InoteveraGaidhi rinotevera
Kurega AI Ichiti Handizivi
Mutauro AI
Mutauro AI GUIDE
BloombergGPT imhando yekutsvagisa yakadzidziswa pamusanganiswa wezvemari uye zvinyorwa zvakajairika, ichiratidza kuti domain data inogona kubatsira sei nemabasa emutauro wezvemari uku uchiramba uchigona mutauro wakakura.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
Pepa rayo rinoshuma mhedzisiro, kwete vimbiso yekuti chero yezvemari LLM inopa yakavimbika investimendi, kutevedzera kana yazvino musika zano.
BloombergGPT i50-bhiriyoni-parameter mutauro modhi inotsanangurwa mune 2023 bepa rekutsvagisa nevatsvagiri veBloomberg. Vatsvagiri vakaunganidza korasi ye363 bhiriyoni emari tokens uye 345 bhiriyoni general-chinangwa tokens. Aya ma corpus saizi, kwete chirevo chekuti chiratidzo chese chakadyiwa mukudzidziswa. Vanyori vakaongorora modhi pazviyereso zvemutauro zvakajairwa, mabhenji emari uye mabasa emukati. Mutauro wezvemari une mazwi akasarudzika, zvimiro zvegwaro uye mabasa. Muenzaniso wakadzidziswa pane zvinyorwa zvedomeini unogona kudzidza mamiririro anobatsira pakuronga manzwiro, kuziva masangano, kupindura mibvunzo kana kupfupisa zvinyorwa zvemari. General data inogona kubatsira kuchengetedza zvakakura kugona. Asi chiyero chedata uye hunyanzvi hazvingoite zvega kuvimbisa chokwadi, kurongeka kwenhamba kana kukodzera kwesarudzo yakadzorwa. Mhedzisiro yebhenji inotsamira pane yakasarudzwa dataset, metrics, baselines uye nguva yekuongorora. Kana uchienzanisa iyo LLM yemari neyakajairwa modhi, tsanangura chinangwa chebasa kutanga. Shandisa anomiririra, akachengetwa mafaira, nhau, zvinyorwa kana mibvunzo yeanalyst, uye dzivirira kuvuza kubva mumienzaniso yebhenji. Ongorora nhamba chaidzo zvakasiyana kubva kumhando yeprose. Yedza zuva-inorerekera mibvunzo, zvinyorwa, makambani asingawanzo, mienzaniso yakaipa uye zvinyorwa zvemitauro yakawanda kana izvo zviri muchikamu. Enzanisa zvinobuda negwaro rekwakabva uye uyera kuti kangani vaongorori vanoda kugadzirisa mhedzisiro. Chigadziro chekutsvaga chinogona kupa zvinyorwa zvemazuva ano kana zvinyorwa zvemusika, asi kutora hakubvisi kudiwa kwekuongorora macalculation uye makotesheni. Chengetedza kwakabva uye zuva regwaro, uye ita kuti sisitimu isiyanise ruzivo rwakadzoserwa kubva kutsanangudzo inogadzirwa nemuenzaniso. Chengetedza mvumo yevanhu mukufambiswa kwebasa uko zvinobuda zvinogona kukanganisa mari, chikwereti, kutevedzera kana sarudzo dzevatengi. Ongorora kuvanzika, kodzero dzedata, latency uye mutengo pamwe nemhando yebasa. Bepa reBloombergGPT rinoratidzawo miganho yezvichemo zveveruzhinji: mabhenji emukati anogona kunge asingagone kudzokororwa akazvimirira, uye mukana wakashumwa pane imwe bvunzo haugadzirise hukuru pamabasa. MaLLM emari anofanirwa kubatwa sezvikamu zvebasa remutauro, kwete zviremera zvinozvimiririra pamari kana mutemo. Vashandise ivo chete mukati meyakajeka yekusimbisa uye yekuzvidavirira maitiro.
Muongorori anoenzanisa modhi-yakasarudzika yezvemari uye muenzaniso wakajairwa pamihoro-yakasarudzika-kufona mibvunzo uchishandisa rubriki yakafanana uye magwaro ekwakabva.
Chikwata chekutsvagisa chinobvunza chimiro chemutauro kuti chiburitse manhamba-shizha kubva mufaira, chobva chasimbisa nhamba dzese zvichienderana netafura yefaira yakanyorwa.
Chikwata chechigadzirwa chinoongorora kupatsanurwa kwemanzwiro pahukuru hwekambani uye nenhau masosi, ichitarisa kuti mhedzisiro inoshanduka here pamamiriro emari asina kujairika.
Kuteedzera kufambiswa kwemabasa kunopa pfupiso yemuenzaniso muzvinyorwa zvemazuva uye inonyora zvakabuda seyamuro yekuongorora kwete kurudziro.
Mamiriro emutauro wezvemari anogona kuvandudzwa sezvo ongororo ichivharira mamwe mabasa, nguva uye marudzi ezvinyorwa, asi mabhenji anowanikwa anofanirwa kusungirirwa kune kongiri yekushanda. Zvikwata zvinofanirwa kuchengetedza date source corpora, kuzorodza bvunzo dzakaitwa uye kuyera nhamba uye zvikanganiso zvekutaura zvakasiyana nekutsetsenura. Enzanisa akasarudzika uye akajairwa masisitimu pahumbowo humwe. Shandisa kusaina-kwemunhu kuita sarudzo dzakakosha uye ongororazve kukodzera kana data, zvigadzirwa kana mamiriro emusika achinja. Rekodha mazuva ekuyedzwa uye kuumbwa kwemuenzaniso kuitira kuti mibairo yakashumwa ienzaniswe mushanduro dzese.
Tsamba yemahara
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
BloombergGPT imhando yekutsvagisa yakadzidziswa pamusanganiswa wezvemari uye zvinyorwa zvakajairika, ichiratidza kuti domain data inogona kubatsira sei nemabasa emutauro wezvemari uku uchiramba uchigona mutauro wakakura. Pepa rayo rinoshuma mhedzisiro, kwete vimbiso yekuti chero yezvemari LLM inopa yakavimbika investimendi, kutevedzera kana yazvino musika zano.
Iro bepa rinotsanangura hombe yemari corpus yakasanganiswa ne general-chinangwa data.
Zvibodzwa zveBenchmark zvinogumira kune yakaedzwa dataset, metric uye evaluation dhizaini.
Kutsetsenura uye kurongeka kwechokwadi kwakasiyana kugona; manhamba nenhambo zvinoda macheki akananga.
Kudzosa kunogona kupa humbowo, asi muongorori achiri kutarisa zvirevo uye maverengero achipesana nekwakabva.
Iyo yakagovaniswa, yakabatwa-yakagadzika seti inoita kuti kuenzanisa kuve nechirevo uye inoderedza kurerekera kwesarudzo.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Kurega AI Ichiti Handizivi
Mutauro AI