Siri thinks she’s a big deal. As far as I’m concerned, she’s all talk and no action. Sure, we’ve made amazing progress in natural language processing and speech recognition, but none of that has made a dent in the reasons that many people still hate many software applications. As a stand-in for all the software that people hate, I’m going to focus here on Microsoft Word (coincidentally). But first, I’d like to break this up into three types of tasks:
- Stuff that Siri/Watson/Google can do already, more or less
- Stuff that Siri/Watson/Google can’t do now, and might take many years to learn to do
- Stuff that why-can’t-MS-Word-do-already, dammit, dammit, dammit!
Stuff that Siri/Watson/Google can do already, more or less: On Friday, a co-worker asked me to confirm that a cannula is classified by FDA as a Class I medical device (lowest level of risk and regulation) but by the EU as a Class IIa. We use this information to determine how much “technical documentation” the cannula requires, and whether the technical documentation has to be submitted to someone or just stored somewhere in a nice, warm drawer. Anyway, this confirmation is just a lookup task, and Siri/Watson/Google could be programmed to handle it, but for now it’s probably cheaper and easier just to ask me.
Stuff that Siri/Watson/Google can’t do now, and might take many years to learn to do: One of my company’s tasks is to help our clients find “predicate devices” for FDA clearance and “comparator devices” for EU approval. In a nutshell, we have to prove that the client’s device is equivalent to a medical device already in use, which might allow the client to forego certain kinds of clinical testing before their new device is sold. This part of the technical documentation might sound a little like marketing (“This revolutionary, new shoelace works exactly the same as all the old shoelaces”). But it’s actually pretty straightforward technical writing, where you dig through lots of academic and engineering source material to create a clear, technical narrative. Siri/Watson/Google are still a long way away from this sort of analysis and synthesis of source texts.
Stuff that why-can’t-MS-Word-do-already, dammit, dammit, dammit!: The article title is one example. Here are a few more commands that people should be able to speak into a microphone:
- Word, standardize the bibliographic citations to all use APA volume (issue)
- Word, use British English spell-checking only for quotes that were published in the UK.
- Word, adjust the table for maximum readability – font size, paragraph spacing, cell margins, table borders, and page margins.
These tasks don’t require incisive editing or the creation of new content. They’re all just manipulations of existing text to make it best suit the reader’s needs – exactly the kind of thing your boss or editor might ask you to do. For some tasks, we write macros and other kludges (yes, a macro is a kludge) to make Word do stuff that’s it’s too dumb to know how to do without our help. And even though Siri/Watson/Google could be programmed to do these tasks specifically, and might be able to use natural language processing to understand new tasks, there’s no hint – yet – that MS Word is going to make these tasks intuitive for people who aren’t trained tech writers.
And remember, these are the non-creative tasks that artificial intelligence should already be able to handle in 2017. Siri/Watson/Google can explain the connection between a cannula and a cannoli (yes, there is one), but they can’t explain why a new cannula with a built-in sterilizer is exactly the same as all the old cannulas. So yeah, maybe Deep Blue can replace Gary Kasparov and Watson can replace Ken Jennings, but they can’t come close to replacing tech writers.