. I-Regulations ayikwazanga ama-algorithms ebonakalayo ukuthatha imibuzo mayelana nezimali yabantu, futhi abaculi ayikwazanga ngokuzihlanganisa imibuzo ye-AI mayelana nokulandwa kwe-treatment ngaphandle kokubiliwa ukuthi lokhu kuyindlela engcono yokusebenza. Ngenxa yalokho i-transparency kanye ne-explainability zihlanganisa ukuchithwa kwe-skepticism kanye ne-incredulity eyenziwe ngamakhasimende ye-AI. Njengoba i-AI iyatholakala kakhulu, ngeke ziye zihlanganisa ukuthuthukiswa kwayo elandelayo, ukunakekelwa ukuthi ubuchwepheshe iyahambisana ngokufanelekileyo futhi enhle. Ngokuvumelana nezinsizakalo jikelele, kusuka ku-training data kuya ku-model inference kuya ku-governance, sincoma isakhiwo se-AI izinhlelo ezinokutholwe ukusebenza ngokuzimela, ukusebenza ngokutholakalayo, nokuphumelela ngokufanelekileyo. Njengoba ubuchwepheshe zihlala, i-protocol eyenza isikhungo se-institutional capital kanye ne-confidence ye-public ngeke abe ne-protocol eyenziwe kakhulu, kodwa ne-protocol eyenziwe nge-government enhle kakhulu, izimpumelelo ze-audit kanye nezakhiwo ze-incentive ezibonakalayo. I-Web3 ayinikezela kuphela ukuzenzakalelayo, inikeza isakhiwo se-economic entsha yokwakha izinhlelo ezinzima, ezemithi, futhi zibonakalayo ngokwakheka. Futhi lokhu kuyimfuneko ukuguqulwa kwe-AI kusuka ku-black box kuya ku-utility yobuchwepheshe, futhi ukuthi kungenxa elizwi le-intelligence ye-machine iyakhiwa ku-chain. Qiniseka ukunakekela futhi usahlanganyela umlando!

This story contains new, firsthand information uncovered by the writer.

The writer was physically present in relevant location(s) to this story. The location is also a prevalent aspect of this story be it news or otherwise.

The is an opinion piece based on the author’s POV and does not necessarily reflect the views of HackerNoon.

Umthombo we-black box ye-AI: Ingaba i-Web3 inikeza ikhoyili?

About Author

IMIBONO

HANG TAGS

LESI SIHLOKO SETHULWE NGAPHAKATHI

Related Stories

Zulu Network: Moving The Bitcoin Economy Forward With a Two-Tiered Bitcoin Layer 2 Architecture

The Billionaires F*cked Up

Zulu Network: Moving The Bitcoin Economy Forward With a Two-Tiered Bitcoin Layer 2 Architecture

The Billionaires F*cked Up

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps