Ukukhawulezisa i-AI-Powered Robotics kunye noKhuseleko lwe-Intanethi: Impumelelo kuQeqesho lweRobhothi yokwenyani kunye noNciphiso lwe-Ahashi.

Ukukhawulezisa iiRobhothi ze-AI-Powered kunye noKhuseleko lwe-Intanethi: Ukuphumelela kuQeqesho lweRobhothi yokwenyani kunye noNciphiso lwe-Agatha

Intsebenziswano yakutshanje phakathi kwe-KAIST kunye ne-Nvidia ibangele ukuphumelela okuphawulekayo: ukunciphisa ixesha loqeqesho lwe-robot yokwenene nge-31-68% ngokusebenzisa umgaqo-nkqubo wokufunda ngokuqhubekayo (RL). Le mpumelelo ineempembelelo ezinkulu ekumiselweni kweerobhothi ezinikwe amandla e-AI kumashishini afana nelemveliso, ulungiselelo, kunye nokhathalelo lwempilo. Ngaphezu koko, ukuqaliswa kwe-Osprey 1.0, injini yemithetho evulekileyo yokunciphisa ukusetyenziswa kakubi kwexesha langempela, inika isixhobo sedemokhrasi kwimibutho yokukhusela abasebenzisi ekusebenziseni kakubi i-intanethi. Eli nqaku liphonononga iinkcukacha zobugcisa kunye nefuthe elinokubakho lale mpumelelo kwishishini.

KAIST kunye noNvidia ACCRUE: A New Standard for Real-Robot Training

Ukusetyenziswa kwe-ACRUE yomgaqo-nkqubo oqhubekayo we-RL kumisele umgangatho omtsha woqeqesho lwerobhothi yokwenyani, ukwehlisa ngokubonakalayo ixesha loqeqesho elifunekayo. Le theyibhile ilandelayo ibonisa uthelekiso phakathi kwe-RL yemveli kunye ne-ACRUE:

Umzekelo, inkampani efana ne Amazon Robotics inokunyusa i-ACRUE ukukhawulezisa ukuthunyelwa kweerobhothi ezinikwe amandla e-AI kwiindawo zazo zokugcina, iphucula ukusebenza kakuhle kunye nokunciphisa iindleko.

Osprey 1.0 Injini yeMithetho eVulekileyo yoKhuseleko lwe-Intanethi

I-Osprey 1.0 ibonelela ngesixhobo esibalulekileyo kwimibutho yokukhusela abasebenzisi bayo ekuxhatshazweni kwe-intanethi. I-snippet yekhowudi elandelayo ibonisa indlela yokuchaza umthetho wokufumana umxholo oxhaphazayo usebenzisa i-Osprey 1.0:

ngenisa i-osprey

# Chaza umthetho wokufumana umxholo oxhaphazayo
umthetho = osprey.Rule(
    igama = "Ukufunyanwa kwempatho-gadalala",
    definition="Ibona umxholo oxhaphazayo kwisicatshulwa esiveliswe ngumsebenzisi",
    iimeko=[
        osprey.Imeko(
            igama="Ukudibanisa kwegama elingundoqo",
            description="Idibanisa amagama angundoqo anxulumene nokuxhatshazwa",
            iiparamitha={
                "keywords": ["abuse", "uchuku", "spam"]
            }
        )
    ],
    izenzo=[
        osprey.Action(
            igama="Flag Content",
            definition="Iphawula umxholo njengoxhaphazayo",
            iiparamitha={
                "flag_type": "ukuxhatshazwa"
            }
        )
    ]
)

# Sebenzisa umgaqo kumxholo owenziwe ngumsebenzisi
umxholo = "Lo ngumzekelo wesiqulatho esixhaphazayo"
iziphumo = osprey.apply_rule(umgaqo, umxholo)
ukuba isiphumo:
    print("Umxholo uphawulwe njengohlukumezayo")

Njengoko kuxelwe ngu Wired, ukusetyenziswa kwe-AI-powered solutions kukhuseleko lwe-intanethi kuya kubaluleka kakhulu, kwaye i-Osprey 1.0 linyathelo elibalulekileyo kweli cala.

💡 Into ephambili yokuThatha: Ukuqaliswa kwe-Osprey 1.0 kunika ithuba elibalulekileyo kwimibutho yokuphucula ukhuseleko lwe-intanethi ngokusebenzisa iinjini zemithetho evulelekileyo. Ngokufikelela kwidemokhrasi kwezi zixhobo, i-Osprey 1.0 inokunceda ukunciphisa umthwalo wokuxhatshazwa kwe-intanethi kunye nokukhuthaza indawo ekhuselekileyo ye-intanethi.

Iimpembelelo kuyilo kunye noRhwebo

Ukuphuhliswa kwe-ACCRUE kunye ne-Osprey 1.0 kuphakamisa imibuzo ebalulekileyo malunga neempembelelo zezakhiwo kunye norhwebo olubandakanyekayo kuyilo lwabo. Umzekelo, ukusetyenziswa komgaqo-nkqubo oqhubekayo we-RL kwi-ACRUE kunokufuna izixhobo ezibalulekileyo zokubala kunye nokugcinwa kwedatha. Ngokufanayo, ukuthunyelwa kwe-Osprey 1.0 kunokufuna ukuqwalaselwa ngokucokisekileyo kwemigaqo yokusebenza kwe-injini kunye ne-scalability.

⚠️ UMngcipheko woYilo: Ukwamkelwa kwe-ACCRUE kunye ne-Osprey 1.0 kunokwazisa imingcipheko emitsha yolwakhiwo, efana nokwanda kobunzima bokubala kunye neemfuno zokugcina iinkcukacha. Amaqela obunjineli kufuneka avavanye ngocoselelo olu rhwebelwano kwaye ayile iisistim zabo ngokufanelekileyo ukuqinisekisa ukusebenza ngokupheleleyo kunye nokwandiswa.

Iingcebiso zokuKhawulezisa iiRobhothi eziXhaswa yi-AI kunye noKhuseleko lwe-Intanethi

Ukukhawulezisa ukwamkelwa kweerobhothi ezine-AI kunye nokuphucula ukhuseleko kwi-intanethi, sicebisa oku kulandelayo:

• Phonononga ukusetyenziswa kwe-RL eqhubekayo ngaphandle komgaqo-nkqubo kuqeqesho lwerobhothi yokwenyani ukunciphisa ixesha loqeqesho kunye nokuphucula ukusebenza kakuhle.
• Sebenzisa i-injini yemithetho evulelekileyo efana ne-Osprey 1.0 ukuphucula ukhuseleko lwe-intanethi kunye nokunciphisa umthwalo wokuxhatshazwa kwi-intanethi.
• Vavanya ngononophelo iimpembelelo zoyilo kunye norhwebo olubandakanyekayo kuyilo kunye nokusasazwa kwezi nkqubo ukuqinisekisa ukusebenza ngokupheleleyo kunye nokukalwa.

Imithombo

• Iindaba ze-AI Namhlanje - Amabali e-AI aphambili kunye nohlaziyo lwangoLwesine, nge-15 kaOkthobha, ngo-2026.
• ExplainX - Amabali e-AI aphezulu achazwe nge-15 ka-Okthobha ngo-2026.
• Eneentambo — Iindaba kunye namanqaku kubukrelekrele bokwenziwa.
• Iirobhothi zaseAmazon - Izisombululo zeRobhothi zogcino kunye nolungiselelo.