Gary Kasparov and the future of PPC
I’ve been reading a book recently, written by a futurologist. He was talking about artificial intelligence and a series of chess games between humans (usually called Gary Kasparov) and computers.
Initially, Kasparov was able to interpret the chessboard in a way that a computer couldn’t, however eventually he was beaten by a really powerful computer.
I thought that was the end of the story — it wasn’t.
It appears that a new super-race of chess players called ‘centaurs’ has emerged from the ashes and they are better at chess than supercomputers. Basically, they are computers that play chess, but with a human sat alongside them, over-ruling if they see an opportunity that the computer doesn’t.
It’s easy to draw comparisons between chess and PPC.
Currently, the ‘intelligence’ of PPC tools is limited — you can successfully use them to offer suggestions for you to accept or reject, but they aren’t at a stage where you’d want to reverse the roles, with you overruling the computer by exception. That said it seems highly likely that one day soon(ish) we’ll reach the point where the machine takes the PPC lead, with the human only stepping in on occasions.
Effectively, PPC accounts will be managed by centaurs (I prefer the fantasy term to the more sci-fi ‘cyborgs’ as I’ve watched too many sci-fi films where cyborgs run amok — centaurs tend not to do that).
In chess, there are no factors outside the game itself, so it seems likely that at some point, computers will have learned the game to the point where they won’t benefit from our assistance. In PPC however, until computers can foresee the downtime of websites, the stock levels of various products and several other factors that affect performance, their ability to predict the future value of a click will always be limited.
There is another advantage that computers have over humans — we really aren’t very good at being objective. Cognitive bias is an interesting area, and whilst I know even less about it than chess, what I do know concerns me. In fact, it makes me question my own ability to manage PPC accounts (and I’m egotistical enough to think that I am one of the best in the business).
There are a number of forms of cognitive bias, but perhaps the simplest one is confirmation bias. Basically, people interpret data in a way that aligns with what they already believe. Put simply, if you believe something is true, you believe any evidence that supports your belief and disregard anything that contradicts it.
There are plenty of examples for this — people believe what they believe, and are remarkably resilient to any evidence that contradicts what they believe. Donald Trump does nothing (or everything) wrong, Brexit is or isn’t a disaster, Global Warming is or isn’t a coming apocalypse, people really are or aren’t abducted by aliens on a daily basis, and the earth is or isn’t flat (it isn’t, though I admit, I tend to disregard any evidence to the contrary).
Another strange quirk of the human condition is that we believe things because we want them to be true (or fear them to be true — never quite sure why that works either).
We are all susceptible to these factors and it makes me wonder just how objective we can be when reviewing our own decisions retrospectively. If I push a bid up on a set of keywords because I believe the performance warrants the higher bid, I want to see an improvement in performance as a result. I’m expecting to see one. So, will I look at the data, and accept it if it supports this, whilst rejecting it as ‘random variation’ if it doesn’t?
I’d like to think that I don’t. As far as I can see, being aware of the risk is the best (and perhaps the only) defence against it. But this is where computers have a distinct advantage over us. They don’t have egos. They don’t care whether they said to do one thing last week and are now saying the opposite. They are completely objective.
50 years from now, I believe that data will link seamlessly in the way that we are starting to see with hardware since the invention of the USB. If you have a USB port, you can plug almost anything into your computer (unless you’re on a Mac!). Is it such a stretch that software, and therefore data, will move this way in the future? If so, then many of our advantages over computers will disappear, and artificial intelligence will take care of the rest, as it has in chess.
Ultimately, computers will be able to do anything that we can — and I very much doubt that the ‘human condition’ will change significantly (it hasn’t in recent millennia).
On top of this, their ability to react to data will be better than anything we can dream of. If you’ve sold out of one of your best-selling toasters, pull the bids on ‘toasters’ back a bit. If you’ve sold out of an unpopular model, don’t change the bids. If the site goes down, pause the account. If a competitor changes their prices, either change your bids, or potentially your prices. And this is just scraping the surface of the possibilities.
Will machines eventually replace PPC Analysts? How can they not? We are already on a path towards a centaur-like relationship and I suspect that by the time I retire, PPC Analysts as they exist now will have been consigned to the history books. But for the foreseeable future, I maintain that humans taking the lead, with the appropriate automated assistance, will still produce the best results.