Market Leader (https://www.warc.com/Store/ProductInfo/4642) is “the strategic marketing journal for business leaders”. It comes out once a quarter and they’ve asked me to contribute a column.
If you’re a WARC subscriber you can read it on their website but they’ve allowed me to stick the ‘original submitted version’ on here, as long as it carries the disclaimer “Unedited Version” and the credit “Reproduced with permission of Market Leader, the strategic marketing journal for business leaders. To subscribe visit www.warc.com/bookstore © Copyright Warc and The Marketing Society.”
So, here it is, without the benefit of editing and, therefore, probably, with a high risk of slander, libel and typos.
#MarketingAI – if it’s not the biggest hashtag in your stream right now it will be soon. All the large tech companies and half the world’s start-ups are about to knock on your door to tell you how their algorithms can transform your business. A tiny percentage of them are right. Here are four things to think about.
1. Don’t Worry About the Words, Worry About The Data
They might say AI, they might say Machine Learning or Deep Learning. They might have invented some dreadful neologism like Deep Marketing or some acronym that munges AI or ROI together. Someone will do that. Don’t worry about it. All that’s happened is that computers have suddenly got smart enough to do human-like tasks because businesses like Google and Amazon have access to vast amounts of data about stuff humans do. Google are really good at ‘natural language’ because they’ve got astounding amounts of human language being typed and spoken at them every day. Combine that with very large numbers of very fast computers and the computers can start to learn for themselves from all the billions of examples. (I’m simplifying dramatically). That’s the capability they’re trying to sell you. It’s worth remembering that you probably don’t have anything close to that amount of data. It might seem like all your transactions and insights and products and everything is a lot. It isn’t. In order to tell you anything useful they’ll probably have to combine it with other data. Or they’ll be manipulating your data with an algorithm that was trained on another set altogether. You should be asking questions about this. These businesses are hungry for data, your data, Their algorithms have to be fed. Make sure you keep a close eye on what happens to it. You don’t want to be paying them to make themselves smarter.
2. You Can’t See What’s Going On
It’s hard to see what Machine Learning systems are actually doing. They don’t work in ways that make ordinary ‘common sense’ to humans and they do lots of things very quickly, so it’s hard to spot mistakes. That’s even more true of the biases that get built in unthinkingly; the assumptions that have lain unexamined in your business for years, or things that are even worse than that. Many data training sets are built on majority populations. You don’t want to be standing in front of your board or a press conference blaming an algorithm for your accidentally racist targeting strategy. As The Atlantic magazine pointed out last year “Flickr’s auto-tagging of online photos label pictures of black men as “animal” or “ape” and researchers determined that “Google search results for black-sounding names are more likely to be accompanied by ads about criminal activity than search results for white-sounding names.” Read Weapons of Math Destruction by Cathy O’Neiland imagine those things happening inside your business.
3. Build A Machine, Get Replaced By A Machine
All the prognostications about the robots taking our jobs focus on the jobs everyone imagines are easy to computerise. People who read stuff and make some decisions, within fairly limited parameters with reasonably little independent creative thought. We dress it up with a lot of hoo-ha but that’s true of quite a lot of marketing/communications jobs too. The more you’ve turned your marketing operation into something science-y the easier it is to replace with a computer. Even if it’s half-baked pseudo science. The computer doesn’t care if your formula is rubbish, if you’ve got a formula it’ll follow it. The more you use things like pre-testing to make decisions the easier you are to replace with a machine.
4. Cheap, quick, algorithmic – pick three.
The ‘creative’ side of the business is not immune either. Computers are getting pretty good at producing perfectly acceptable art, music and writing. None of it’s genius but all of it’s cheap. It’s just the sort of thing you might need as a soundtrack for your product videos or as a way of instantly translating your brochures. To explore how easy this is I built a little twitterbot that generates a new corporate tagline on the hour every hour. It can do thousands a second. Some of them are good. All of them are cheaper than paying an agency. (Have a look @taglin3r.) If I were one of your creative agencies I’d be getting out of the commodity side of things – the art-working, retouching, mood music side – and retrenching to the stuff that only authentically, originally creative people can do. If I were you I’d be doing the same.
Follow @fourthingsabout on twitter for a stream of links and articles related to this quarters topic.
Russell Davies is Chief Strategic Officer at BETC, a contributing editor for Wired and a relentless mucker-about on the internet.