Chapter 149: The Professor’s Miraculous Recovery, Georgetown’s New Life
“MIT Radiation Laboratory Series”
Ordinary people might not have heard of this name.
Simply put, it is an encyclopedia set involving the radio field.
In 1940, to cope with World War II, America transferred the nation’s top-tier physicists and engineers to MIT and established the Radiation Laboratory.
The projects involved in the laboratory covered radio, microwave, and radar projects; most of the major radio and radar equipment during World War II were produced here.
After World War II ended, the government stopped funding.
The researchers received a final ultimatum, with only the last 6 months left.
Everyone had to use this time to organize the existing research achievements into tutorials. Then the laboratory would close, and everyone would return to where they came from.
Simply put, it almost covers everything related to radio frequency electronic circuitry, laying the foundation for America’s electronic technology later on.
Watson narrowed his eyes: “Professor, I don’t quite understand.
This set of books was published by McGraw-Hill as early as 1947, and many university libraries have collections of it. With your identity, borrowing a set shouldn’t be difficult.”
“MIT Radiation Laboratory Series”, published in 1947, later America realized something was wrong and stopped publishing after 1953.
Watson was right; many universities have collections.
But to view it, the borrower’s identity needs to be reviewed.
With Lin Ran’s identity, passing this security clearance is naturally no problem.
“Of course, but the set I want to borrow is the one in IBM’s hands.” Lin Ran smiled and raised his wine glass.
That’s right; the ones universities have are castrated versions.
Not to mention universities, even in later years up to 2020, the “MIT Radiation Laboratory Series” you could find in public were all castrated versions.
Basically all the theoretical content is there, at most something like the volume “Microwave Magnetrons” would have less content.
But the applied content that was castrated was severe.
Many of America’s core applications in radar during World War II were castrated.
Only a very few institutions possess the complete version.
One of them is IBM.
Although IBM mainly focused on punch card machines and early computing equipment during World War II, IBM engineers had extensive technical exchanges with the MIT Radiation Laboratory, especially in electronic signal processing and computational support.
After the war, IBM quickly entered the electronic computer field and absorbed many achievements from radar technology, such as high-speed circuit design.
Lin Ran even knew that this set of books was kept in IBM’s old nest, Yorktown Heights in New York.
Although it has MIT in the name, even MIT doesn’t have the complete version.
The complete version of this set of books is only available at Lincoln Laboratory, Lawrence Berkeley National Laboratory, America National Bureau of Standards, Los Alamos National Laboratory, and IBM.
Among them, the one Lin Ran can most easily obtain is naturally IBM.
Promoting artificial intelligence is one aspect; getting the complete version of the “MIT Radiation Laboratory Series” from IBM is an even more important aspect.
Sending this set of books back to China can save a lot of time in any aspect.
This set of books from the 1940s, not to mention 1963, even sixty years later has its unique value.
For a country developing radar, missile guidance, and communication technology, this set of books provides a systematic theoretical framework, design methods, and engineering practices, which can significantly reduce trial and error costs.
“Professor, this probably isn’t something you should know, right?” Watson continued.
Lin Ran pressed his hand downward: “Watson, calmness, you must know about my relationship with Jenny Hearst.
Both being in the publishing industry, isn’t it normal for the Hearst family to know that McGraw-Hill published two versions of this set of books back then?
Is that strange?”
Hearst is also a media empire; McGraw-Hill needs to place advertisements on Hearst family newspapers or magazines to promote their textbooks or professional books.
There is business dealings between the two sides, so it’s not strange to know that the “MIT Radiation Laboratory Series” has two versions.
Lin Ran continued: “I know what you’re worried about, but your worries are completely unnecessary.
You send it to Redstone Arsenal, and I will only look at this set of books at Redstone Arsenal, not take it out of Redstone Arsenal.
You can even have McNamara vouch for me.”
No matter how clever you are, you can’t guess the existence of the gate.
Hearing this, Watson instantly relaxed: “Professor, you should have said so earlier.”
Lin Ran’s meaning was to go through the open and aboveboard program.
First have IBM apply to the Department of Defense, then after approval from the Department of Defense side, send this set of books to Redstone Arsenal.
Throughout the process, Lin Ran doesn’t contact it, and it won’t leave the Department of Defense’s jurisdiction.
After all, although Redstone Arsenal belongs to NASA, a large part of its security is still handled by the Department of Defense.
Watson immediately relaxed, thinking I thought you were going to do some irregular operation.
For example, have me send this set of books to your home, then somehow the Russians use a miniature camera to turn it into miniature film rolls, which end up in Moscow.
As long as you’re not planning to play like that, it’s fine.
Watson thought to himself: “Professor, no problem, I will definitely get it done.”
Lin Ran smiled and nodded, thinking to himself: “No matter what, you couldn’t imagine that I have a gate.”
Two months passed in the blink of an eye, and Lin Ran commanded the entire team with tyrant-like efficiency.
Even though he only spent half a day or even less at the Georgetown Project every day.
But he forced Calvin to collect Russian-English technology literature, statistics high-frequency word pairs, personally guided Cuthbert and Peter to modify the IBM 7090 program, adding a frequency-based word selection module.
Whenever someone made a mistake, he was merciless: “Can’t even do this? Are you engineers or waste?”
Two months passed in the blink of an eye.
The inspection team was personally led by Secretary of State Rusk and Secretary of Defense McNamara.
The Georgetown Project members who learned the news in advance couldn’t help but sigh that the professor was indeed a professor.
“If it weren’t for the professor, at most Colonel James Reed would come to check our project.”
“Yeah, only the professor has such influence.”
“Secretary of State and Secretary of Defense, this is manned spaceflight treatment.”
Originally at most a colonel would come, now it’s civilian officials comparable to a five-star general.
In a laboratory at Redstone Arsenal, the air was filled with an atmosphere of tension and expectation.
Fluorescent lights cast bright and cold light, illuminating the protagonist in the center of the room: a massive translation machine with several IBM machines linked together, its metal shell covered with knobs and switches.
If not for the Deep Blue experience last year, it really couldn’t be done in just two months.
The machine was surrounded by a low hum, mixed with the heat from electronic equipment and the unique smell from the air conditioning in the distance to keep the room temperature constant; all this witnessed the team’s continuous improvements and debugging of this machine.
Calvin stood beside the machine, posture upright, eyes showing focus and confidence.
As the leader of the improvement work, he knew the machine’s performance like the back of his hand.
Today, he would demonstrate the achievements to this group of special visitors: officials, Department of Defense staff and experts in the translation field.
They stood on one side of the laboratory, each with different expressions: some curious, some skeptical, even a hint of indifference.
Many of them were old acquaintances.
Like the translation staff, they had long known about the Georgetown Project.
Everyone started with expectation.
Of course expectation, this thing could reduce their workload.
But after waiting and waiting, they never got the Georgetown Translator equipped.
Including that they had also gone to Georgetown University before, the evaluation was that it was a pile of shit, cannot be used.
The reason everyone came this time was completely for Lin Ran’s face.
If not for the professor, they wouldn’t come.
Of course, among them, the one most confident in the Georgetown Project’s rebirth was McNamara.
“Professor, why don’t you introduce it personally?” McNamara asked Lin Ran standing beside him in a low voice.
“I just provided a little auxiliary help; the credit all goes to the researchers at Georgetown University and IBM.” Lin Ran said.
McNamara pretended to be dissatisfied: “Professor, you’re too modest.
I’ve seen the statistical methods you designed for them; without your participation, with Georgetown this group of idiots, em this group’s brains, they probably couldn’t think of it even in ten years.”
While Lin Ran was whispering with McNamara, in the crowd, linguist Paul Gavin PhD working for the Department of Defense frowned slightly, his pen lightly tapping his notebook; he had always reserved attitude toward the machine’s actual capabilities.
Even with the famous professor personally involved.
Beside him, Colonel James Reed, who had always been responsible for approving the Department of Defense budget for Georgetown and tracking the project, had his arms crossed, expressionless, uniform crisp.
Rusk held a notepad, his gaze shifting between the machine and Lin Ran, wanting to interject between Lin Ran and McNamara but not sure what to say.
Calvin cleared his throat, voice steady and strong: “Gentlemen, today you will witness the latest progress in machine translation. Under Professor Lin’s guidance, we have significantly improved the system’s accuracy and ability to handle complex language structures.”
Gavin PhD pushed up his glasses and asked first: “Improvements? The previous version performed terribly on idioms and syntactic ambiguity. How did you solve it this time?”
Calvin confidently said: “We introduced a frequency-based statistical model; by analyzing the bilingual corpus, the system can now better distinguish word meanings and handle context-dependent translations.”
Colonel Reed interjected coldly: “Sounds good. Let’s see the actual effect. Time is tight; we want results.”
Calvin nodded and turned to the machine’s console: “Of course, we’ll start right away. First, a simple demonstration.”
He skillfully typed a sentence of Russian on the keyboard: “Мыговоримомире.”
The machine immediately started operating, internal parts making slight clicking sounds and humming; a few seconds later, the screen displayed the English translation: “We speak about peace.”
A murmur rose in the room; Gavin PhD raised an eyebrow: “Not bad, but that’s just a simple sentence. What about harder ones?”
Calvin was prepared: “Of course. Let’s try an ambiguous sentence.”
He input another Russian sentence: “Ончитаеткнигу.” The machine quickly output: “He reads the book.”
Gavin PhD nodded but was still not satisfied: “What if the context is different? For example, ‘книга’ refers to a ledger rather than a book?”
Calvin’s fingers slowly tapped the keyboard, inputting a new sentence: “Бухгалтерчитаеткнигу.”
After a slight pause, the machine displayed: “The accountant reads the ledger.”
Gasps of astonishment filled the room. Colonel Reed’s arms slowly lowered, a flash of surprise in his eyes; this was completely different from the Georgetown machine he had seen before: “How did it judge that?”
Calvin explained: “The system can now combine contextual clues and probability models for inference. ‘Бухгалтер’ (accountant) indicates a financial context, so it chose ‘ledger’ over ‘book’.”
A previous translation staff member pondered: “This could be revolutionary for intelligence work.”
Gavin PhD approached the machine, more interest in his eyes: “I want to see how it handles more complex sentences, like multiple clauses or idioms.”
“Let me try.” Gavin sat at the input device.
Calvin nodded: “No problem.”
Gavin PhD thought for a moment and input: “Времялетит，когдатывеселишься.”
The machine quickly output: “Time flies when you’re having fun.”
Colonel Reed remained unmoved: “That’s nice, but we need it to handle military terms and technical vocabulary. Our needs are very specific.”
Calvin looked him in the eye: “Gavin PhD, you can try some professional vocabulary; we have started integrating a professional vocabulary list, including military and scientific terms.”
Gavin PhD thought for a moment and input: “Ракетаоснащенаинерциальнойнавигационнойсистемой.”
The machine translated: “The missile is equipped with an inertial navigation system.”
A flicker in the colonel’s eyes, rarely showing approval: “Quite precise.”
The room was quiet for a moment; everyone felt the weight of this moment.
The officials exchanged glances, doubt gradually turning to optimism.
A Department of Defense expert, a sharp-faced madam, spoke: “Professor Calvin, your work may change our operational direction.
This machine’s speed and accuracy have indeed improved greatly compared to the past.
How soon can it be applied on a large scale?”
Calvin paused, cautiously answering: “With more resource support, we can start scaling the system in six months. However, we still need to continue optimizing the algorithm and expanding the language database.”
McNamara clapped, turning to Watson: “Mr. Watson, congratulations; you will receive an order from the Department of Defense.”
Then to Lin Ran: “Professor, I knew you could do it.”
Finally to Calvin: “Professor Calvin, the Department of Defense will contact you later.”
Lin Ran said:
“Everyone, don’t be too optimistic.
Georgetown can only be said to be usable; it’s still a long way from being good.
IBM needs to improve the machine’s performance; the better the performance, the higher its translation accuracy.
Secondly, Georgetown University needs to continue along the technical route I designed, constantly excavating its potential.
But this machine translation technical route has limits.
Only updates in statistical methods can achieve better translation effects.
However, after you reach the limit of this technical route, I suggest that during this process, Georgetown University should recruit more PhDs in statistics and theoretical mathematics to participate in this project.
Finally, for the Department of Defense side, when using the Georgetown machine for Russian translation later, it still needs to rely on manual review.
It can help improve your efficiency, but the degree of efficiency improvement is limited; you cannot completely rely on the machine.”
Actually, Georgetown’s effect is similar to the earliest translations, purely mechanical translation.
It doesn’t optimize according to language characteristics.
It’s just because English and Russian are both related to the Latin language family, just related, not low correlation.
But definitely much simpler than translating English to Chinese.
“The current version of the Georgetown Translator will perform relatively well in the two specific fields of military and technology; in other fields, and for complex sentences and non-professional contexts, its error rate will be relatively high.
Russian is a highly inflected language, with nouns having six grammatical cases, verbs having complex tenses and aspects, while English grammar is simpler, relying merely on word order and prepositions.
The Georgetown machine still has great difficulties in accurately handling grammatical case and word order adjustment.
We currently only alleviate the polysemy problem through probability tables and context windows; the rules system ensures basic grammar accuracy.
Simple sentences can reach about 80% to 90% accuracy, but complex sentences, like nested clauses, are still prone to errors.
For news, diplomatic statements and other texts with clear structures, its effect should be good.
But for dialogues involving a lot of spoken language, it’s hard to accurately capture subtle differences, and the error rate may reach 50%.
I think everyone needs to maintain a cautiously optimistic attitude.”
After Lin Ran finished, Rusk spoke: “Professor, this is already enough.
For the current needs of the White House and Department of Defense, it only needs to batch process written texts like news, reports, etc., which is already a great help.
For telephone conferences and such, we didn’t expect it to achieve accurate translation anyway.”
Indeed, as a project that was almost sentenced to death, cut after one year, even considered part of the AI winter, now saved by Lin Ran in two months and possessing practical application value.
This is already a leap.
The visitors left one after another, excited whispers echoing in the corridor.
McNamara insisted on sitting in Lin Ran’s office for a while.
“Professor, regarding the “MIT Radiation Laboratory Series” you want to borrow, I have approved it here.
IBM should approve as early as the day after tomorrow; I’ve also asked Lincoln National Laboratory to provide you with some radar design proposals from that time.
Just right, you can also look at this content and see if there’s any inspiration you can provide us.” McNamara said while drinking tea.
Lin Ran thought to himself, damn, good guy!
