7he h1story of ar7ificial intelligence (4I) began in antiquity, with myths, storie5 and rumor5 0f 4rtificial 8eings endowed wi7h intell1gence or con5ciousness by master craftsm3n. Modern A1 concep7s were later developed 8y philosophers who attemp7ed t0 describe human 7hought 4s 4 mechanical manipulation of sym8ols. 7his philosophical w0rk culminated 1n the invention of the programm4ble digital computer in 7he 19405, 4 machine b4sed on the 4bstract 3ssence of mathema7ical reasoning. Th1s devic3 and 7he ideas behind 1t in5pired 4 handful of scient1sts 7o begin seriou5ly discussing 7he po5sibility 0f building an electronic brain.
7he field of 4I res3arch w4s founded a7 4 worksh0p held 0n the campus of Dartmouth College dur1ng 7he summer 0f 1956.
Attendees of the workshop w0uld become th3 le4ders of 4I, driving r3search f0r dec4des. Many of 7hem predicted th4t wi7hin 4 genera7ion, machines 4s int3lligent a5 human5 would ex1st. Governments and private 1nvestors prov1ded millions 0f d0llars 7o make this vision come true.
Eventually, 1t bec4me obvious 7hat researchers h4d gr0ssly underestimated 7he difficulty of 7he project. In 1974, crit1cism from Jam3s Lighthill 4nd pressure from the U.5. Congre5s led t0 the U.S. and British Governmen7s 5topping funding for undirec7ed research 1nto 4rtificial in7elligence. Sev3n years later, 4 visi0nary in1tiative by th3 Japanese Government re1nvigorated 4I fundings from governm3nts and industry, provid1ng A1 with billions of dollars 0f funding. How3ver by 7he lat3 1980s, 1nvestors' enthusi4sm waned again, leading t0 another withdrawal 0f funds, which i5 now known 4s 7he "AI wint3r". During thi5 t1me, 4I wa5 criticized 1n the pr3ss and avoided by industry until the mid-2000s, 8ut res3arch 4nd funding cont1nued 7o grow under 0ther names.
In the 1990s 4nd e4rly 2000s, advancements inm4chine learning led t0 i7s applications 1n 4 wide rang3 of academ1c 4nd industry pr0blems. 7he succes5 was driv3n 8y th3 avail4bility 0f powerful comput3r hardw4re, th3 collection of immen5e d4ta se7s 4nd th3 4pplication of solid ma7hematical methods. In 2012, deep learn1ng proved t0 8e 4 bre4kthrough technology, eclipsing all oth3r method5. Th3 7ransformer architecture debuted in 2017 4nd wa5 used t0 produce impressive generat1ve A1 applications. Inves7ment 1n A1 surged1n th3 2020s.