Computer Science. The English Language Perspective - Беликова (1176925), страница 38
Текст из файла (страница 38)
For one thing thereis our fascination with intelligence, which seemingly imparts tous humans a special place among life forms. Questions arisesuch as “What is intelligence?”, “How can one measureintelligence?” or “How does the brain work?” All thesequestions are meaningful when trying to understand artificialintelligence. However, the central question for the engineer,especially for the computer scientist, is the question of the225intelligent machine that behaves like a person, showingintelligent behavior. The attribute artificial might awaken muchdifferent associations. It brings up fears of intelligent cyborgs.
Itrecalls images from science fiction novels. It raises the questionof whether our highest good, the soul, is something we shouldtry to understand, model, or even reconstruct. With suchdifferent offhand interpretations, it becomes difficult to definethe term artificial intelligence or AI simply and robustly.In 1955, John McCarthy, one of the pioneers of AI, was the firstto define the term artificial intelligence, roughly as follows: Thegoal of AI is to develop machines that behave as though theywere intelligent.To test this definition, imagine the following scenario. Fifteen orso small robotic vehicles are moving on an enclosed squaresurface.
One can observe various behavior patterns. Somevehicles form small groups with relatively little movement.Others move peacefully through the space and gracefully avoidany collision. Still others appear to follow a leader. Aggressivebehaviors are also observable. Is what we are seeing intelligentbehavior? According to McCarthy’s definition these robots canbe described as intelligent, thus it is clear that this definition isinsufficient.In the Encyclopedia Britannica one finds a definition that goeslike: AI is the ability of digital computers or computercontrolled robots to solve problems that are normally associatedwith the higher intellectual processing capabilities of humans . .. But this definition also has weaknesses.
It would admit, forexample, that a computer that can save a long text and retrieveit on demand displays intelligent capabilities, for memorizationof long texts can certainly be considered a higher intellectualprocessing capability of humans, as can, for example, the quickmultiplication of two 20-digit numbers. According to thisdefinition, then, every computer is an AI system. This dilemmais solved elegantly by the following definition by Elaine Rich:Artificial Intelligence is the study of how to make computers dothings at which, at the moment, people are better.226Rich, tersely and concisely, characterizes what AI researchershave been doing for the last 50 years.
Even in the year 2050, thisdefinition will be up to date.Tasks such as the execution of many computations in a shortamount of time are the strong points of digital computers. Inthis regard they outperform humans by many multiples. Inmany other areas, however, humans are far superior tomachines. For instance, a person entering an unfamiliar roomwill recognize the surroundings within fractions of a secondand, if necessary, just as swiftly make decisions and planactions.
To date, this task is too demanding for autonomousrobots. According to Rich’s definition, this is, therefore, a taskfor AI. In fact, research on autonomous robots is an important,current theme in AI. Construction of chess computers, on theother hand, has lost relevance because they already play at orabove the level of grandmasters.It would be dangerous, however, to conclude from Rich’sdefinition that AI is only concerned with the pragmaticimplementation of intelligent processes. Intelligent systems, inthe sense of Rich’s definition, cannot be built without a deepunderstanding of human reasoning and intelligent action ingeneral, because of which neuroscience is of great importanceto AI.
This also shows that the other cited definitions reflectimportant aspects of AI. A particular strength of humanintelligence is adaptivity. We are capable of adjusting to variousenvironmental conditions and change our behavior accordinglythrough learning. Precisely because our learning ability is sovastly superior to that of computers, machine learning is,according to Rich’s definition, a central subfield of AI.In 1950, computer pioneer Alan M. Turing suggested aproductive approach to evaluating claims of artificialintelligence in what became known as the Turing test.
He gavea definition of an intelligent machine, in which the machine inquestion must pass the following test. The test person Alice sitsin a locked room with two computer terminals. One terminal isconnected to a machine, the other with a non-malicious person227Bob. Alice can type questions into both terminals. She is giventhe task of deciding, after five minutes, which terminal belongsto the machine.
The machine passes the test if it can trick Aliceat least 30% of the time.Computer programs have been able to pass the Turing test to alimited extent. The AI pioneer and social criticJosephWeizenbaum developed a program named Eliza, whichis meant to answer a test subject’s questions like a humanpsychologist. He was in fact able to demonstrate success inmany cases. Supposedly his secretary often had longdiscussions with the program. Today in the internet there aremany so-called chatterbots, some of whose initial responses arequite impressive. After a certain amount of time, however, theirartificial nature becomes apparent.Notes:John McCarthy (1927 - 2011) was a legendary computerscientist at Stanford University who developed time-sharing,invented LISP, and founded the field of Artificial Intelligence.Elaine Rich works as Distinguished Senior Lecturer at theUniversity of Texas at Austin. Books: Automata, Computabilityand Complexity: Theory and Applications (author), ArtificialIntelligence (co-author).Joseph Weizenbaum (1923 - 2008) was a German-Americancomputer scientist who is famous for his development of theEliza program in 1966 and for his views on the ethics ofartificial intelligence.
He became sceptical of artificialintelligence and a leading critic of the AI field following theresponse of users to the Eliza program.Assignments1. Translate the sentences from the texts into Russian inwriting paying attention to the underlined words andphrases:2281. The term artificial intelligence stirs emotions.
For onething there is our fascination with intelligence, whichseemingly imparts to us humans a special place amonglife forms.2. With such different offhand interpretations, it becomesdifficult to define the term artificial intelligence or AIsimply and robustly.3. AI is the ability of digital computers or computercontrolled robots to solve problems that are normallyassociated with the higher intellectual processingcapabilities of humans.4.
In this regard they outperform humans by manymultiples. In many other areas, however, humans are farsuperior to machines.5. It would be dangerous, however, to conclude fromRich’s definition that AI is only concerned with thepragmatic implementation of intelligent processes.2. Answer the following questions:1. What is the key AI problem to be addressed bycomputer scientists?2. Why is McCarthy’s definition called “insufficient”?3. What is wrong with the definition of AI in theEncyclopedia Britannica?4.
Where do machines outperform humans? Where dopeople win?5. What is the essence of the Turing test?3. Translate into English:Эрик Браун, 45-летний исследователь из IBM, отвечаетза мозг суперкомпьютера Ватсон, который в 2011 г. получилизвестность победами над людьми в популярнойтелевикторине. Самая большая трудность для Брауна, как229наставника машины, не в том, чтобы впихнуть в Ватсонакак можно больше знаний, но в том, чтобы придатьтонкость его пониманию языка. Например, научить слэнгу.Как проверить, может ли компьютер «мыслить»?Классический тест — так называемый тест Тьюринга —прост: он предполагает способность вести светскую беседу.Если бы компьютер сумел бы не выдать свою двоичнуюсущность в непринужденном разговоре, он бы доказал своеинтеллектуальное превосходство. Но пока ни одноймашине это не удалось.Два года назад Браун попытался натаскать Ватсона спомощью популярного веб-сайта Urban Dictionary.Словарные статьи на сайте составляются обычнымипользователямииредактируютсядобровольнымиредакторами по достаточно произвольным правилам.
Тутесть всевозможные актуальные аббревиатуры как bb (англ.bye bye) — пока, hf (англ. have fun) — отличноповеселиться, w8 (англ. wait) — жди. В том числе иогромное количество всяких слэнговых конструкций,таких, как hot mess — «горячая штучка».Но Ватсон не мог различить салонную лексику ислэнговую — которой в Urban Dictionary хватает. Крометого, из-за чтения Википедии Ватсон приобрел некоторыедурные привычки. В ответах на вопросы исследователя втестах он использовал малоцензурные словечки.В конечном счете команда Брауна разработала фильтр,чтобы отцеживать брань Ватсона, и выскребла UrbanDictionary из его памяти.
Это испытание доказывает,насколько тернист будет путь любого железногоинтеллектуала к “лёгкой болтовне”. Теперь Браунподготавливает Ватсона к использованию в качестведиагностического инструмента в больнице: там знаниевсяких модных аббревиатур не потребуется.4. Give the summary of the text using the key terms.230APPROACHES AND TECHNIQUESRead the following words and word combinations and usethem for understanding and translation of the text:conventional - общепринятый, традиционныйcomputational intelligence - вычислительный интеллектmachine learning - машинное обучениеcase-based reasoning - вывод (рассуждения), основанныена прецедентахbehavior-based AI - поведенческий ИИreferred to as - под названием, именуемыйneural networks - нейронные сетиfuzzy logic - нечеткая логикаneats versus scruffies - чистюли против неряхad hoc rules - ситуативные правилаinference engine - механизм логического выводаforward chaining - прямой логический выводbackward chaining - обратный логический выводdirected acyclic graph - ориентированный циклическийграфarc - дугаconditional dependence - условная зависимостьto be subject to controversy - вызывать спорыtrack record - послужной список, достиженияThe artificial intelligence community can be roughly dividedinto two schools of thought: conventional AI and computationalintelligence.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
