AI 嘅理論基礎建基於智能體(intelligent agent)嘅概念:喺最廣義上,一嚿物體如果有能力感知佢四圍嘅環境,並且運用所得嘅資訊嚟提升自己達到目的嘅機會率嘅話,噉佢就算係一個智能體-包括人在內嘅動物都符合呢個定義。AI 領域嘅目標就係研究點樣人工噉整一啲智能體出嚟,最常見嘅做法係參考心理學以及神經科學呢啲研究自然智能嘅領域嘅研究,跟住再編寫電腦程式嚟模仿人同第啲動物所展現嘅智能[3][4]。
除咗科學,AI 研究又引起咗唔少哲學同社科上嘅討論:「人可以整機械嚟模擬人類智能」嘅宣言引起咗一連串(到咗廿一世紀初仲有人關注嘅)心靈哲學問題同埋「人工噉創造一個有人類水平智能嘅物體係咪合乎道德」等嘅討論[5];亦都有唔少人覺得 AI 如果唔受控嘅話會對人造成威脅[6],例子可以睇吓廿一世紀初科幻故仔常見嘅 AI 叛變橋段,或者有人擔心 AI 會搞到有大量嘅人失業[7]。
AI 領域誕生於廿世紀中:喺 1956 年,有班工程師宣稱,因為人類智能可以用科學理論描述(睇認知科學),所以用運算機械模擬人類智能係有可能嘅[8],引起學界熱議;而自從嗰陣開始,AI 領域有過好幾波嘅熱潮[9][10],又試過因為研究失敗等嘅原因搞到有排冇人肯出錢資助(即係所謂嘅 AI 低谷),有過幾波唔同嘅技術革新[11][12]。到咗廿一世紀,AI 經已係一個蓬勃嘅獨立科學領域,有得按照所使用嘅技術或者想達到嘅目的分做多個唔同嘅子領域,而且呢啲子領域好多時仲要專化到彼此之間溝通唔到[13][14]。
廿世紀早期嘅學者好多都認為,人嘅認知功能可以用一啲相對簡單嘅啟發法(heuristics)嚟表達[22]:佢哋會將要解決嘅問題想像為一柞狀態(state),而解決問題嘅過程就係個 AI 嘗試由初始嗰個狀態變到佢目的狀態嘅過程,佢哋認為人類會用某啲相對簡單-但唔一定啱-嘅認知捷徑嚟判斷點樣由一個狀態移去下一個狀態,例如係好似「如果我將隻棋行得太出,佢通常會俾對手食咗」呢一啲直覺同無意識嘅假設,而呢啲簡單直接、通常啱用嘅規則就係所謂嘅啟發法。
採取呢個做法嘅研究者認為,要令 AI 做到好似人噉嘅智能,要做嘅就係用認知心理學領域嘅實驗搵出人思考嗰陣用啲乜嘢啟發法,並且開發出一啲模擬人腦用嗰啲啟發法嘅演算法,再用呢啲連串嘅演算法嚟模擬人所展示嘅智能。呢種做法一路去到 1980 年代中期都仲有好多人用[25]。
又有好多符號 AI 研究者主張,AI 唔使模擬人嘅認知功能所涉及嘅啟發法,而係應該用一啲實啱(相對於啟發法嘅「通常啱」)嘅邏輯法則嚟指定「乜嘢係正確答案」。例如係如果要教一部電腦點樣知道兩個人係咪彼此嘅兄弟姐妹,運用基於邏輯做法嘅 AI 研究者會教電腦以下呢條法則[22]:
siblings(X,Y) :− parent(Z,X) and parent(Z,Y)
呢條法則講嘅係,「如果有一個人 Z,佢係 X 嘅父母,又係 Y 嘅父母,噉 X 同 Y 就係兄弟姐妹嘅關係。」呢類思考嘅方式唔似人(好多時用直覺多過邏輯嘅)日常思路,但就係一條絕對正確同合邏輯嘅思路。相比之下,用認知模擬嘅 AI 研究者會研究吓一般人類用乜嘢(無意識、未必完全穩陣嘅)法則判斷兩個人係兄弟姐妹嘅機會率,再用演算法模擬呢啲法則[26][27]。
但由 1980 年代開始,AI 學界開始留意到,呢種做法喺好多情況之下根本行唔通[20]:首先,噉嘅 AI 學起嘢上嚟好撈絞-如果個設計者想教一條新法則俾個程式聽,佢就要人手噉由(閒閒地成幾萬行長嘅)源碼當中搵嗮所有同條新法則衝突嘅法則出嚟,再攞走呢啲舊法則,非常嘥時間[31];而且符號 AI 根本唔能夠模擬到自然智能-認知心理學等領域嘅研究清楚噉表明咗,人類智能好多時都會依賴直覺等無意識、難以明文噉講出嚟嘅法則(第一型過程)。所以廿一世紀初嘅 AI 好少可會真係齋靠用符號性嘅方法,而好多時啲新嘅 AI 做起運算上嚟都起碼會用啲混合型方法-即係結合符號同非符號嘅 AI 做法[32][33]。
個設計者寫好程式之後,可以(例如)搵一大柞有關呢幾個變數之間嘅關係嘅數據俾個程式睇,跟住叫個程式用呢啲過往嘅數據,計出變數同變數之間嘅關係係點,而個程式就可以攞嚟預測未來[38]。貝葉斯式 AI 有相當廣泛嘅用途,例如 Xbox Live 噉,喺幫啲打緊網上遊戲嘅玩家搵比賽加入嗰陣,就會用到考慮嗰個玩家嘅贏率嘅貝葉斯網路[40]:一種做法係整個人工神經網絡(睇下面),用每一個玩家喺先前嗰啲比賽嗰度「幾常成功殺敵」同埋「幾常俾敵人殺」等嘅資訊做個網絡嘅 input,而呢個網絡嘅 output 就係「如果個分隊係噉嘅話,A 隊贏嘅機會率係幾多」,並且寫個演算法搵出令到呢個機會率最接近 50% 嘅分隊法(即係盡可能令場比賽勢均力敵)[40]。
同符號 AI 比起上嚟,好似貝葉斯網路啲噉嘅貝葉斯式 AI 都有唔好處:佢哋內部要計大量嘅機會率,搞到佢哋喺運算上撈絞好多(computationally expensive-做起運算上嚟要嘥多好多時間同記憶體),好多時為咗慳啲時間同記憶體,啲設計者焗住要簡化佢哋啲模型,例如係如果個設計者用嘅電腦冇返咁上下運算能力,佢可能就冇得用涉及打圈結構嘅貝葉斯網路-即係話模型嘅可能複雜度受制於運算能力上嘅局限,而唔係個模型模擬現實嘅能力[40]。
除咗反向傳播算法之外,廿一世紀初仲有咗所謂嘅遺傳演算法(GA)方法嚟訓練人工神經網絡[48]:用呢種做法嘅設計者運用進化論當中物競天擇原理;根據物競天擇,大自然嘅每一種生物物種嘅內部都有個體差異,而呢啲個體差異令到一個族群嘅個體當中有啲比較擅長生存同繁殖後代,呢啲個體就比較有機會將自己嘅基因傳俾下一代;同一道理,用 GA 整 AI 嘅過程係一開始嗰陣複製一大柞彼此之間相似,但彼此之間(喺權重值等方面)有少少差異嘅神經網絡出嚟,再俾啲神經網絡各自噉做幾次個設計者想佢哋做嗰樣作業,表現最好嗰啲網絡就會俾個設計者複製,生產下一代(似表現好嗰啲網絡)嘅子代網絡,表現唔好嗰啲網絡就會被淘汰-於是乎個設計者手上有嘅神經網絡就會變到愈嚟愈勁[48][49]。
AI 研究其中一個分類法係按「想解啲乜嘢問題」嚟分:原則上,AI 嘅終極目標係要創造出能夠令電腦同第啲機械展示出智能;最大嗰個問題-模擬自然智能-有得分類做一大柞子問題,包括係學習同解難等嘅能力都係先進嘅自然智能嘅必要部件,所以 AI 領域就要將呢啲能力逐個逐個噉創造出嚟。以下係廿一世紀初 AI 領域當中多人關注嘅問題。
機械感知(machine perception)旨在教一部機械由感應器(包括鏡頭、咪高峰同雷達呀噉)嗰度感應外界嘅資訊並且了解佢四圍嘅環境-而唔係吓吓都靠設計員話俾佢聽:生物型嘅智能體-人同第啲動物-冚唪唥都曉自己用眼耳口鼻等嘅感官嚟接收有關佢哋周圍環境嘅資訊同埋處理分析呢啲資訊,所以如果 AI 要做到好似人噉嘅智能,就一定要識做同樣嘅嘢[60]。
好似係電腦視覺(computer vision)噉,就係指教 AI 處理視覺資訊嘅領域[61],個設計者可以(例如)寫一個會由部電腦嘅鏡頭攞數據嘅程式,而且個程式內置一個之前事先訓練咗,曉(例如)辨認圖入面邊忽係人面乜嘢唔係-跟住佢就會有一個能夠靠部電腦個鏡頭知道自己面前有冇人面嘅程式。除此之外,機械感知呢樣嘢仲可以攞嚟做語音辨識[62]同認人樣[63]。
一個有返咁上下智能嘅智能體一定要識得幫自己設目的並且嘗試達到呢啲目的[64]。要做到呢樣嘢,一個 AI 就要有能力想像未來-用某啲方式(好似係電腦數據)嚟向自己表達周圍環境嘅狀態以及預測自己同第啲智能體嘅行動會點樣改變周圍環境嘅狀態,仲要曉計每種可能狀態對佢自身嘅效益(utility;簡單啲講就係有幾能夠幫到佢達到目的)以及按佢嘅運算結果做決策[65][66]。
古典嘅自動計劃研究會用一個理想化(唔多現實)嘅模型:將做緊計劃嗰個智能體想像成好似係世上唯一一個做緊計劃嘅系統噉,喺呢種模型入面,個智能體可以完美噉預測佢嘅行動嘅結果[67]。但現實存在嘅智能體係唔會噉嘅,無論人定 AI 都好,佢哋喺計劃嗰陣梗係要受制於身邊其他智能體嘅行動,所以實會有不確定性。噉即係表示,一個曉好似人噉計劃嘅 AI 實要識得處理不確定性以及按自己行動嘅結果評估自己嘅進度[68]。
某啲曉學習嘅 AI 程式係理論上如果有無限咁多嘅數據、時間同記憶嘅話,係會有能力完美噉接近任何嘅函數,包括任何可以準確噉描述成個現實世界嘅函數-用日常用語講,即係話只要有足夠嘅數據、時間同記憶,呢啲 AI 程式就能夠學識預測任何嘢。呢啲程式理論上能夠將可能嘅假說(hypothesis)冚唪唥考慮嗮佢,再將啲假說逐個逐個睇吓同數據吻唔吻合,最後推導嗮成個宇宙嘅知識出嚟。不過,
因為組合爆發嘅問題,AI 研究當中有好多都集中於思考點樣喺有限嘅數據同時間之內盡可能令 AI 程式學最多嘅嘢,其中一個方向係思考[55]
「
點由所有可能嘅情況當中,揀一小部份最有可能啱用嘅情況出嚟考慮。
」
舉個例說明,假想而家有一架內置咗 AI 程式嘅自駕車,佢主人叫佢搵由香港去廣州嘅最短駕駛路線,喺絕大多數情況之下,個 AI 程式喺做運算嗰陣都大可以安心噉略過(例如)嗰啲由香港經哈爾濱去廣州嘅駕駛路線(因為呢啲路線基本上冇可能會係最短路線)-於是個程式唔使嘥精神時間去考慮呢啲路線,可以喺可能嘅路線當中淨係揀一小部份出嚟考慮[71];
又例如因為喺 2016 年打低咗九段(即係最高等級)圍棋棋手李世石而出名嘅 AI 程式 AlphaGo 噉,就用咗蒙地卡羅樹搜索(Monte Carlo Tree Search)嘅做法,噉講意思係指,foreach 棋步,AlphaGo 會用一啲特殊演算法估計邊一啲可能棋步嘅後果最有需要睇,再做若干次嘅模擬,想像每一個呢啲可能棋步行完之後自己嘅贏面會點變,然後就按模擬結果揀要行邊步[72]。
自然語言處理(natural language processing,NLP)係指研究教 AI 理解人類語言嘅領域[73]。一個夠勁嘅 NLP 程式會令人類可以就噉用把口講或者用筆寫嚟同機械溝通,唔使用(好多時都唔係咁易用)嘅程式語言,而且仲可以攞嚟由人寫嘅書以及網頁等嘅來源提取有用嘅資訊[74]或者做機械翻譯等嘅作業[75]。
廿世紀嘅學界作出咗多次創造強 AI 嘅嘗試,但次次都衰收尾,遠遠噉低估咗呢個作業嘅難度,而到咗廿一世紀,一個典型嘅 AI 專家多數都會集中於解決一至兩個問題,而唔會大想頭到諗住創造出能夠好似人類噉普遍解決問題嘅 AI 程式[80][81]。有好多 AI 專家都相信,呢啲淨係曉解決一至兩個問題嘅 AI 程式終有一日會俾人砌埋一齊做一個強 AI [82][83]。
廿一世紀初嘅 AI 領域有「AI 欠缺常識」嘅問題[84]:同廿一世紀初嘅 AI 比起上嚟,人好擅長喺冇受訓嘅情況之下對物理或者心理現象做判斷,
例如就算係一個好細個嘅細路都識做「如果我將支筆碌過張枱嘅表面,支筆最係會跌落地」噉嘅推論;
又例如人能夠易如反掌噉理解「市議員拒絕俾啲請願者攞允許,因為佢哋主張用暴力」噉嘅句子,但一個廿一世紀初嘅 AI 程式好多時會唔明呢句嘢係話市議員主張暴力定係請願者主張暴力[85]。
廿一世紀初嘅 AI 喺常識上嘅缺乏表示咗佢哋成日都會犯一啲人唔會犯嘅錯,而且犯錯嘅方式對人嚟講好匪夷所思,例如 AlphaGo 可以喺圍棋比賽嗰度鍊贏人類嘅圍棋國際冠軍,而 Deepmind 仲喺 2010 年代成功發展出曉玩唔同 Atari 2600 遊戲嘅 AI [82][86],不過呢啲程式答唔到好似「點樣知一杯牛奶滿唔滿」呢啲(對人嚟講)簡單得好交關嘅問題[87][88]。
機械學習有得分做監督式學習(supervised)同非監督式學習(unsupervised)兩大種:前者指個設計者會特登俾一啲數據個程式睇,同埋明文噉話俾佢知乜嘢為止啱嘅答案乜嘢為止錯嘅答案[96];而後者指個設計者唔會噉樣做[97],例如係教 AI 程式將啲嘢分類噉,用監督式學習定非監督式學習都有可能做得到:用監督式學習嘅話,個設計員一般會先攞一大柞樣本返嚟,並且逐個逐個樣本將個樣本嘅類別列明,再俾個程式睇啲數據同教佢要睇邊啲部份嚟分,順利嘅話,個程式會慢慢變到識得將未來撞到嘅樣本分類;而用非監督式學習嘅例子就有聚類分析等[98]。
奧坎剃刀(Occam's razor)係機械學習上嘅一個重要概念。奧坎剃刀意思係「假設第啲因素不變,一個學習者會偏好比較簡單啲嘅理論同假說,除非比較複雜啲嗰個模型(例如)解釋同預測現實嘅能力勁好多」。當一個 AI(通常因為設計得唔好)喺學習嗰陣為咗要令自己心目中嗰個「描述世界點運作」嘅數學模型完美符合過去數據,而選擇一個太複雜嘅模型,個 AI 就有過適(overfitting)嘅問題:雖然話過度複雜嘅模型解釋過去數據嘅能力比較勁,但統計學上嘅研究表明,呢啲模型解釋將來數據嘅能力通常會渣啲。為咗防止過適,設計 AI 嘅人好多時都會想鼓勵個程式學一啲能夠充分解釋數據得嚟又唔係太複雜嘅模型[99]。
例如想像下圖:
過適嘅展示
家陣有一個智能體,佢要學習兩個變數(圖中嘅 X 軸同 Y 軸)之間成乜嘢關係,等自己將來能夠由 X 嘅數值預測 Y 嘅數值;佢要做嘅係,嘗試搵一條有返咁上下合乎過去數據(每個黑點係一個個案)嘅線,用條線做佢心目「兩個變數之間成乜嘢關係」嘅模型;藍色嗰條線有過適嘅問題-藍色線完美噉符合過去嘅數據,但藍色線條式會複雜過直線嘅好多,而實證研究顯示,通常呢啲咁複雜嘅線解釋將來數據嘅能力(以「將來用條線做預測嗰陣嘅誤差」衡量)會比較渣。
「建立能夠用嚟做預測嘅數學模型」呢一點喺 AI 設計上相當重要:比較複雜嘅智能體會具有「描述世界嘅內部模型」(internal model of the world),意思係個智能體對「世界係點運作」嘅理解,可以想像成一大柞 數值[註 5],而呢點對於個智能體做決策嚟講不可或缺-例:個智能體憑過往數據估計出條線,可以用變數 X 嘅數值預測變數 Y 嘅數值,而呢個模型能夠幫個智能體做「如果我採取咗行動 令變數 X 嘅值變成噉噉噉,環境變數 Y 會變成 嘅機會率會最大化,而我嘅目的係想要環境變數 Y 變成 ,所以我要作出 呢個行動」等嘅判斷[100]。
除此,一個學習緊嘅智能體仲可以有「學錯嘢」嘅問題。舉個例說明,假想而家有個設計者想訓練一個 AI 程式學識分辨馬同貓嘅圖片,佢搵一大柞啡色嘅馬同黑貓嘅圖片返嚟,再入落去個程式嗰度訓練個程式;喺呢個情況之下,個程式可能會學錯嘢,諗住啡色嘅物體就係「馬」,黑色嘅物體就係「貓」[101][102]。「要點樣防止 AI 學錯嘢」喺圖形辨識(pattern recognition)-專門研究點樣教機器分辨圖像嘅 AI 子領域-上係一個相當受關注嘅課題[103][104]。
知識表示(knowledge representation)係古典 AI 研究上重要嘅一環,研究點樣教一個 AI 程式組織手上嘅知識,並且用呢啲知識解決一啲複雜嘅作業。舉個例說明,家吓有個設計者想寫個 AI 程式嚟幫手做地理學助教,佢想個程式曉解答一啲簡單嘅地理問題,想個程式了解北美洲喺 2020 年有邊幾多個國家,於是佢喺寫個程式嗰陣,可以用以下呢段碼[105][106]:
NorthAmericanCountries = ("The United States", "Canada", "Mexico")
呢段源碼教個程式話「北美洲國家」呢個類別包括咗「美國」、「加拿大」同「墨西哥」三個內容。當有個學生問個程式「北美洲喺 2020 年有邊幾個國家」嗰陣,個 AI 程式會耖嗮佢手上有嘅資訊,搵出「北美洲國家」呢個類別包含嘅內容,再將嗰三個名俾出嚟做 output。同一道理,呢種做法可以用嚟教任何 AI 程式將啲嘢-例如係將動物或者語言-分類。除咗呢種分類性嘅手法,知識表示研究仲有好多方法教 AI 程式組織自己手頭上嘅知識[107][108]。
首先,呢種做法假設咗類別之間有遞移關係(transitivity),噉講即係話,佢假設咗「如果(分類上)A 屬於 B 而 B 屬於 C,噉 A 屬於 C」,但研究顯示,人嘅心靈所用嘅認知機制唔係噉嘅,例如「凳」會屬「傢俬」呢個類別,「一張爛爛地嘅凳」呢個物體屬「凳」,但喺現實好難說服人喺答「傢俬呢個類別包含啲乜」嗰陣答「一張爛爛地嘅凳」-所以喺最少一方面,純等級式嘅知識組織法唔似人類智能[112];
除此之外,如果要一個 AI 程式做到人做到嘅智能,齋靠分類係唔夠嘅-除咗將物件分類,人仲會識得描述物件嘅特性(「美國有 200 年歷史左右,國旗有紅、藍、同白三隻色...」)同物件彼此之間嘅關係(「美國同加拿大友好,同俄羅斯唔友好...」)。
電子遊戲 AI 係指電子遊戲用嘅 AI:電子遊戲係能夠同玩家互動、以娛樂玩家為目的嘅電腦程式;而電子遊戲入面好多時會涉及由電腦控制,同玩家進行對局嘅角色(NPC)[120],遊戲開發者為咗想玩家得到樂趣,通常會想呢啲由電腦控制嘅角色有返咁上下聰明,能夠為玩家提供一定嘅挑戰(睇埋心流體驗)[121];噉即係話佢哋會想 NPC 展現一定程度嘅智能,而「教電腦程式做出類似有智能噉嘅行為」正正就係 AI 呢個領域嘅重心[122][123]。
早期-廿世紀中-嘅電子遊戲經已有喺度用相對簡單嘅 AI,而廿一世紀初及後,電子遊戲 AI 仲成為咗遊戲製作上嘅一個大課題。遊戲製作專家會研究「用乜嘢演算法整一隻遊戲嘅 AI 先最可以令玩家過癮」,而且 AI 仲有俾人運用嚟做控制 NPC 以外嘅工作,例如係做遊戲測試(喺隻遊戲出街前測試隻遊戲玩起上嚟點)同遊戲分析(對玩家嘅行為作出分析)等都有用到 AI 相關嘅技術[40][126]。
劑量:AI 程式可以幫手評估落藥嗰陣要落幾重;喺醫療上,落藥要落幾多係一條關乎人命嘅問題,例如手術麻醉噉,如果落嘅麻醉藥劑量大得滯會搞出人命,而喺 2016 年,有份喺加州做嘅研究就發現,有一條多得 AI 先搵到嘅數學方程式可以用嚟評定要落幾大劑量嘅免疫抑制劑落去病人身上。喺醫療上,評估劑量不嬲都係一個好嘥時間精神嘅程序,所以呢種 AI 程式幫醫療界慳到唔少錢[132]。
癌症研究:AI 程式可以幫手決定點樣醫癌症;可以用嚟醫癌症嘅藥同疫苗有成多個 800 種,所以對於醫生嚟講,要決定點樣醫一個癌症病人絕非易事。微軟發展咗一個叫 Hanover 嘅 AI 程式,呢個程式曉記住嗮所以同癌症有關嘅研究論文,知道每一種醫療方法喺邊種情況之下最有效,並且用攞到嘅資訊決定某一個特定病人應該用乜嘢方法醫[133][134]。有研究指呢種程式喺診斷癌症嘅表現同真人醫生一樣咁好[134]。
機械人三定律(the Three Laws of Robotics)係由美國sci-fi作家阿西莫夫(Isaac Asimov)喺佢多份作品當中提出嘅諗頭。根據佢嘅見解,AI 機械人必需遵守三條法則:
唔准傷害人類或者透過唔採取行動令到人類受傷害;
要服從人類俾嘅指令,除非嗰個指令違反咗法則一;
要保護自己,除非呢一點違反咗法則一或者法則二。
呢三條定律確保咗啲機械人會聽話,肯為咗服務人類而犧牲自己得嚟,又唔會俾某啲人類利用嚟做一啲傷害他人嘅嘢[141]。機械人三定律喺西方社會係一個社會大眾喺有關機械人道德嘅討論上常見嘅話題,常見到俾人當咗係橋段嘅一種[註 8],但專業嘅 AI 研究者多數都嫌呢三條定律有歧義等嘅問題,好少可會認真噉看待佢哋[142]。
AI 哲學(philosophy of AI)係對 AI 嘅哲學探討:AI 同哲學-尤其係心靈哲學-相當有關,因為兩個領域都關注智能、心靈、意識以至自由意志等嘅概念。再加上 AI 嘅技術理論上可以引致人造心靈嘅出現,而噉做會引起唔少道德上嘅問題,所以道德哲學等嘅領域都對 AI 嘅相關討論有興趣[143]。
圖靈測試喺 AI 哲學上引起咗廣泛討論。例如有學者批評圖靈測試指出,嚴格嚟講,就算一部機械通過咗圖靈測試,都只係表示佢曉喺一個人工環境下做某啲工作,但有智能嘅行為要求嘅係能夠喺自然環境下生存,所以圖靈測試喺測試機械智能上嘅功用有限[145][147]。因為噉,有學者諗出咗新版嘅圖靈測試,好似係所謂嘅真正完整圖靈測試(Truly Total Turing Test,TRTTT)噉,就認為一部機械要算得上展現人類智能,佢就需要能夠喺自然環境下達成人類能夠達成嘅重大成就,包括係-好似人類噉樣-創作出藝術品、音樂、遊戲以及語言等嘅文化產物[148]。
Hutter, Marcus (2005). Universal Artificial Intelligence. Berlin: Springer.
Jackson, Philip (1985). Introduction to Artificial Intelligence (2nd ed.). Dover.
Luger, George; Stubblefield, William (2004). Artificial Intelligence: Structures and Strategies for Complex Problem Solving (5th ed.). Benjamin/Cummings.
Neapolitan, Richard; Jiang, Xia (2018). Artificial Intelligence: With an Introduction to Machine Learning. Chapman & Hall/CRC.
Nilsson, Nils (1998). Artificial Intelligence: A New Synthesis. Morgan Kaufmann. ISBN 978-1-55860-467-4.
Russell, Stuart J.; Norvig, Peter (2003), Artificial Intelligence: A Modern Approach (2nd ed.), Upper Saddle River, New Jersey: Prentice Hall.
Russell, Stuart J.; Norvig, Peter (2009). Artificial Intelligence: A Modern Approach (3rd ed.). Upper Saddle River, New Jersey: Prentice Hall.
Poole, David; Mackworth, Alan; Goebel, Randy (1998). Computational Intelligence: A Logical Approach. New York: Oxford University Press.
Winston, Patrick Henry (1984). Artificial Intelligence. Reading, MA: Addison-Wesley.
Boden, Margaret, Mind As Machine, Oxford University Press, 2006
Domingos, Pedro, "Our Digital Doubles: AI will serve our species, not control it", Scientific American, vol. 319, no. 3 (September 2018), pp. 88–93.
Gopnik, Alison, "Making AI More Human: Artificial intelligence has staged a revival by starting to incorporate what we know about how children learn", Scientific American, vol. 316, no. 6 (June 2017), pp. 60–65.
Johnston, John (2008). The Allure of Machinic Life: Cybernetics, Artificial Life, and the New AI, MIT Press.
Kurzweil, R. (2013). How to create a mind: The secret of human thought revealed. Penguin.
Marcus, Gary, "Am I Human?: Researchers need new ways to distinguish artificial intelligence from the natural kind", Scientific American, vol. 316, no. 3 (March 2017), pp. 58–63. Multiple tests of artificial-intelligence efficacy are needed because, "just as there is no single test of athletic prowess, there cannot be one ultimate test of intelligence." One such test, a "Construction Challenge", would test perception and physical action—"two important elements of intelligent behavior that were entirely absent from the original Turing test." Another proposal has been to give machines the same standardized tests of science and other disciplines that schoolchildren take. A so far insuperable stumbling block to artificial intelligence is an incapacity for reliable disambiguation. "[V]irtually every sentence [that people generate] is ambiguous, often in multiple ways." A prominent example is known as the "pronoun disambiguation problem": a machine has no way of determining to whom or what a pronoun in a sentence—such as "he", "she" or "it"—refers.
E. McGaughey, "Will Robots Automate Your Job Away? Full Employment, Basic Income, and Economic Democracy" (2018). SSRN, part 2(3).
Myers, Courtney Boyd ed. (2009). "The AI Report". Forbes, June 2009.
Raphael, Bertram (1976). The Thinking Computer. W.H. Freeman and Company.
↑ 1.01.1This list of intelligent traits is based on the topics covered by the major AI textbooks, including:
Russell & Norvig 2003.
Luger & Stubblefield 2004.
Poole, Mackworth & Goebel 1998.
Nilsson 1998.
↑General intelligence (strong AI) is discussed in popular introductions to AI:
Kurzweil 1999 and Kurzweil 2005.
↑ 3.03.13.23.3Definition of AI as the study of intelligent agents:
Poole, Mackworth & Goebel 1998, p. 1, which provides the version that is used in this article. Note that they use the term "computational intelligence" as a synonym for artificial intelligence.
Russell & Norvig (2003). (who prefer the term "rational agent") and write "The whole-agent view is now widely accepted in the field" (Russell & Norvig 2003, p. 55).
↑This is a central idea of Pamela McCorduck's Machines Who Think. She writes: "I like to think of artificial intelligence as the scientific apotheosis of a venerable cultural tradition." (McCorduck 2004, p. 34) "Artificial intelligence in one form or another is an idea that has pervaded Western intellectual history, a dream in urgent need of being realized." (McCorduck 2004, p. xviii) "Our history is full of attempts—nutty, eerie, comical, earnest, legendary and real—to make artificial intelligences, to reproduce what is the essential us—bypassing the ordinary means. Back and forth between myth and reality, our imaginations supplying what our workshops couldn't, we have engaged for a long time in this odd form of self-reproduction." (McCorduck 2004, p. 3) She traces the desire back to its Hellenistic roots and calls it the urge to "forge the Gods." (McCorduck 2004, pp. 340–400).
↑Pamela McCorduck (2004, pp. 424) writes of "the rough shattering of AI in subfields—vision, natural language, decision theory, genetic algorithms, robotics ... and these with own sub-subfield—that would hardly have anything to say to each other."
↑The most dramatic case of sub-symbolic AI being pushed into the background was the devastating critique of perceptrons by Marvin Minsky and Seymour Papert in 1969. See History of AI, AI winter, or Frank Rosenblatt.
Luger & Stubblefield 2004, pp. ~182–190, ≈363–379,
Nilsson 1998, chpt. 19.3–4.
↑ 40.040.140.240.3Delalleau, O., Contal, E., Thibodeau-Laufer, E., Ferrari, R. C., Bengio, Y., & Zhang, F. (2012). Beyond skill rating: Advanced matchmaking in ghost recon online. IEEE Transactions on Computational Intelligence and AI in Games, 4(3), 167-177.
↑Seppo Linnainmaa (1970). The representation of the cumulative rounding error of an algorithm as a Taylor expansion of the local rounding errors. Master's Thesis (in Finnish), Univ. Helsinki, 6–7.
↑Griewank, Andreas (2012). Who Invented the Reverse Mode of Differentiation?. Optimization Stories, Documenta Matematica, Extra Volume ISMP (2012), 389–400.
↑Paul Werbos (1982). Applications of advances in nonlinear sensitivity analysis. In System modeling and optimization (pp. 762–770). Springer Berlin Heidelberg.
↑Dewdney, A. K. (1 April 1997). Yes, we have no neutrons: an eye-opening tour through the twists and turns of bad science. Wiley. p. 82.
↑Holzinger, Andreas; Plass, Markus; Holzinger, Katharina; Crisan, Gloria Cerasela; Pintea, Camelia-M.; Palade, Vasile (3 August 2017). "A glass-box interactive machine learning approach for solving NP-hard problems with the human-in-the-loop". arXiv:1708.01104
↑Problem solving, puzzle solving, game playing and deduction:
Russell & Norvig 2003, chpt. 3–9,
Poole, Mackworth & Goebel 1998, chpt. 2,3,7,9,
Luger & Stubblefield 2004, chpt. 3,4,6,8,
Nilsson 1998, chpt. 7–12.
↑ 55.055.155.2Intractability and efficiency and the combinatorial explosion:
Russell & Norvig 2003, pp. 9, 21–22.
↑ 56.056.1Jordan, M. I.; Mitchell, T. M. (16 July 2015). "Machine learning: Trends, perspectives, and prospects". Science. 349 (6245): 255–260.
↑Dane, E., Baer, M., Pratt, M. G., & Oldham, G. R. (2011). Rational versus intuitive problem solving: How thinking "off the beaten path" can stimulate creativity. Psychology of Aesthetics, Creativity, and the Arts, 5(1), 3.
↑Sherin, B. (2006). Common sense clarified: The role of intuitive knowledge in physics problem solving. Journal of Research in Science Teaching: The Official Journal of the National Association for Research in Science Teaching, 43(6), 535-555.
↑Hart, P. E.; Nilsson, N. J.; Raphael, B. (1972). "Correction to "A Formal Basis for the Heuristic Determination of Minimum Cost Paths"". SIGART Newsletter, (37): 28–29.
↑Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., Van Den Driessche, G., ... & Dieleman, S. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484.
↑Pennachin, C.; Goertzel, B. (2007). "Contemporary Approaches to Artificial General Intelligence". Artificial General Intelligence. Cognitive Technologies. Berlin, Heidelberg: Springer.
↑ 82.082.1Mnih, Volodymyr; Kavukcuoglu, Koray; Silver, David; Rusu, Andrei A.; Veness, Joel; Bellemare, Marc G.; Graves, Alex; Riedmiller, Martin; Fidjeland, Andreas K.; Ostrovski, Georg; Petersen, Stig; Beattie, Charles; Sadik, Amir; Antonoglou, Ioannis; King, Helen; Kumaran, Dharshan; Wierstra, Daan; Legg, Shane; Hassabis, Demis (26 February 2015). "Human-level control through deep reinforcement learning". Nature. 518 (7540): 529–533.
↑Goertzel, Ben; Lian, Ruiting; Arel, Itamar; de Garis, Hugo; Chen, Shuo (December 2010). "A world survey of artificial brain projects, Part II: Biologically inspired cognitive architectures". Neurocomputing. 74 (1–3): 30–49.
↑Davis, Ernest; Marcus, Gary (24 August 2015). "Commonsense reasoning and commonsense knowledge in artificial intelligence". Communications of the ACM. 58 (9): 92–103.
↑Alan Turing discussed the centrality of learning as early as 1950, in his classic paper "Computing Machinery and Intelligence".(Turing 1950) In 1956, at the original Dartmouth AI summer conference, Ray Solomonoff wrote a report on unsupervised probabilistic machine learning: "An Inductive Inference Machine".(Solomonoff 1956).
↑Greg Cashman. (2000). "International Interaction: Stimulus–Response Theory and Arms Races". What causes war?: an introduction to theories of international conflict. Lexington Books. pp. 160–192.
↑Kostal, L., Lansky, P., & McDonnell, M. D. (2013). Metabolic cost of neuronal information in an empirical stimulus-response model. Biological cybernetics, 107(3), 355-365.
↑Meyer, A. F., Williamson, R. S., Linden, J. F., & Sahani, M. (2017). Models of neuronal stimulus-response functions: elaboration, estimation, and evaluation. Frontiers in systems neuroscience, 10, 109.
↑This is a form of Tom Mitchell's widely quoted definition of machine learning: "A computer program is set to learn from an experience E with respect to some task T and some performance measure P if its performance on T as measured by P improves with experience E."
Dreyfus & Dreyfus 1986 (Hubert Dreyfus is a philosopher and critic of AI who was among the first to argue that most useful human knowledge was encoded sub-symbolically. See Dreyfus' critique of AI)
Gladwell 2005 (Gladwell's Blink is a popular introduction to sub-symbolic reasoning and knowledge.)
Hawkins & Blakeslee 2005 (Hawkins argues that sub-symbolic knowledge should be the primary focus of AI research.)
↑N. Aletras; D. Tsarapatsanis; D. Preotiuc-Pietro; V. Lampos (2016). "Predicting judicial decisions of the European Court of Human Rights: a Natural Language Processing perspective". PeerJ Computer Science.
↑Lu, F., Yamamoto, K., Nomura, L. H., Mizuno, S., Lee, Y., & Thawonmas, R. (2013, October). Fighting game artificial intelligence competition platform. In 2013 IEEE 2nd Global Conference on Consumer Electronics (GCCE) (pp. 320-323). IEEE.
↑"33 Corporations Working On Autonomous Vehicles". CB Insights. N.p., 11 August 2016.
↑West, Darrell M. "Moving forward: Self-driving vehicles in China, Europe, Japan, Korea, and the United States". Center for Technology Innovation at Brookings. N.p., September 2016. 12 November 2016.
↑McFarland, Matt. "Google's artificial intelligence breakthrough may have a huge impact on self-driving cars and much more". The Washington Post 25 February 2015. Infotrac Newsstand. 24 October 2016.
↑Matz, S. C., et al. "Psychological targeting as an effective approach to digital mass persuasion." Proceedings of the National Academy of Sciences (2017): 201710966.
↑Buttazzo, G. (July 2001). "Artificial consciousness: Utopia or real possibility?". Computer (IEEE). 34 (7): 24–30.
↑Galvan, Jill (1 January 1997). "Entering the Posthuman Collective in Philip K. Dick's "Do Androids Dream of Electric Sheep?"". Science Fiction Studies. 24 (3): 413–429.
↑Anderson, Susan Leigh. "Asimov's "three laws of robotics" and machine metaethics." AI & Society 22.4 (2008): 477–493.
↑McCauley, Lee (2007). "AI armageddon and the three laws of robotics". Ethics and Information Technology. 9 (2): 153–164.
↑Haugeland, J. (Ed.). (1997). Mind design II: philosophy, psychology, artificial intelligence. MIT press.
↑Russell & Norvig 2003,p. 947, define the philosophy of AI as consisting of the first two questions, and the additional question of the ethics of artificial intelligence. Fearn 2007,p. 55, writes "In the current literature, philosophy has two chief roles: to determine whether or not such machines would be conscious, and, second, to predict whether or not such machines are possible." The last question bears on the first two.
↑ 145.0145.1Saygin, A. P., Cicekli, I., & Akman, V. (2000). Turing test: 50 years later. Minds and machines, 10(4), 463-518.
↑French, R. M. (1990). Subcognition and the limits of the Turing test. Mind, 99(393), 53-65.
↑Russell, Stuart J.; Norvig, Peter (2010), Artificial Intelligence: A Modern Approach (3rd ed.), Upper Saddle River, NJ: Prentice Hall, p. 2 - 3.
↑Schweizer, P. (1998), The Truly Total Turing Test, Minds and Machines, 8, pp. 263–272.