Chii chaitika
MarkTechPost inoshuma kuti Generalist AI yakaburitsa GEN-1.5, multimodal robhoti hwaro modhi inoshandisa 3-12 yechipiri sensorimotor kuratidzira mukati meiyo 30-yechipiri mamiriro hwindo kuita basa remuviri. Chirevo chinoti modhi iyi yakawana budiriro ye59% pamabasa gumi ekunyengera pasina kurongeka uye yakasvika 83% mushure menhanho gumi pamaminetsi mashanu edata rebasa. Zvikumbiro hazvina kusimbiswa zvakazvimiririra kubva kune kwakapihwa.
MarkTechPost inoshuma kuti Generalist AI's GEN-1.5 inogamuchira vhidhiyo, sensor, mutauro, uye proprioceptive yekupinda, inochengetedza makumi matatu-sekondi yemamiriro hwindo, uye inoburitsa chiito trajectories pa100 Hz. Iyo yepakati meshini, iyo iyo kambani inodana "yemuviri kukurudzira," inoisa pfupi sensorimotor muenzaniso mune iwo mamiriro. Muenzaniso unogona kusanganisira kamera uye sensor hova padivi peyakaratidzwa chiitiko trajectory. Sekureva kwaMarkTechPost, modhi yacho inobva yapinza basa rakatarisirwa kubva mukuratidzira uye kuita pasina kuraira kwemutauro, gradient yekuvandudza, nhanho yekugadzirisa zvakanaka, kana chirongwa chakanangana nebasa. Chirevo chinoti modhi iyi yakaramba ichidzidziswa kweanopfuura mwedzi misere pane data-yekudyidzana data yakatorwa mudzimba, matura, uye mafekitori.
MarkTechPost inoshuma avhareji yebudiriro ye59%, ine mwero wekutsauka kwegumi muzana mapoinzi, pamabasa gumi akasiyana-siyana ekunyengedza apo modhi yakadzidziswa yakagamuchira chiratidziro chimwe chete. Chinyorwa chinoti 10 gradient nhanho uchishandisa maminetsi mashanu edata rebasa rega rega rakasimudza avhareji yekubudirira kusvika 83%, nekutsauka kwakajairwa kwe9 muzana. Inoshumawo kuti imwe gradient nhanho uchishandisa miniti imwe yedata yakasvika 66.5% pane basa rakaitwa. Chinyorwa chinoratidza izvi mhedzisiro seyakaderera-data fomu yekuyedza-nguva yekudzidziswa uye inoti iyo yakashumwa uremu inoshanduka mushure mematanho gumi aive pazasi 0.15%. Nzvimbo yakapihwa haipe runyorwa rwebasa, maverengero emuenzaniso, ruzivo rwehardware, tsananguro yekutanga, kana yakazara evaluation protocol inodiwa kuongorora nhamba idzodzo wakazvimirira.
Chirevo chinotsanangura akati wandei ekutamisa maitiro ayo MarkTechPost anoratidza kune Generalist AI kuongororwa. Mazano maviri akarekodhwa akarekodhwa anogona kunge akasungirirwa kuita kumwe kuenderera mberi, nemuenzaniso uchigadzira mafambiro ekupindira akadai sekuisazve, kubatazve, uye kudzoreredza kukanganisa. Chiratidziro chakarekodhwa mukuenzanisa chinonzi chakaendeswa kune robhoti chairo kunyangwe data rekutanga kudzidzisa rakanga risina vhidhiyo inofananidzwa kana simba. Chinyorwa ichi chinotsanangurawo kuratidzwa kunoitwa nemaoko emunhu tichitarisa makamera erobhoti, zvichiteverwa nekuberekana uchishandisa maoko erobhoti. Mushure mekugadzirisa kwakareruka, MarkTechPost inoti sisitimu iyi inogona kushandisa bhanana sebhurasho, shandisa guruva kufambisa block, kubvisa bepa rinovhara ndiro, uye kushanda ambidextrously. Chirevo ichi chinotiwo maitiro anokurudzirwa anoramba ari brittle pane ayo akarongedzerwa.
Kwakabva mashoko: marktechpost.com ↗
Nei zvichikosha
Chirevo chinotsanangura shanduko yakakosha mukugadziriswa kwerobhoti: zviratidziro zvinogona kushanda sekukurudzira-mukati-kati pane kuda kudzidziswa kwakakura-kwakanangana nebasa. Kana zvakawanikwa zvikabata pamusoro peiyo kambani kuongororwa, vanogona kuderedza iyo data uye mutoro weinjiniya unobatanidzwa mukudzidzisa marobhoti maitiro matsva. Humbowo hunoramba huripo kare, uye mabasa acho aive akapfava, pfupi-pfupi, uye asina kuzvimiririra.
Izvo zvakakosha zvakakosha zveshumo ndezvekuti inogadzira kudzidza kwerobhoti sedambudziko remamiriro ezvinhu pamwe nedambudziko rekugadzirisa uremu. Yakajairwa basa inogadziriswa inogona kuda kuunganidza data uye kuita akawanda optimization matanho ega ega maitiro matsva. MarkTechPost inoshuma kuti GEN-1.5 inogona kushandisa ratidziro pfupi nekukasira, nepo diki yekuwedzera kudzidziswa inovandudza mashandiro zvakanyanya. Kune vanoita zvemarobhoti, izvo zvinogona kupedzisira zvaita kuti zvive nyore kugadzirisa yakajairika-chinangwa sisitimu yekuchinja zvinhu, marobhoti, kana maitiro pasina kuvaka yakaparadzana mutemo webasa rega rega. Ichi chinoramba chiri mhedzisiro yekutsvagisa yakashumwa, kwete humbowo hwekuti nzira yakagadzirira kushandiswa kwemaindasitiri.
Mienzaniso yekufambisa yakashumwa ine basa nekuti inoedza kupfuura kukopa zvakananga. MarkTechPost inoti modhi iyi yakagadzira mafambisirwo ekubatanidza pakati pezviratidziro zvakapatsanurwa, ichibva pazviratidziro zvekuyedzerwa kuenda pakuurayiwa kwemuviri, uye yakagadziridza kuratidzira kwevanhu kuita robhoti. Aya maitiro, kana akagadzirwazve, anogona kuratidza kuti modhi yakadzidza yakafararira yemuviri nguva dzose pane kuyeuka imwe chete yakagadziriswa trajectory. Zvakadaro, kwakabva hakuburitse kuti kangani kuchinjirwa uku kwakabudirira, kuti kupindira kwevanhu kwakabvumidzwa kwakadii, kana kuti kuratidzira nenzvimbo dzekuyedza dzakasarudzwa nekambani. Mienzaniso yakashumwa saka inoratidza gwara rekutsvagisa kwete vimbiso yakayerwa yekufunga-chinangwa chemuviri kufunga.
Mhedzisiro yacho inoratidzawo musiyano pakati pekugona uye kuwanikwa. MarkTechPost inoshuma kuti GEN-1.5 ndeyekutsvagisa kuburitswa kunoitwa paGeneralist AI's ega marobhoti zvikepe uye data injini. Iko kunonzi hakuna huremu hweveruzhinji, API, peji remitengo, kana wega-sevhisi chigadzirwa, uye vangangove vashandisi vanofanirwa kutsvaga kudyidzana kwakananga. Izvo zvinomisa kudzokorora kwakazvimirira uye zvinoita kuti zviome kumasangano ekunze kuenzanisa modhi nemamwe marobhoti-yekudzidza masisitimu. Zvinoreva zvakare kuti kukurumidza kukanganiswa neveruzhinji kunonyanya kuve kwesainzi uye nehurongwa: basa rinogona kukanganisa magadzirirwo evanoongorora marobhoti akawanda, asi vaverengi havasati vakwanisa kuyedza system yacho zvakananga kana kuongorora mutengo wayo, kuvimbika, kana mashandiro ayo.
Interactive Mechanism: Iyo Inonyatsoshanda
Ongorora ari pasi tekinoroji kuseri kwekusimudzira uku uchipindirana.
Which component of an AI application is the machine-learning model itself?
Zvekutarisa zvinotevera
Mibvunzo mikuru ndeyokuti GEN-1.5 inowanza kupfuura mabasa akashumwa, maitirwo aanoita pasi pekuchengetedza-akakosha kana mamiriro asina kujairika, uye kana mhedzisiro yacho inogona kudhindwa kunze kweGeneralist AI's ngarava uye data data. MarkTechPost inoshuma kuti iyo sisitimu haina huremu hweveruzhinji, API, mitengo, kana wega-sevhisi kuwana, saka yakazvimiririra yekuyedza uye inoshanda kuendesa kunoramba kusipo.
Chinonyanya kukosha ndechekuzvimiririra kusimbiswa. MarkTechPost's report haipei mazita emabasa e10, nhamba yezviedzo pamabasa, mapoka ekukundikana, nguva dzekuvimba, maitiro ekuenzanisa, kana mapuratifomu emuviri anoshandiswa. Mashoko iwayo akakosha pakududzira maavhareji akashumwa 59% uye 83%. Ongororo dzekutevera dzinofanira kuyedza zvinhu zvitsva, mwenje wakasiyana uye marongero epanzvimbo yekushandira, kutevedzana kwebasa refu, uye zviratidziro zvakanyorwa nevanhu kana marobhoti kunze kwepombi yedata yeGeneralist AI. Kudhindwa nevaongorori vakazvimiririra kwaizobatsira kuona kana iyo yakashumwa kugona inzvimbo yakafara yemuenzaniso kana mhedzisiro inotarisisa kuunganidzwa kwekambani uye sarudzo dzekuongorora.
Kuchengeteka uye kusimba ndeimwe nzvimbo isina kugadziriswa. MarkTechPost inoshuma kuti iyo modhi inomhanya yakavharwa-loop, inogona kushivirira kumwe kuvhiringidzika, kupora kubva mukukanganisa, uye kugadzirisa, ukuwo ichiramba yakashata kupfuura shanduro dzakarongeka. Kwakabva hakutsanangure mitengo yekudhumhana, miganhu yekumanikidza, kumira kwechimbichimbi, kukundikana kunosanganisira vanhu, kana kuita munzvimbo dzakazara uye dzisingafungidzike. Vasati vatumirwa mudzimba, matura, kana mafekitori, vaongorori vaizoda humbowo hwezvinoitika kana ratidziro isinganzwisisike, chinhu chisipo, robhoti rinorasikirwa nebasa, kana kukanganisa kunogona kukonzera kukuvara kana kukuvara. Chirevo hachipi humbowo hwekuti GEN-1.5 yakaongororwa mamiriro aya.
Kuwana uye kutonga kucharatidza kana tsvakiridzo ichinyatsobatsira. MarkTechPost inoshuma kuti Generalist AI haana kuburitsa huremu, API, mitengo, kana chigadzirwa chekuzvishandira, ichisiya sisitimu iripo chete kuburikidza nekudyidzana kwakananga. Zvidzoreso zveramangwana zvinofanirwa kujekesa kana vaongorori vekunze vachiwana mukana wekuongorora, ndezvipi zvirambidzo zvekutonga-data zvinoshanda kune rekodhi-yekudyidzana kurekodha, uye kana modhi yacho ichigona kuongororwa pasina kufumura mavhidhiyo anonetsa kubva kudzimba kana kumabasa. Izvo zvichave zvakakoshawo kuona kana iyo inonzi yakaderera-data adapta inobata pachiyero chakakura uye pamabasa ese anoda kurongwa kwakasimba. Kusvika iyo mibvunzo yapindurwa, GEN-1.5 inofanirwa kubatwa sechiratidzo chekutanga chekutsvagisa pane kutumirwa general-chinangwa marobhoti system.