I-HodlX Guest Post  Thumela Iposi Lakho

 

I-AI (i-artificial intelligence) ibangele ukujabula okukhulu phakathi kwabathengi namabhizinisi ngokufanayo - eqhutshwa inkolelo eshisekayo yokuthi ama-LLM (amamodeli ezilimi ezinkulu) namathuluzi afana ne-ChatGPT azoguqula indlela esifunda ngayo, esisebenza ngayo nesiphila ngayo.

Kodwa njengasezinsukwini zakuqala ze-inthanethi, abasebenzisi bangena ngaphandle kokucabangela ukuthi idatha yabo yomuntu siqu isetshenziswa kanjani - anomthelela lokhu okungaba nawo kubumfihlo babo.

There have already been countless examples of data breaches within the AI space. In March 2023, OpenAI temporarily took ChatGPT okungaxhunyiwe ku-intanethi after a ‘significant’ error meant users were able to see the conversation histories of strangers.

Leso siphazamisi sisho ulwazi lokukhokha lwababhalisile - okuhlanganisa amagama, amakheli e-imeyili nengxenye yezinombolo zekhadi lesikweletu - nazo zazisesidlangalaleni.

In September 2023, a staggering 38 terabytes of Microsoft data was inadvertently ligcwele by an employee, with cybersecurity experts warning this could have allowed attackers to infiltrate AI models with malicious code.

Researchers have also been able to Sebenzisa AI systems into disclosing confidential records.

Emahoreni ambalwa nje, iqembu elibizwa ngokuthi i-Robust Intelligence likwazile ukucela ulwazi olungamhlonza umuntu ku-software ye-Nvidia futhi ladlula izivikelo eziklanyelwe ukuvimbela isistimu ukuthi ingaxoxi ngezihloko ezithile.

Izifundo zafundwa kuzo zonke lezi zimo, kodwa ukwephulwa ngakunye kubonisa ngamandla izinselelo okudingeka zinqotshwe ukuze i-AI ibe amandla athembekile futhi athembekile ezimpilweni zethu.

I-Gemini, i-chatbot ye-Google, ivuma nokuthi zonke izingxoxo zicutshungulwa ababuyekezi abangabantu - egcizelela ukungabi sobala ohlelweni lwayo.

“Ungafaki noma yini ongeke uthande ukuthi ibuyekezwe noma isetshenziswe,” kusho isexwayiso esiya kubasebenzisi.

I-AI ishesha ukudlula ithuluzi abafundi abalisebenzisela umsebenzi wabo wesikole noma izivakashi ezithembele kulo ukuze bathole izincomo ngesikhathi sohambo oluya e-Rome.

Kuya ngokuya kuncika kuzo ezingxoxweni ezibucayi - futhi sinikeze yonke into kusukela kumibuzo yezokwelapha kuya ezinhlelweni zethu zokusebenza.

Ngenxa yalokhu, kubalulekile ukuhlehla futhi ucabange ngezinkinga ezintathu eziphezulu zobumfihlo bedatha ezibhekene ne-AI namuhla, nokuthi kungani zibalulekile kithi sonke.

1. Ukwaziswa akuyimfihlo

Amathuluzi afana ne-ChatGPT abamba ngekhanda izingxoxo ezedlule ukuze ubuyele kuzo kamuva. Nakuba lokhu kungathuthukisa ulwazi lomsebenzisi futhi kusize ukuqeqesha ama-LLM, kuza nengcuphe.

Uma uhlelo lugetshengwa ngempumelelo, kunengozi yangempela yokuthi ukwaziswa kudalulwe kunkundla yomphakathi.

Imininingwane engaba amahloni evela kumlando womsebenzisi ingase iputshuzwe, kanye nolwazi olubucayi kwezohwebo lapho i-AI isetshenziselwa izinjongo zomsebenzi.

Njengoba sibonile ku-Google, konke okuthunyelwe kungagcina sekucutshungulwa ithimba lakhona lokuthuthukisa.

Samsung took action on this in May 2023 when it banned employees from using generative AI tools altogether. That came after an employee alayishiwe confidential source code to ChatGPT.

Umdondoshiya wezobuchwepheshe ubekhathazeke ngokuthi lolu lwazi kuzoba nzima ukulithola nokuyisusa, okusho ukuthi i-IP (impahla yengqondo) ingase igcine isatshalaliswa emphakathini wonkana.

I-Apple, i-Verizon kanye ne-JPMorgan bathathe isinyathelo esifanayo, ngemibiko ephakamisa ukuthi i-Amazon yethule ukwehla ngemuva kwezimpendulo ezivela ku-ChatGPT zifana nedatha yayo yangaphakathi.

Njengoba ubona, ukukhathazeka kudlulela ngale kwalokho okuzokwenzeka uma kukhona ukwephulwa kwedatha kodwa ethembeni lokuthi ulwazi olufakwe ezinhlelweni ze-AI lungahlehliswa kabusha futhi lusatshalaliswe kubabukeli abaningi.

Companies like OpenAI are already ebhekene nazo multiple lawsuits amid allegations that their chatbots were trained using copyrighted material.

2. Amamodeli e-AI angokwezifiso aqeqeshwe izinhlangano awawona angasese

Lokhu kusiletha ngobunono ephuzwini lethu elilandelayo - kuyilapho abantu ngabanye nezinkampani bengakwazi ukusungula amamodeli abo e-LLM angokwezifiso ngokusekelwe emithonjeni yabo yedatha, ngeke babe yimfihlo ngokugcwele uma bekhona ngaphakathi kwengxenyekazi efana ne-ChatGPT.

Ekugcineni ayikho indlela yokwazi ukuthi okokufaka kuyasetshenziswa yini ukuqeqesha lezi zinhlelo ezinkulu - noma ulwazi lomuntu siqu lungagcina lusetshenziswa kumamodeli azayo.

Njenge-jigsaw, amaphuzu edatha avela emithonjeni eminingi angahlanganiswa ukuze kwakheke ukuqonda okuphelele nokunemininingwane ekhathazayo mayelana nokuthi ungubani kanye nesizinda sakhe.

Izinkundla ezinkulu zingase futhi zehluleke ukunikeza izincazelo ezinemininingwane yokuthi le datha igcinwa futhi icutshungulwe kanjani, ngokungakwazi ukuphuma ezicini umsebenzisi angakhululekile ngazo.

Ngaphandle kokuphendula ukwaziswa komsebenzisi, amasistimu e-AI aya ngokuya enekhono lokufunda phakathi kwemigqa futhi athole yonke into kusukela endaweni yomuntu kuya kubuntu bakhe.

Esimeni sokuphulwa kwedatha, imiphumela emibi ingenzeka. Ukuhlasela kobugebengu bokweba imininingwane ebucayi okuyinkimbinkimbi ngendlela emangalisayo kungase kuhlelwe - kanye nabasebenzisi abaqondiswe ngolwazi abalufunze ngokuyimfihlo ohlelweni lwe-AI.

Ezinye izimo ezingaba khona zihlanganisa le datha esetshenziselwa ukucabanga ukuthi ungubani, kungakhathaliseki ukuthi lokho kungenxa yezinhlelo zokusebenza zokuvula ama-akhawunti asebhange noma amavidiyo ajulile.

Abathengi badinga ukuhlala beqaphile ngisho noma bengayisebenzisi i-AI ngokwabo. I-AI iya ngokuya isetshenziswa ukunika amandla amasistimu okuqapha kanye nokuthuthukisa ubuchwepheshe bokubona ubuso ezindaweni zomphakathi.

Uma ingqalasizinda enjalo ingasungulwa endaweni eyimfihlo ngempela, inkululeko yomphakathi kanye nezimfihlo zezakhamizi ezingenakubalwa zingaphulwa ngaphandle kokwazi kwazo.

3. Idatha yangasese isetshenziselwa ukuqeqesha amasistimu e-AI

Kunokukhathazeka ngokuthi amasistimu amakhulu e-AI aqoqe ubuhlakani bawo ngokuhlola inqwaba yamakhasi ewebhu.

Izilinganiso ziphakamisa ukuthi amagama ayizigidi eziyizinkulungwane ezingu-300 asetshenziselwe ukuqeqesha i-ChatGPT - lokho kungu-570 gigabytes wedatha - ngezincwadi nokufakiwe kweWikipedia phakathi kwamasethi edatha.

Ama-algorithms nawo aziwa ngokuthi ancike emakhasini ezinkundla zokuxhumana kanye namazwana aku-inthanethi.

Ngeminye yale mithombo, ungaphikisa ngokuthi abanikazi balolu lwazi bebelindele ubumfihlo obuphusile.

Kodwa nansi into - amathuluzi amaningi nezinhlelo zokusebenza esisebenzisana nazo nsuku zonke sezivele zithonywe kakhulu yi-AI - futhi sisabele ekuziphatheni kwethu.

I-Face ID ku-iPhone yakho isebenzisa i-AI ukulandelela izinguquko ezicashile ekubukekeni kwakho.

I-TikTok kanye ne-Facebook's AI-powered algorithms yenza izincomo zokuqukethwe ngokusekelwe kuziqeshana nokuthunyelwe owake wakubuka esikhathini esidlule.

Abasizi bezwi abafana no-Alexa no-Siri bancike kakhulu ekufundeni komshini, nabo.

Umlaza onesiyezi wokuqalwa kwe-AI ungaphandle, futhi ngayinye inenjongo ethile. Nokho, ezinye zisobala kakhulu kunezinye mayelana nokuthi idatha yomsebenzisi iqoqwa, igcinwe futhi isetshenziswe kanjani.

Lokhu kubaluleke kakhulu njengoba i-AI yenza umthelela emkhakheni wezokunakekelwa kwempilo - kusukela ekucabangeni kwezokwelapha kanye nokuxilongwa kuya ekugcineni amarekhodi kanye nemithi.

Izifundo zidinga ukufundwa emabhizinisini e-inthanethi abanjwe amahlazo obumfihlo eminyakeni yakamuva.

Flo, a women’s health app, was umsolwa by regulators of sharing intimate details about its users to the likes of Facebook and Google in the 2010s.

Siyaphi sisuka lapha

I-AI izoba nomthelela ongacimeki kuzo zonke izimpilo zethu eminyakeni ezayo. Ama-LLM aba ngcono nsuku zonke, futhi amacala amasha okusebenzisa ayaqhubeka nokuvela.

Kodwa-ke, kunengozi yangempela yokuthi abalawuli bazozabalaza ukuhambisana njengoba imboni ihamba ngesivinini esikhulu.

Futhi lokho kusho ukuthi abathengi badinga ukuqala ukuvikela idatha yabo futhi baqaphe ukuthi isetshenziswa kanjani.

Ukwahlukaniswa kwezindawo kungadlala indima ebalulekile lapha futhi kuvimbele amanani amakhulu edatha ekuweleni ezandleni zezinkundla ezinkulu.

Ama-DePIN (amanethiwekhi engqalasizinda amisiwe) anamandla okuqinisekisa ukuthi abasebenzisi bansuku zonke bathola izinzuzo ezigcwele ze-AI ngaphandle kokuthi ubumfihlo babo bubekwe engcupheni.

Ukwaziswa okubethelwe akukwazi kuphela ukuletha imiphumela yomuntu siqu kakhulu, kodwa ama-LLM agcina ubumfihlo azoqinisekisa ukuthi abasebenzisi banokulawula okugcwele kwedatha yabo ngaso sonke isikhathi. - kanye nokuvikelwa ekusetshenzisweni kabi.

---

Chris Were is a CEO of I-Verida, a decentralized, self-sovereign data network empowering individuals to control their digital identity and personal data. Chris is an Australian-based technology entrepreneur who has spent more than 20 years devoted to developing innovative software solutions.

 

Isithombe Esenziwe: Midjourney

• I-SEO Powered Content & PR Distribution. Khuliswa Namuhla.
• I-PlatoData.Network Vertical Generative Ai. Zinike Amandla. Finyelela Lapha.
• I-PlatoAiStream. I-Web3 Intelligence. Ulwazi Lukhulisiwe. Finyelela Lapha.
• I-PlatoESG. Ikhabhoni, I-CleanTech, Amandla, Environment, Ilanga, Ukuphathwa Kwemfucuza. Finyelela Lapha.
• I-PlatoHealth. I-Biotech kanye ne-Clinical Trials Intelligence. Finyelela Lapha.
• Source: https://dailyhodl.com/2024/03/28/top-three-data-privacy-issues-facing-ai-today/