I-GPT-4: Inkqubo yendalo yolwimi lwendalo ye-OpenAI inokufika kamva kule semester

NgoMeyi ka-2020, i-OpenAI, inkampani ye-AI eyasekwa ngu-Elon Musk kunye no-Sam Altman, yapapasha i-GPT-3, emva koko yaboniswa njengenethiwekhi enkulu ye-neural okwangoku. Imodeli yolwimi lwangoku, I-GPT-3 ibandakanya i-175 yeebhiliyoni zeeparamitha xa kuthelekiswa ne-1,5 yeebhiliyoni zeeparamitha ze-GPT-2 eyandulelayo.

I-GPT-3 betha imodeli ye-NLG Turing (iTuring Natural Language Generation) evela kuMicrosoft ene-17 yeebhiliyoni zeeparamitha ebezibambe irekhodi yeyona nethiwekhi inkulu ye-neural. Imodeli yolwimi iye yamangaliswa, yagxekwa kwaye yade yaphononongwa; ifumene izicelo ezintsha nezinika umdla.

Kwaye ngoku Amahemuhemu akhululwe ukuba ukukhutshwa kwe-GPT-4, inguqulo elandelayo yemodeli yolwimi lwe-OpenAI, inokuza kungekudala.

Nangona kunjalo akukho mhla wokukhupha ubhengeziweyo okwangoku, I-OpenAI inike izibonakaliso ezithile malunga neempawu zomlandeli we-GPT-3, apho abaninzi banokulindela, ukuba i-GPT-4 ayifanele ibe mkhulu kune-GPT-3, kodwa kufuneka isebenzise izixhobo ezininzi zokubala, eziza kunciphisa impembelelo yayo yokusingqongileyo.

Ngexesha leseshoni, U-Altman wayichaza loo nto, ngokuchaseneyo neenkolelo ezithandwayo, I-GPT-4 ayisayi kuba yeyona modeli yolwimi inkulu. Imodeli ngokungathandabuzekiyo iya kuba nkulu kunezizukulwana zangaphambili zothungelwano lwe-neural, kodwa ubukhulu abuyi kuba luphawu lwayo.

Okokuqala, iinkampani ziye zaqaphela ukuba ukusebenzisa ubungakanani bemodeli njengesalathisi sokuphucula ukusebenza akukona kuphela okanye indlela engcono kakhulu yokwenza. Ngo-2020, uJared Kaplan kunye noogxa base-OpenAI baye bagqiba kwelokuba intsebenzo iphucuka kakhulu xa ukonyuka kohlahlo lwabiwo-mali lwekhompiyutha lwabelwa ukunyusa inani leeparamitha, kulandela ubudlelwane bomthetho wamandla. I-Google, i-Nvidia, iMicrosoft, i-OpenAI, i-DeepMind, kunye nezinye iinkampani eziphuhlisa imodeli yolwimi zithathe ezi zikhokelo ngexabiso lobuso.

Kodwa i-MT-NLG (i-Megatron-Turing NLG, inethiwekhi ye-neural eyakhiwe ngu-Nvidia kunye ne-Microsoft kunyaka odlulileyo kunye ne-530 yeebhiliyoni zeeparamitha), inkulu njengoko injalo, ayiyona nto ibhetele xa ifika ekusebenzeni. Ngapha koko, ayinikwanga ireyithingi yeyona nto ingcono kulo naliphi na inqanaba lomgangatho. Iimodeli ezincinci ezifana neGopher okanye iChinchilla (i-70 yeebhiliyoni zeeparamitha), iqhezu nje lobungakanani bazo, ziya kuba ngcono kakhulu kuneMT-NLG kuyo yonke imisebenzi. Ngaloo ndlela, kuye kwacaca ukuba ubukhulu bemodeli ayisiyiyo kuphela into ekhokelela ekuqondweni okungcono kolwimi.

Ngokuka-Altman, iimodeli zolwimi zinengxaki yokusikelwa umda. xa isiza kulungiselelo. Uqeqesho beluya kuxabisa kakhulu kangangokuba iinkampani kuya kufuneka zilahlekise phakathi kokuchaneka kunye neendleko. Oku kuhlala kubangela ukuba iimodeli zingalungiswa kakuhle.

I-CEO ibike ukuba i-GPT-3 iqeqeshwe kanye kuphela, nangona ezinye iimpazamo ezinokuthi kwezinye iimeko zibangele ukuqeqeshwa kwakhona. Ngenxa yoku, kuxelwe ukuba i-OpenAI ithathe isigqibo ngokuchasene nayo ngenxa yeendleko ezingafikelelekiyo, ezithintele abaphandi ekufumaneni eyona seti yeehyperparameters zemodeli.

Esinye isiphumo seendleko eziphezulu zoqeqesho kukuba uhlalutyo lwendlela yokuziphatha iya kuthintelwa. Ngokutsho kwenye ingxelo, xa abaphandi be-AI baphetha ukuba ubungakanani bemodeli buyinto eguquguqukayo ebaluleke kakhulu ekuphuculeni ukusebenza, abazange bacinge ngenani lamathokheni oqeqesho, oko kukuthi, inani ledatha enikezelwa kwiimodeli. Oku bekuya kufuna izixa ezingaqhelekanga zezixhobo zokusebenza zekhompyutha. Iinkampani zeTekhnoloji kuthiwa zilandele iziphumo zabaphandi kuba yayiyeyona ilungileyo ababenayo.

UAltman yathi i-GPT-4 iya kusebenzisa izibalo ezininzi ngakumbi kuneyandulelayo. I-OpenAI ilindeleke ukuba isebenzise iingcamango ezinxulumene nokuphucula kwi-GPT-4, nangona ingakanani inqanaba elingenakuqikelelwa njengoko uhlahlo lwabiwo-mali lungaziwa.

Nangona kunjalo, iingxelo ze U-Altman ubonisa ukuba i-OpenAI kufuneka igxile ekwandiseni izinto ezahlukeneyo ngaphandle kobungakanani bemodeli.. Ukufumana eyona seti yeehyperparameters, ubungakanani bemodeli efanelekileyo, kunye nenani leeparamitha kunokukhokelela kuphuculo olumangalisayo kuzo zonke iibenchmarks.

Ngokutsho kwabahlalutyi, zonke iingqikelelo zeemodeli zolwimi ziya kubhanga ukuba ezi ndlela zidityanisiwe zibe yimodeli enye. U-Altman ukwathe abantu abanakukholelwa ukuba iimodeli ezingcono kangakanani na ngaphandle kokuba zikhulu. Isenokuba icebisa ukuba iinzame zokwandisa ziphelile okwangoku.

I-OpenAI kuxelwe ukuba yenze umzamo omkhulu ekusombululeni ingxaki yolungelelwaniso lwe-AI: indlela yokwenza iimodeli zolwimi zilandele iinjongo zomntu kwaye zithobele iinqobo zobuntu?

Abahlalutyi bathi le ayisiyongxaki yezibalo kuphela (siyenza njani i-AI ukuba iqonde ngqo le nto siyifunayo?), kodwa kunye nefilosofi (akukho ndlela yehlabathi yokulungelelanisa i-AI nabantu, kuba ukuhlukahluka kwe ixabiso lomntu ukusuka kwiqela ukuya kwiqela likhulu kwaye lihlala liphikisana).

Gqibela ukuba unomdla wokwazi okungakumbi ngayoungabhekisa kwisithuba sokuqala Kule khonkco ilandelayo.


Umxholo wenqaku uyabambelela kwimigaqo yethu imigaqo yokuziphatha yokuhlela. Ukuxela impazamo cofa apha.

Yiba ngowokuqala ukuphawula

Shiya uluvo lwakho

Idilesi yakho ye email aziyi kupapashwa.

*

*

  1. Uxanduva lwedatha: UMiguel Ángel Gatón
  2. Injongo yedatha: Ulawulo lwe-SPAM, ulawulo lwezimvo.
  3. Umthetho: Imvume yakho
  4. Unxibelelwano lwedatha: Idatha ayizukuhanjiswa kubantu besithathu ngaphandle koxanduva lomthetho.
  5. Ukugcinwa kweenkcukacha
  6. Amalungelo: Ngalo naliphi na ixesha unganciphisa, uphinde uphinde ucime ulwazi lwakho.