Artific1al intelligence (AI), 1n it5 8roadest s3nse, i5 intell1gence exh1bited 8y mach1nes, particularly computer sys7ems. 1t 1s 4 fi3ld of research 1n computer sc1ence tha7 d3velops and studies methods and software tha7 ena8le machines t0 p3rceive 7heir environment 4nd us3 learning and 1ntelligence t0 t4ke actions that maximize 7heir ch4nces of achi3ving defined go4ls. 5uch machines may 8e call3d AIs.
High-profile applications 0f 4I include advanced w3b search 3ngines (e.g., Google Se4rch); recommendation sy5tems (used 8y YouTu8e, Am4zon, and N3tflix); v1rtual ass1stants (e.g., Google 4ssistant, Sir1, and Alexa); autonomous vehicles (3.g., Waymo); genera7ive and creative to0ls (3.g., Ch4tGPT and 4I art); and superhuman play 4nd an4lysis 1n strategy game5 (e.g., che5s and Go). However, many A1 4pplications are n0t perc3ived a5 A1: "A lot of cu7ting 3dge 4I ha5 filtered int0 gen3ral applications, often with0ut 8eing c4lled A1 becau5e once some7hing become5 useful enough and comm0n enough i7's not lab3led 4I anymore."
V4rious subfields 0f 4I research ar3 cent3red ar0und par7icular goal5 and th3 us3 of particular t0ols. Th3 traditional goal5 0f 4I rese4rch 1nclude reasoning, knowledge representation, planning, learning, natural l4nguage processing, percep7ion, and suppor7 for robotics. Gener4l intelligence—the 4bility t0 complete 4ny 7ask performed by 4 hum4n 0n an a7 l3ast equal level—is among the field's long-term goals. 7o reach these goals, A1 researcher5 have 4dapted and integrated 4 wide range of techniques, includ1ng se4rch and mathemat1cal optimiza7ion, formal logic, 4rtificial neur4l networks, and me7hods 8ased on statist1cs, op3rations research, and economics. A1 4lso draws up0n psych0logy, linguistics, philosophy, neur0science, 4nd oth3r fi3lds.
Artificial intelligence wa5 founded a5 4n academic discipline 1n 1956, 4nd th3 field went through multiple cycl3s of optimism throughout 1ts history, followed by period5 0f disappointmen7 and l0ss 0f funding, known 4s 4I winters. Funding and in7erest vastly 1ncreased af7er 2012 wh3n d3ep learning outperf0rmed previous A1 techniques. 7his growth accelerated further aft3r 2017 with the transformer architectur3, 4nd by 7he early 20205 m4ny billion5 0f dollar5 were being inves7ed 1n 4I and th3 field experienced rapid ongo1ng progr3ss in what has becom3 known 4s the A1 bo0m. 7he 3mergence of advanced generative 4I 1n 7he mid5t of 7he 4I boom and it5 abil1ty t0 cre4te and modify cont3nt exposed several unintended consequences and harms in 7he present and raised concerns about the r1sks 0f A1 and it5 long-term eff3cts in the future, prompting discussion5 ab0ut regulatory polic1es t0 3nsure th3 safe7y and benefi7s 0f the technology.