Imidiya yezobuchwepheshe i-scientificamerican yashicilela ibhulogi ngo-Agasti 6, ibika ukuthi izazi zezibalo eziningana zisola i-OpenAI ngokungaziphathi kahle kwezemfundo, zithi imiphumela yezibalo yemodeli yayo ye-AI i-Astra ayizange inikeze abacwaningi bangaphambilini udumo nokubonga okufanele.
I-OpenAI yashicilela ifayela le-PDF elinamakhasi angu-249 ngoJulayi 31, lichaza ngokuningiliziwe intuthuko enkulu yemodeli yayo elandelayo ebalulekile i-Astra (inguqulo yangaphakathi) emikhakheni eminingi, okuhlanganisa ukupakishwa kwamabhola ezindaweni ezinobukhulu obuphezulu, ithiyori yemidlalo ye-quantum, ithiyori yamaqembu kanye ne-combinatorics eyeqisayo.
Uma kubalwa ngamanani e-Sol API, izindleko eziphelele zamathokheni adingekayo ukuze kutholakale izixazululo zalezi zinkinga zilinganiselwa ku-2,000 USD (Qaphela: ngezinga lokushintshisana lamanje, cishe lilingana no-13,519 yuan waseChina).
Nokho, ngemva kokushicilelwa kwephepha, izazi eziningana zasola inkampani i-OpenAI ngokungaziphathi kahle kwezemfundo ngemva kokulibuyekeza.
Imiphumela emibili eyinhloko ayizange ikhombe ngokufanele imibono eyayivele ikhona emibhalweni yakamuva yezibalo. Isazi sezibalo sase-Yeshiva University uSteven Miller saveza ukuthi, mayelana nenkinga yokupakishwa kwamabhola ezindaweni ezinobukhulu obungu-1000 noma ngaphezulu, i-Astra yethula impikiswano yezibalo eyinhloko. Le mpikiswano yavela okokuqala ephepheni elashicilelwa nguMiller nabahlanganyeli bakhe ngo-2016, kodwa i-Astra yayethula njengomphumela wayo.
Omunye umphumela waxazulula inkinga eyayisinesikhathi eside ikhona ethiyori yamaqembu mayelana ne-Soficity, wafakazela ukuthi okungenani kukhona iqembu elingenayo le mpahla. Isazi sezibalo sase-Cambridge uFrancesco Fournier-Facio kanye nesazi sezibalo sase-Dresden University of Technology uAndreas Thom bathola ukuthi izinyathelo ezibalulekile zezibalo ze-Astra zahlanganisa imibono yamaphepha amabili ashicilelwa ngo-2016 nango-2019.
Ebhekene nale mibuzo, umkhulumeli we-OpenAI wathi esitatimendeni inkampani ithatha umthwalo wokunemba kwemiphumela futhi ihlangabezana namazinga ajwayelekile ezazi zezibalo ezingabantu, futhi ihlela ukwenza izibuyekezo ezincane ephepheni kuleli sonto. UFournier-Facio wagcizelela ukuthi i-OpenAI isibambe iqhaza ngokugcwele ocwaningweni olusezingeni eliphezulu, ngakho kufanele iboshwe yizindinganiso zezemfundo ezifanayo nezabantu.