An article in the Daily Telegraph two days ago written by Charles Cumming caught my eye. It is titles “The Idea of AI Stitching Together a Book is Still Novel”.
This article is part of a series by Harry de Quetteville, the Telegraph’s Special Correspondent, Technology on the Future of Work. During the series he asks holders of representative jobs what they do and then he gives an assessment of the risk of the job being performed by AI.
The chart below, prepared by ONS is a scatter blot of nine groups of jobs against the risk that they well be automated. (Sorry, the text didn’t copy with the chart. The far left vertical line is 15% and the far right vertical line is 75%. The top group 1 includes managers, group 2 is professionals, group 3 is technicians, group 4 is administrators, 5 is skilled trades, 6 is caring and service, 7 is sales & customer service, 8 is machine & plant operators, 9 is elementary operations.)
Amazon’s page on the author says, “Charles Cumming is a British writer of spy fiction. He was educated at Eton College (1985-1989) and the University of Edinburgh (1990-1994), where he graduated with 1st Class Honours in English Literature. The Observer has described him as ‘the best of the new generation of British spy writers who are taking over where John le Carré and Len Deighton left off’.”
The article says, “What does your job involve?
I write spy thrillers. I am a procrastinator but I try to write at least 1000 words a day. Of course then I delete 900. It’s hard to seize full control of characters and the suspense of the plot; there are perhaps one in seven great days. When I’m starting a new book, I feel my way into it, sketching out a general idea of what’s going to happen and where it will take place. Then I plot it out, make research visits, and read about the politics and espionage on the subject. That leads to three or more ideas, at which stage there could be a lot of plates spinning: I have to remember my characters and be true to them. And I sit with those components for the year or two the book takes to write.
Has Covid affected life?
Writing is solitary anyway. But contemporary writers do now have to make the decision to ignore Covid or incorporate it. For my latest, I’m incorporating.
How long have you been doing it?
What training did you get?
None. Just reading other books, and a sense of self-confidence that I could write decent sentences.
How much does it pay?
They say there are 50 writers in the UK who make a living solely from writing books. Everyone else has to have another job – screenwriting or teaching or whatever. I’ve had years where I’ve made 30k and one Hollywood year where I made 350k. Income is so variable it’s hard to plan, to know how much you’re going to have at any point.
What took longest to learn?
Understanding that it’s a business, that you’re only as good as your last book and will be jettisoned if your books are not making money. I certainly began with an artistic hat on, but I realised that books will be read as entertainments, as stories and escapism. To imagine they are anything else is a pipe dream. As long as people appreciate them, though, I’m happy.
What is the most boring bit?
Will you always do this?
If you gave me a vineyard with no frosts, I would certainly consider a sabbatical. I’m not one of those writers who must write.
Do you think your job will be the same when you retire?
There is already screenplay software that will fill in a lot of structure, and similar for fiction. But the idea of AI being able to tie it all together feels some distance off.
Analysis (at this point Harry de Quetteville takes over)
Computer writing – or at least manipulating words – became one of 2020’s hottest technological topics. In May, the American lab OpenAI announced GPT-3 software which produces almost eerily good text. GPT (which stands for Generative PreTrained Transformer) is itself an extension of Natural Language Processing (NLP) which uses a brain-like computing process known as deep learning to extract information in documents and which itself has come on in leaps and bonds. The big question is, do computers ‘understand’ what they are ‘reading’ and ‘writing’? The answer is no, at least for the moment. But does that matter, given that their appearance of understanding is increasingly perfect?
Would machines actually do better?
The dramatic improvement of AI’s handling of language suggests it’s not impossible.
Bottom line: Risky
We could well see a GPT-3 bestseller within three years. Implications for this job group: Group 3 includes many creative types, and whatever instincts may tell you, they are far from immune. The ONS for example, thinks artists (35pc at risk) are more vulnerable than paramedics (28pc). Most vulnerable of all? Sports players – half of whom are expected to lose their jobs.
ONS jobs risk estimate: 35c.”
For those of us who write, this sobering bit of information is unlikely to be welcome in the midst of a pandemic.