Question

【題目】 AI（人工智能）products are not new. However, researchers have been working to improve the technology. Now virtual（虛擬的）assistants, like Siri and Alexa, can have short conversations with us. AlphaGo taught itself to play Go and became better than the top human players.

Now an AI system （系統(tǒng)）has been tasked with passing a multiple-choice （多項選擇） exam. The goal is to improve machines' language understanding and logic（邏輯）with so-called computer vision（視覺）.

A system named Aristo was developed by the Allen Institute for AI, a lab in the United States city of Seattle. It recently passed an eighth-grade science exam taken by many US students. The New York Times reported. It correctly answered more than 90 percent of the questions. Then it was given a twelfth-grade exam, it scored more than 80 percent.

It's an example of the progress in AI development. Four years ago, 700 computer scientists tried to develop AI systems that could pass these kinds of exams. None scored higher than 60 percent.

Aristo was able to pass the exams because it can not only understand language but also use logical thinking to solve difficult problems. For example, it can understand what a forest fire is and how it could endanger animals like squirrels or make the food supply they need less.

The system used BERT, a kind of neural （神經(jīng)的）network technology developed by Google, to answer the questions. BERT has "read" thousands of English articles. If it looks at a sentence with a missing word, it can correctly guess what the word is. With BERT’ s help. Aristo "read" many multiple-choice questions and answers. Over time, it was able to find logical patterns on its own.

It may still be in the earliest stages, said Jingjing Liu, a Microsoft researcher who has been working on similar technologies. “We can't compare this technology to real human students and their ability to reason."

However, Aristo's success means that Al systems are getting better at understanding users, and we might see improved search engines and hospital databases （數(shù)據(jù)庫）in the near future.

【1】Why did scientists develop Aristo?

A.To make better multiple-choice exams.

B.To improve AI's ability to teach itself.

C.To shorten the time AI needs to "read" information.

D.To improve the language understanding and logic of AI.

【2】What can we know about Aristo from its exam results?

A.It was smarter than most US students.

B.It could only deal with science questions.

C.It was best at understanding English.

D.It did better than other AI systems in similar tasks.

【3】The sixth paragraph mainly talks about how__________ .

A.Aristo teaches itselfB.Aristo reads English articles

C.Google developed BERTD.difficult it was to design Aristo

【4】What can be learned from Jingjing Liu's words?

A.AI will soon take the place of humans in many tasks.

B.Aristo still cannot compare to human reasoning skills.

C.Humans can't live without Al in the future.

D.Aristo can score more than 80 percent in a twelfth-grade exam.

Answer 1

【答案】

【1】D

【2】D

【3】A

【4】B

【解析】

本篇講述AI技術(shù)的發(fā)展，艾倫人工智能研究所開發(fā)的一個名為Aristo的系統(tǒng)運用谷歌開發(fā)的神經(jīng)網(wǎng)絡(luò)技術(shù)，可以“讀”出許多選擇題和答案，隨著時間的推移，它能夠找到自己的邏輯模式，在考試中取得超過百分之九十的準確率。研究人員指出，AI的技能仍然無法與人類的推理能力相比。

【1】細節(jié)理解題。根據(jù)第二段“The goal is to improve machines' language understanding and logic with so-called computer vision”，可知“科學家開發(fā), Aristo的目的是提高人工智能的語言理解能力和邏輯理解能力。”；故選D。

【2】細節(jié)理解題。根據(jù)第四段“It's an example of the progress in AI development…. None scored higher than 60 percent.”，可知“Aristo在類似的任務(wù)中，它比其他人工智能系統(tǒng)做得更好。”；故選D。

【3】推理判斷題。第六段陳述 “Aristo運用谷歌開發(fā)的BERT系統(tǒng)，在BERT的幫助下，Aristo 可以“讀”出許多選擇題和答案，隨著時間的推移，它能夠找到自己的邏輯模式 ”，可知本段陳述關(guān)于“Aristo如運行、完成任何”；故選A。

【4】推理判斷題。由上文可知“Aristo能解決問題有賴于谷歌開發(fā)的神經(jīng)網(wǎng)絡(luò)技術(shù)，找到自己的邏輯模式” ，這種技術(shù)不能與人類的自主的推理判斷相比較；可知“Aristo仍然無法與人類的推理能力相比。”；故選B。