AI in the Near Future: Dr Merlin Stone on the Parallels of Marketing and Medicine
World-renown researcher, Dr. Merlin Stone, talks with Zuant CEO, Peter Gillett, about the future of AI and the link between marketing and medicine. Learn how data analytics is being used to diagnose and treat today's toughest health issues as well as marketing's biggest challenges.
Why Artificial Intelligence? Why Now?
Peter: Why are so many people looking at AI right now?
Dr. Stone: It all started in the 1960s when the university philosophers got interested in how computers were replicating human decisions. Edinburgh University was the home of that work, but I was at Sussex as a student and there were philosophers interested in AI because they were interested in the power of computers to do better than the human brain. We've seen the story of Big Blue playing chess and its serious application for the diagnosis of cancer. So we know that AI can deal with some of the most complex and sophisticated diagnostic problems to help us make decisions. On the flip side, marketing problems are often soft problems such as uncertainty about data quality of. We aren't certain about the theories or the ideas we use to turn data into a diagnosis of what's going on, or we're not certain about what is the best solution when we have got the diagnosis. That's the world of marketing, but it's also the world of medicine.
As a youngster, I became aware of the idea of AI with HAL in Stanley Kubrick’s 2001 A Space Odyssey, but you pre-date me a tad. Let’s go back to university when you became aware of AI as an undergraduate.
Well, back in the swinging 60s I met lots of interesting people who were philosophers very interested in AI. One of those was a chap called Don Michie who was a British researcher in AI. During World War II he worked for the Government Code and Cypher School at Bletchley Park, contributing to the war effort to solve "Tunny," a German teleprinter cipher. I think the idea of AI then was actually a misnomer; it was more very clever programming for a long period until we got to the world of neural networks. This was when computers could teach themselves things as opposed to clever programmers looking at all the things that they could see and helping what was basically a dumb machine.
That area is also related to another area of my research, which is how do you make sense of information where there are lots of experts? And that's not the AI bit; it’s actually another area of interest to marketers, which is where in that wood can you see the wood? That's not just a question of AI, it’s a question of perception: How do you know what's really going on? If your sales go down, is it because your customers are leaving or because your products are rubbish? The simplest question, but actually often it would be a combination of both - you need a sophisticated analysis tool to find out. We're talking about big data now.
AI vs. The Dark Side of Marketing
Peter: What are some current, real world examples to illustrate this?
Dr. Stone: It used to be fairly simple. Much of my work now is looking at information management — One of the journals I publish in is called The Dark Side of Marketing, which is about marketers and business and their attempts to bend data to fit a story about which they had a narrative they had previously given, so that they wouldn’t be blamed for a failure. For instance, we can see that in the papers with some of the governments, and the banks in the credit crisis a few years ago used this smokescreen. And there's Fred Goodwin, the head of Royal Bank of Scotland who said no, RBS were not exposed to the American mortgage market. This was at the last shareholders meeting before the bank went bust. … Their most recent acquisition had in fact been an American mortgage bank, which was already deeply exposed.
So for sure, AI is also a good way of overcoming the problems of bias — What we have had in the last few years is all the macro economic stuff, which shows us the incredible power of bias to influence the way we interpret a situation and say what works and what doesn't work. So if you're a marketer and you're not truly databased, you're letting the data speak, or you're telling people you are, but actually what you're doing is making the decisions up. That is why I pin a lot of hope on AI. It challenges people's prejudices about what's working and what isn’t, and why it works — This is true of science, medicine, and many other areas.
Peter: Let's move away from fake news, or at least ‘selective information presentation,’ and look at a positive future with AI.
Dr. Stone: We just had this terrible case of the hospital in Portsmouth. An estimated 250 were killed by over prescription of certain drugs. Now the data was at the National Health and Social Care Information Centre in Leeds, which should have been ringing alarm bells. If you're sitting at the top of the tree, you need some degree of data mining AI to sound the alarm bells automatically. It’s different for the people on the ground. For instance if you work at Dillards, you don’t need to be told what's going on — if you're a shop assistant you know because you can see customers walking away, but if you're sitting on the top of the tree, you often need help particularly if you're not listening to the people at the bottom of the tree. And if you look at the National Health Service in the UK its outcomes are really poor and the people at the top of the tree are looking hard at the outcomes. You can see the marketing analogy very clearly. Then what happens is ‘bad decisions.’
If I were to follow that through, which is politically very, very controversial, why do I say that the National Health Service (NHS) has such poor results compared to the rest of Europe? It’s because the more money spent on it, the lower the quality gets because the spending mechanism is not about improving outcomes, sadly; and all the statistics now show that we're roughly, in terms of outcomes, rather like an Eastern European country — Simply put: no one is challenging the data.
Everybody has good stories about the NHS, but many people have bad stories about failed diagnoses and failing to have efficient treatment. It's not about public or private medicine, it's about data. So in Europe most health systems are public, but they’re split between the provider and the insurer, and the insurer can challenge the provider looking at the data about achievements and saying why aren't you doing blood tests early on which is what we don't do here in the UK? And we're spending lots of money rectifying stuff, which you wouldn't have had to rectify if you’d done a blood test to start.
A Better Future?
Peter: Are there any countries that you know of that are performing well by analyzing the data?
Dr. Stone: Yes, because the health insurer does it; that’s how they develop the view about whether they should pay the provider — by looking at the data — and that's what you see in Germany, Italy, France and all the Nordic countries and so on. A simplistic view perhaps, but the structure is part of it — you need the people who are delivering of course, but you also need somebody to look at the data and challenge the quality of the delivery.
Peter: But is that really AI?
Dr. Stone: Yes, they use AI to do it.
In the UK, are we just analyzing tons of raw data without applying intelligence besides human intelligence looking at results?
When it’s a very complex set of data you need AI. You need data mining, which is an early part of AI before you have an idea. You just tell me what’s happening in the system. You can argue that no hypothesis based work would have picked up the killer doctor, Harold Shipman for instance. Why would you have identified this guy? You might have done some simple outline analysis, but basically you want to look at the whole system and tell me what’s going wrong.
Peter: Shouldn’t mortality rates per doctor have shown disparities?
Dr. Stone: But you've got a million things to look at — why would you look at mortality rates per doctor? Just tell me what's going wrong. That's in a way what AI does — it goes to the meta-level. It says: actually don't worry about getting it right. That's our job in the end; I’m the artificial brain; I'm going to help you by always identifying the things that are going wrong. And not just bad things that are going on, they may be good things. In this field, the Nuffield organisation, which is a research body, clearly says that efficient hospitals are also better hospitals. The ones who are under tough financial pressure and manage with it are also the better hospitals because they're better managed. But that's quite a big statement and has been backed by loads of researchers. My preference would be to feed the data into a system, which tells me what's working and what's not — rather than hiring lots of highly specialized researchers!
Peter: It can be depressing looking at how things are at present, but you sound very positive about the future of AI.
Dr. Stone: My dream is that one day all the stuff that’s done by all these highly skilled analysts and stats people won't be needed anymore because we're just going to say: ‘Alexa, tell me what's working in my local health service?’ Yes that clearly will require AI. Alexa will have access to the data. The computer power is there now.