The Last Word On Nothing

For two years, I have felt like Eloise in the beloved 1950s children’s classic of the same name. A sudden need for shared office space led me to a reasonably priced desk in downtown Ottawa, but the exact location stunned me. Much like the entitled Eloise who lives in New York’s Plaza Hotel, I also have been in a “room on the tippy-top floor” of a grand hotel—one built within five years of the Plaza: Ottawa’s Château Laurier.

Now, it happens that the top floors of the Château are not penthouse suites, but more like the cramped storage spaces in the rafters that could never be passed off as hotel rooms. Where my office lies, once were the women’s dormitories for chamber maids. On the same floor were traveling salesmen’s rooms, with a discretely curtained off area where they could display their wares on a table.

The Château Laurier was opened by the Grand Trunk Railway in conjunction with Ottawa’s rail terminal across the street in 1912. There was no real festive opening in the end, because the president of the railway sank on the Titanic on his way here, but the first registered guest was Prime Minister Wilfred Laurier, after whom it the hotel is named.

Continue reading →

Earlier this month, the International RBD Study Group published its findings that a certain class of sleep disorder is strongly linked to the risk of developing Parkinson’s disease later in life. As the group’s name implies, the disorder is known as RBD – REM sleep behaviour disorder, which causes you act out your dreams to sometimes fatal consequences for your sleep partner. All the people in the study suffered from this disorder, and after a 12 year followup, they were found to have a high chance of developing Parkinson’s or dementia.

It may not sound like it, but this is actually good news. By the time most people are usually diagnosed with Parkinson’s, the neurodegenerative disease has already done irreversible damage to their brain. Researchers have spent years looking for a reliable early warning sign, so that treatments can become prevention. They might have just found it in RBD.

Continue reading →

Like most journalists, I dread transcribing interviews. I can’t afford to pay other people to do it, so I’ve been experimenting with computer programs that use artificial intelligence to transcribe for me. Last week I tried one of the more advanced transcription programs, Otter, and its performance was nearly flawless.

I felt an odd sense of loss as I watched perfect phrases tumble down my screen at close to the speed of real speech. I wasn’t sad about the hours I’ll save hand-correcting transcripts. Instead, I mourned the bizarre, sometimes profane errors of my older transcription service, Trint.

I am generally more likely to cover the evolution of sponges than a Presidential race or sex scandal. But Trint seems to want me and my interviewees to talk about what everyone else is apparently talking about: politics and sex.

Trint inserts “Trump” and “Melania” into my transcripts seemingly at random, as well as words like “dick.” “There’s a lot of data on this, in terms of being a dick,” read the program’s mis-transcription of one of my recent interviews with a soft-spoken botanist.

I suppose Trint’s proclivities should not surprise me. Like all speech recognition programs, Trint uses statistical algorithms trained on large archives of recorded human speech to predict what is being said. Machine learning technology has a well-documented tendency to reflect the uglier aspects of society, including our racial biases, and Trint, launched in 2016, has been reared on our society’s collective chatter since Trump’s election.

I am tired of spending hours cleaning up my transcripts for the fact-checkers, and eager to find a more accurate program. But I will miss some of Trint’s bizarre word choices and its constant attempts to help me join the crowd. Trint’s errors can also be remarkably creative, like the brilliant AI-generated cookie names published on the blog AI Weirdness: Hand Buttersacks, Apricot Dream Moles, and Walps.

Can artificial intelligence be talented? Does it have anything important to say? I have no idea, but before abandoning Trint I want to celebrate its weirdness with the following poems. I didn’t write these poems, Trint did. It is not my fault that they insult Melania Trump or read like dystopian Radiohead lyrics. I am merely Trint’s transcriptionist. Please share your own AI-generated poems in the comments!

Trint’s mistakes – suggestions? Strokes of genius? — are in bold.

Emily Underbite

Hello, Siberia

It’s Emily Underbite

I’m a relic

I haven’t found a gold star yet

My dream has no leverage

I hurt like imitation

Electoral thinking

What organism has analogous or homologous structures and functions to Melania’s? A slug, a lobster?

People have a lot of trouble with electoral thinking, which is essentially understanding the difference between two sponges.

I mean, what can you expect from a sponge?

Continue reading →

Yesterday I was thinking about how much I love the central dogma of molecular biology.

“Central dogma” is a funny name for it. It sounds like it has something to do with religion, but it’s not; it’s just the thing that makes all living cells work. A cell has DNA; that DNA has a code, which gets transcribed into the almost-the-same-but-not-quite code of RNA; and then that RNA gets read and used as a template for building proteins. This is happening all the time in all living cells–the cells in the skin on your right ear, the cells in your pancreas, the cells in the rootlets of the plants outside, the cells in a polar bear’s liver.

Continue reading →