D

   Hollywood’s theory that machines with evil(邪恶) minds will drive armies ofkiller robots is just silly. The real problem relates to the possibility thatartificial intelligence(AI) may become extremely good at achieving somethingother than what we really want. In 1960 a well-known mathematician NorbertWiener, who founded the field of cybernetics（控制论）, put it this way: “If we use, to achieve our purposes, a mechanicalagency with whose operation we cannot effectively interfere(干预), we had better be quite surethat the purpose which we really desire.”

   A machine with a specific purpose hasanother quality, one that we usually associate with living things: a wish topreserve its own existence. For the machine, this quality is not in-born, noris it something introduced by humans; it is a logical consequence of the simplefact that the machine cannot achieve its original purpose if it is dead. So ifwe send out a robot with the single instruction of fetching coffee, it willhave a strong desire to secure success by disabling its own off switch or evenkilling anyone who might interfere with its task. If we are not careful, then,we could face a kind of global chess match against very determined, superintelligent machines whose objectives conflict with our own, with the realworld as the chessboard.

   The possibility of entering into and losingsuch a match should concentrating the minds of computer scientists. Someresearchers argue that we can seal the machines inside a kind of firewall,using them to answer difficult questions but never allowing them to affect thereal world. Unfortunately, that plan seems unlikely to work: we have yet to inventa firewall that is secure against ordinary humans, let alone super intelligentmachines.

   Solving the safety problem well enough tomove forward in AI seems to be possible but not easy. There are probablydecades in which to plan for the arrival of super intelligent machines. But theproblem should not be dismissed out of hand, as it has been by some AIresearchers. Some argue that humans and machines can coexist as long as theywork in teams—yet that isnot possible unless machines share the goals of humans. Others say we can just “switchthem off” as if super intelligent machines are too stupid to think of that possibility.Still others think that super intelligent AI will never happen. On September11, 1933, famous physicist Ernest Rutherford stated, with confidence, “Anyonewho expects a source of power in the transformation of these atoms is talkingmoonshine.” However, on September 12, 1933, physicist Leo Szilard invented theneutron-induced(中子诱导) nuclearchain reaction.

67.Paragraph 1mainly tells us that artificial intelligence may         .

A. run out ofhuman control

B. satisfy human’sreal desires

C. command armiesof killer robots

D. work fasterthan a mathematician

68.Machines withspecific purposes are associated with living things partly because they mightbe able to        .

A. preventthemselves from being destroyed

B achieve theiroriginal goals independently

C. do anythingsuccessfully with given orders

D. beat humans ininternational chess matches

69.According tosome researchers, we can use firewalls to           .

A. help superintelligent machines work better

B. be secureagainst evil human beings

C. keep machinesfrom being harmed

D. avoid robots’affecting the world

70.What does theauthor think of the safety problem of super intelligent machines?

A. It willdisappear with the development of AI.

B. It will getworse with human interference.

C. It will besolved but with difficulty.