Kulokhu okuthunyelwe, sabelana nawe ngeqoqo lemithombo yolwazi oluwusizo mayelana neSayensi Yedatha evela kubasunguli kanye ne-CTO ye-DAGsHub, umphakathi kanye nenkundla yewebhu yokulawula inguqulo yedatha nokusebenzisana phakathi kososayensi bedatha nonjiniyela bokufunda bemishini. Ukukhethwa kuhlanganisa imithombo ehlukahlukene, kusukela kuma-akhawunti we-Twitter kuya kumabhulogi wobunjiniyela agcwele, ahloselwe labo abazi kahle ukuthi bafunani. Imininingwane ngaphansi kokusikwa.
Kusuka kumbhali:
Uyilokho okudlayo, futhi njengesisebenzi solwazi udinga ukudla okuhle kolwazi. Ngifuna ukwabelana ngemithombo yolwazi mayelana neSayensi Yedatha, Ubuhlakani Bokwenziwa kanye nobuchwepheshe obuhlobene engibuthola buwusizo kakhulu noma bukhanga kakhulu. Ngethemba ukuthi lokhu kuzokusiza nawe!
Isiteshi se-YouTube esikufanelekela kahle ukuhambisana nemicimbi yakamuva. Isiteshi sibuyekezwa njalo futhi umsingathi unesasasa elithelelanayo nomdlandla wazo zonke izihloko ezixoxwayo. Lindela ukumbozwa komsebenzi othokozisayo hhayi ku-AI kuphela, kodwa nakuzithombe zekhompiyutha nezinye izihloko ezikhangayo.
Esiteshini sakhe se-YouTube, u-Yannick uchaza ucwaningo olubalulekile ekufundeni okujulile ngemininingwane yobuchwepheshe. Esikhundleni sokuzifundela wena isifundo, kuvame ukushesha futhi kube lula ukubuka enye yamavidiyo aso ukuze uthole ukuqonda okujulile kwezihloko ezibalulekile. Izincazelo zidlulisa ingqikithi yama-athikili ngaphandle kokudebeselela izibalo noma ukulahleka ngophayini abathathu. U-Yannick uphinde wabelane ngemibono yakhe yokuthi izifundo zihlangana kanjani, ukuthi ukuthatha imiphumela kubaluleke kangakanani, ukutolika okubanzi, nokuningi. Kunzima kakhulu kwabaqalayo (noma abangebona abezemfundo) ukuthi bazitholele lezi zinto ngokwabo.
Ngamazwi abo:
Ucwaningo lokufunda ngomshini kudingeka lucace, luguquleke futhi lube nempilo. Futhi i-Distill yadalelwa ukusiza ngocwaningo.
I-Distill ishicilelwe eyingqayizivele enocwaningo emkhakheni wokufunda ngomshini. Ama-athikili anokubukwa okumangalisayo athuthukiswa ukuze anikeze umfundi ukuqonda okunembile kwezihloko. Ukucabanga kwendawo nomcabango kuvame ukusebenza kahle kakhulu ekusizeni ukuqonda izihloko Zokufunda Ngomshini kanye Nesayensi Yedatha. Ngakolunye uhlangothi, amafomethi oshicilelo endabuko, avame ukuqina esakhiweni sawo, amile futhi omile, futhi ngezinye izikhathi. . UChris Olah, omunye wabadali be-Distill, naye unebhulogi yomuntu siqu emangalisayo kuyo . Ayikabuyekezwa isikhathi eside, kodwa isalokhu iyiqoqo lezincazelo ezingcono kakhulu esihlokweni sokufunda okujulile esake sabhalwa. Ikakhulukazi, kwangisiza kakhulu I-LSM!
USebastian Ruder ubhala ibhulogi enokuqonda kakhulu kanye nencwadi yezindaba, ngokuyinhloko mayelana nokuphambana kwamanethiwekhi emizwa kanye nokumbiwa kwemibhalo yolimi lwemvelo. Ubuye abe nezeluleko eziningi zabacwaningi nezikhulumi zengqungquthela, ezingasiza kakhulu uma usesikhungweni semfundo ephakeme. Ama-athikili ka-Sebastian ngokuvamile athatha uhlobo lokubuyekezwa, afinyeze futhi achaze isimo socwaningo lwamanje nezindlela emkhakheni othile. Lokhu kusho ukuthi ama-athikili awusizo kakhulu kubasebenzi abafuna ukuthola ama-bearings abo ngokushesha. USebastian naye uyabhala .
U-Andrei Karpathy akadingi singeniso. Ngaphezu kokuba omunye wabacwaningi abafunda ngokujulile abadume kakhulu eMhlabeni, udala amathuluzi asetshenziswa kabanzi, isib. njengamaphrojekthi aseceleni. Abantu abaningi bangene kulo mkhakha ngezifundo zakhe zaseStanford , futhi kuyoba usizo kuwe ukukwazi ukuqeqeshwa kwenethiwekhi ye-neural. Ngiphinde ngincome ukuyibuka mayelana nezinselele zomhlaba wangempela u-Tesla okufanele azinqobe lapho ezama ukusebenzisa umshini wokufunda ngesilinganiso esikhulu emhlabeni wangempela. Inkulumo ifundisa, ihlaba umxhwele futhi iyasangulukisa. Ngaphezu kwezindatshana ezimayelana ne-ML uqobo, u-Andrei Karpathy uyanikeza ngoba . Funda okungakumbi ngo- Andrey futhi kuqhubeke .
Ibhulogi yobunjiniyela be-Uber ihlaba umxhwele ngempela ngezinga layo nobubanzi bokuhlanganisa, ihlanganisa izihloko eziningi, ikakhulukazi. . Engikuthanda kakhulu ngesiko lobunjiniyela bakwa-Uber ukuthambekela kwabo ukukhiqiza okuthakazelisa kakhulu nokubalulekile umthombo ovulekile ngesivinini esikhulu. Nazi izibonelo:
- Futhi uhlu luqhubeka luqhubeka... Hats off, Uber
Izingxabano eceleni, ibhulogi ye-OpenAI iyamangalisa ngokungenakuphikwa. Ngokuhamba kwesikhathi, ibhulogi ithumela okuqukethwe kanye nemibono mayelana nokufunda okujulile okungafika kuphela esikalini se-OpenAI: Hypothetical. ukwehla okukabili okujulile. Ithimba le-OpenAI livamise ukuthumela njalo, kodwa lezi izinto ezibalulekile.
Ibhulogi ye-Taboola ayaziwa njengeminye yemithombo kulokhu okuthunyelwe, kodwa ngicabanga ukuthi ihlukile - ababhali babhala ngezinkinga eziphansi kakhulu, ezingokoqobo lapho bezama ukusebenzisa i-ML ekukhiqizeni "okujwayelekile". " amabhizinisi: okuncane mayelana nezimoto ezizishayelayo kanye nama-RL awina ompetha bomhlaba, okuningi mayelana "ngazi kanjani ukuthi imodeli yami manje isibikezela izinto ngokuzethemba okungamanga?" Lezi zinkinga zibalulekile cishe kuwo wonke umuntu osebenza ensimini, futhi bathola ukufakwa okuncane kwabezindaba kunezihloko ze-AI ezivamile, kodwa kusadingeka ithalente elisezingeni lomhlaba ukuxazulula lezi zinkinga ngendlela efanele. Ngenhlanhla, uTaboola unakho kokubili leli thalente kanye nokuzimisela kanye nekhono lokubhala ngalo ukuze nabanye abantu bafunde.
Kanye ne-Twitter, akukho lutho olungcono ku-Reddit kunokuhileleka ocwaningweni, amathuluzi, noma ukuhlakanipha kwesixuku.
Okuthunyelwe kushicilelwa njalo ngonyaka, kodwa kugcwele ulwazi oluminyene kakhulu. Uma kuqhathaniswa neminye imithombo ekulolu hlu, lolu lufinyeleleka kakhulu kubantu bebhizinisi okungebona ubuchwepheshe. Engikuthandayo ngezingxoxo ukuthi izama ukunikeza umbono ophelele wokuthi imboni nocwaningo kuphokophele kuphi, ihlanganisa ndawonye intuthuko yehardware, ucwaningo, ibhizinisi, kanye ne-geopolitics ngokubuka iso lenyoni. Qiniseka ukuthi uqala ekugcineni ukufunda ngokungqubuzana kwezintshisekelo.
Ama-podcast
Eqinisweni, ngicabanga ukuthi amaphodikasti awakufanelekeli kahle ukuhlola izihloko zobuchwepheshe. Phela, basebenzisa umsindo kuphela ukuchaza izihloko, futhi isayensi yedatha iyinkambu ebonakalayo kakhulu. Amaphodikhasti avame ukukunikeza izaba zokwenza ucwaningo olujulile ngokuhamba kwesikhathi noma ube nezingxoxo ezithokozisayo zefilosofi. Nokho, nazi ezinye izincomo:
- , njengoba ekhuluma nabacwaningi abavelele emkhakheni wobuhlakani bokwenziwa. Iziqephu zikaFrancois Chollet zinhle kakhulu!
- . Kuhle ukuzwa ngamathuluzi engqalasizinda yedatha entsha.
Uhlu Olumangalisayo
Kuncane ongakulandelela lapha, kodwa izinsiza ezengeziwe eziwusizo uma wazi okufunayo:
UMatty uthola izindlela ezinhle, ezinobuhlakani zokusebenzisa amanethiwekhi emizwa, futhi kuyajabulisa ukubona imiphumela yakhe kokuphakelayo kwakho kwe-Twitter. Okungenani ake ubheke ngokushesha.
U-Ori uwumshini wokushayela nje . Ubhala kabanzi ngezinkinga nezixazululo zososayensi bedatha. Qiniseka ukuthi ubhalisa ukuze waziswe uma isihloko sishicilelwa. Okwakhe , ikakhulukazi, ihlaba umxhwele ngempela.
Umsunguli we-fast.ai, umthombo ophelele wokuqamba nokukhiqiza.
Unjiniyela we-ML wezisebenzi e-Github, u-Hamel Hussain umatasatasa emsebenzini edala futhi ebika ngamathuluzi amaningi amakhodi edatha.
Umdali we-Keras, manje buyekeza ukuqonda kwethu ukuthi buyini ubuhlakani nokuthi singabuhlola kanjani.
Ucwaningo lweSayensi ku-Google Brain.
isiphetho
Okuthunyelwe kwangempela kungase kubuyekezwe njengoba umbhali ethola imithombo emihle yokuqukethwe okungaba amahloni ukungayifaki ohlwini. Zizwe ukhululekile ukuxhumana naye ku , uma ufuna ukuncoma umthombo omusha! Futhi futhi i-DAGsHub Advocate [approx. ukuhumusha umsebenzi womphakathi] ku-Data Science, ngakho-ke uma udala okwakho okuqukethwe kwe-Data Science, zizwe ukhululekile ukubhalela umbhali wokuthunyelwe.
Zithuthukise ngokufunda imithombo enconyiwe, nokusebenzisa ikhodi yokuphromotha I-HABR, ungathola u-10% owengeziwe esaphulelo esikhonjiswe kusibhengezo.
Izifundo ezengeziwe
Izihloko Ezifakiwe
Source: www.habr.com