WASHINGTON, United States of America – The controversy over Facebook data gathering has a Canadian backstory far older than the adventures of the young whistleblower from British Columbia who worked for the contentious firm Cambridge Analytica.
This story involves the firm’s billionaire bankroller.
Robert Mercer is the 71-year-old funder of an assortment of right-wing causes, ranging from Cambridge Analytica to Breitbart News to Republican campaigns. The story of his wealth includes a breakthrough in the 1980s and Canadian bilingualism.
Back then, Mercer was a talented mid-career computer programmer at IBM. He’d grown up in New Mexico with an interest in computers, and as there were no courses in the field in the 1960s at his college, he sought experience writing programs for a mainframe in the weapons lab at the Kirtland Air Force Base.
There, he said in a 2013 lecture, he learned he loved everything about computers. But he also learned he did not love big government. He described that summer as a formative experience on his political path. Frustrated with a clunky program that calculated the effect of fusion bombs, he recalled rewriting the program to make it 100 times faster.
“Then a strange thing happened,” Mercer said in a 2014 speech.
“Instead of running the old computations in one-100th of the time, the powers that be at the lab ran computations that were 100 times bigger. I took this as an indication: that one of the most important goals of government-financed research is not so much to get answers as it is to consume the computer budget.
“Which has left me ever since with a jaundiced view of government-financed research.”
He graduated college in 1972 and took a job at IBM. There, he helped develop the field of computer-generated translation. He explained in a paper co-written with colleagues that computer-translation efforts are almost as old as the modern computer itself.
In the 1940s, fellow-travellers had sought to have machines translate language. But they stumbled into two main obstacles: the weak processing power of older computers, and the shortage of translated text in digital format for programmers to study.
Advances in computing gradually solved the first problem.
The second problem was solved by a tip in the 1980s about where to score massive amounts of translated text, which would allow the programmers to detect patterns in data and develop algorithms based on that.
That news came from older IBM colleague John Cocke.
“John was on a plane and … he struck up a conversation with the guy next to him and then suggested they have a drink together,” Mercer’s colleague Peter Brown recalled at the 2013 Conference on Empirical Methods on Natural Language Processing.
“Before he knew it the guy was telling John about the proceedings of the Canadian House of Parliament which were — and probably still are — kept in computer-readable form in French and in English.”
Canadian government employees had already done the work.
They had translated millions of words spoken in Canada’s Parliament from English to French, and vice-versa, at a reliable quality, with literal and figurative meanings of phrases swapped between languages, all of it was in the public domain, available for use by researchers.
“That’s what I liked about the Canadian Hansards data,” Mercer told the 2013 conference.
“I think that’s what appealed to John about it as well.”
Mercer and his colleagues scooped up about 100 million words’ worth of data. That’s how one of the landmark research papers in the field of computer translation wound up including the words, “Bobby Orr,” and “fuddle-duddle,” the insult phrase made famous by former prime minister Pierre Trudeau.
One of their papers said the algorithm reduced the work flow of a translator by 60 per cent, requiring 776 keyboard strokes to repair mistakes within a certain sample, versus 1,916 keystrokes to start from scratch.
When it granted Mercer its lifetime achievement award decades later, the Association for Computational Linguistics noted the thousands of research papers that cite his work, and credited him with technologies we now use every day.
“(Their approaches) now dominate the field of machine translation,” the ACL said, “and provide the underpinning of many of the tools that people now regularly use, such as speech recognizers on mobile phones, context-sensitive spelling correction, and web-based machine translation systems.”
His research led to tremendous wealth.
It turns out that his ability to detect patterns in linguistic translation was useful in stock trading. In 1993, Mercer and Brown got recruitment letters from the hedge fund Renaissance Technologies.
They initially tossed the letters, but later had a change of heart, as Mercer struggled with college tuition bills for his daughters. They accepted a 50-per-cent pay raise, changed jobs, and eventually recruited other members of their old IBM team.
The finance work is a little more hush-hush but Brown described it as based on their previous research: “From building speech-recognition systems and translation systems … we definitely used that skill set.”
Mercer and Brown wound up running the fund after its founder retired.
In recent years he’s become famous for his political investments: the research group that looked into Hillary Clinton’s alleged conflicts of interest and bankrolled the book “Clinton Cash,” the Government Accountability Institute; Breitbart; and the campaign of Donald Trump.
He founded Cambridge Analytica; teamed up with young researcher Christopher Wylie, a Canadian Liberal and spent millions to amass voter data, including from Facebook users.