1 . Improve Cloud Security
Sensitive customer data has constantly been found exposed on cloud servers without password protection. To ease the problem, database software makers have been trying to make security easier for cloud database managers. At the Enigma Conference in San Francisco, Kenn White, a security manager at database software maker MongoDB, will describe a new technique, called field level encryption, to make data safer on the cloud.
Field level encryption works by scrambling data before it’s sent to a cloud database and rearranging it in order when the data is needed for use. The promise of the product is to protect the contents of a cloud database, even if bad guys access it.
MongoDB’s new feature comes as more and more companies move user data to cloud servers, rather than run their own costly data centers. It was predicted that cloud computing would be a $214 billion industry by the end of 2019. That would be up more than 17% from 2018, when it was $182 billion.
Companies have rushed to the cloud without understanding all of the possible security consequences. Many companies have left countless databases exposed, revealing personal data. A database containing details about who lives in 80 million US households was left unprotected in 2019, just like the data on Facebook users.
Database managers want to store their data in an unreadable form, but they also want to be able to find specific pieces of information in the database with a simple search term. For example, someone might want to look up health care patients by their Social Security numbers, even if those numbers are stored as random characters. To make this possible, field level encryption lets database managers encrypt a search term on their machine and send it to the database as a query. The database matches the encrypted version of the search term with the record it’s storing and then sends it back to you.
This approach only works with specific kinds of data. For example, field level encryption isn’t useful for long text entries, like notes in a patient’s medical chart, because you can’t search for individual words.
Still, for data like account numbers, passwords and government ID numbers, field level encryption protects data and maintains a usable database.
Most importantly, White said, it’s simple to set up. Database managers turn it on with a one-time configuration change when they set up the database. “That’s really powerful,” he said in an interview.
1. The underlined word “scrambling” in paragraph 2 probably means________.| A．mixing | B．collecting | C．hiding | D．storing |
| A．Secure the safety of Internet pages. | B．Protect files with a unique style of storage. |
| C．Stop bad guys from accessing the database. | D．Enable companies to store files on the cloud. |
| A．Companies should move user data to cloud servers. |
| B．Cloud computing achieved a 17% increase in 2019. |
| C．Companies may be unaware of the risks of the cloud. |
| D．No companies were willing to run their own data centers. |
| A．present some facts | B．offer security advice |
| C．introduce a technique | D．recommend a product |
2 . Albert Einstein’s 1915 masterpiece “The Foundation of the General Theory of Relativity” is the first and still the best introduction to the subject, and I recommend it as such to students. But it probably wouldn’t be publishable in a scientific journal today.
Why not? After all, it would pass with flying colours the tests of correctness and significance. And while popular belief holds that the paper was incomprehensible to its first readers, in fact many papers in theoretical physics are much more difficult.
As the physicist Richard Feynman wrote, “There was a time when the newspapers said that only 12 men understood the theory of relativity. I do believe there might have been a time when only one man did, because he was the only guy who caught on, before he wrote his paper. But after people read the paper a lot understood the theory of relativity in some way or other, certainly more than 12.”
No, the problem is its style. It starts with a leisurely philosophical discussion of space and time and then continues with an exposition of known mathematics. Those two sections, which would be considered extraneous today, take up half the paper. Worse, there are zero citations of previous scientists’ work, nor are there any graphics. Those features might make a paper not even get past the first editors.
A similar process of professionalization has transformed other parts of the scientific landscape. Requests for research time at major observatories or national laboratories are more rigidly structured. And anything involving work with human subjects, or putting instruments in space, involves piles of paperwork.
We see it also in the Regeneron Science Talent Search, the Nobel Prize of high school science competitions. In the early decades of its 78-year history, the winning projects were usually the sort of clever but naive, amateurish efforts one might expect of talented beginners working on their own. Today, polished work coming out of internships(实习) at established laboratories is the norm.
These professionalizing tendencies are a natural consequence of the explosive growth of modern science. Standardization and system make it easier to manage the rapid flow of papers, applications and people. But there are serious downsides. A lot of unproductive effort goes into jumping through bureaucratic hoops(繁文缛节), and outsiders face entry barriers at every turn.
Of course, Einstein would have found his way to meeting modern standards and publishing his results. Its scientific core wouldn’t have changed, but the paper might not be the same taste to read.
1. According to Richard Feynman, Einstein’s 1915 paper ________.| A．was a classic in theoretical physics |
| B．turned out to be comprehensible |
| C．needed further improvement |
| D．attracted few professionals |
| A．Unrealistic. | B．Irrelevant. |
| C．Unattractive. | D．Imprecise. |
| A．The application of research findings. |
| B．The principle of scientific research. |
| C．The selection of young talents. |
| D．The evaluation of laboratories. |
| A．What makes Einstein great? |
| B．Will science be professionalized? |
| C．Could Einstein get published today? |
| D．How will modern science make advances? |
3 . Elizabeth Spelke, a cognitive (认知的) psychologist at Harvard, has spent her career testing the world's most complex learning system-the mind of a baby. Babies might seem like no match for artificial intelligence (AI). They are terrible at labeling images, hopeless at mining text, and awful at video games. Then again, babies can do things beyond the reach of any AI. By just a few months old, they’ve begun to grasp the foundations of language, such as grammar. They’ve started to understand how to adapt to unfamiliar situations.
Yet even experts like Spelke don’t understand precisely how babies — or adults, for that matter — learn. That gap points to a puzzle at the heart of modern artificial intelligence: We're not sure what to aim for.
Consider one of the most impressive examples of AI, Alpha Zero, a programme that plays board games with superhuman skill. After playing thousands of games against itself at a super speed, and learning from winning positions, Alpha Zero independently discovered several famous chess strategies and even invented new ones. It certainly seems like a machine eclipsing human cognitive abilities. But Alpha Zero needs to play millions more games than a person during practice to learn a game. Most importantly, it cannot take what it has learned from the game and apply it to another area.
To some AI experts, that calls for a new approach. In a November research paper, Francois Chollet, a well-known AI engineer, argued that it’s misguided to measure machine intelligence just according to its skills at specific tasks. “Humans don’t start out with skills; they start out with a broad ability to acquire new skills,” he says. “What a strong human chess player is demonstrating is not only the ability to play chess, but the potential to fulfill any task of a similar difficulty.” Chollet posed a set of problems, each of which requires an AI programme to arrange colored squares on a grid (格栅) based on just a few prior examples. It’s not hard for a person. But modern machine-learning programmes-trained on huge amounts of data — cannot learn from so few examples.
Josh Tenenbaum, a professor in MIT's Center for Brains, Minds & Machines, works closely with Spelke and uses insights from cognitive science as inspiration for his programmes. He says much of modern AI misses the bigger picture, comparing it to a cartoon about a two-dimensional world populated by simple geometrical (几何形的) people. AI programmes will need to learn in new ways — for example, by drawing causal inferences rather than simply finding patterns. “At some point — you know, if you’re intelligent — you realize maybe there's something else out there,” he says.
1. Compared to an advanced AI programme, a baby might be better at _______________.| A．labeling images | B．identifying locations |
| C．playing games | D．making adjustments |
| A．Stimulating. | B．Measuring. | C．Beating. | D．Limiting. |
| A．AI is good at finding similar patterns |
| B．AI should gain abilities with less training |
| C．AI lacks the ability of generalizing a skill |
| D．AI will match humans in cognitive ability |
| A．What is exactly intelligence? |
| B．Why is modern AI advanced? |
| C．Where is human intelligence going? |
| D．How do humans tackle the challenge of AI? |
