79. Sounds like marketing sci-fi, is the future

PLEASE NOTE: This transcript has been created using fireflies.ai – a transcription service. It has not been edited by a human and therefore may contain mistakes.  

 
00:03 
Dom Hawes 
You're listening to unicorny, and I'm your host, Dom Hawes. Welcome back to part two of the most sobering podcast of the year. Today we are talking to Steven Milman about groundbreaking technology that will transform AI's capabilities, enabling unprecedented speed, efficiency and real time data processing. Today, we discuss the profound implications for marketers, from hyper personalisation to real time customers insights and the ethical considerations that come with these advances. It's all getting a bit Sci-Fi at unicorny, but stick around. We're about to bring up general and super intelligence for AI. It is scary, but it's not all doom and digital gloom. So in the last part of this episode, we took a look at AI and how Quantum may affect it very briefly. And then we spent some time talking about marketing, market research, and use of tools.  

 
01:01 
Dom Hawes 
Specifically, let's start with our artificial general intelligence. What is it?  

 
01:06 
Steven Millman 
Sure. So, artificial general intelligence, which is today hypothetical, is the point at which an artificial intelligence becomes as capable, or more than a human at the broad range of tasks that humans do. So, for example, there's an extraordinary chess engine. I forget the name. My son is a chess player, Sam, and he's fantastic. There's this chess AI, which is essentially unbeatable at this point. But that's not general intelligence, because that's all it does. It can't write a story, it can't answer questions about other things. But when we get to artificial general intelligence, we are going to have an AI that is basically able to do everything that we can do cognitively, but will be vastly superior to us in a lot of ways.  

 
01:51 
Steven Millman 
Perfect recall, vastly larger amount of data available to it, the ability to multitask far more effectively than us, and literally anyone's more effective at multitasking than me these days. And the reason why this is an important stage is it is widely believed that once we hit this stage, the speed at which AI is going to grow is going to be almost unfathomable, because what we can do now is we can tell the artificial general intelligence, build a better one, and it will be able to start iterating on itself.  

 
02:21 
Dom Hawes 
And what about artificial superintelligence?  

 
02:23 
Steven Millman 
So, artificial superintelligence, this is the one where listeners might want clean underwear handy. So, artificial superintelligence is when artificial general intelligence becomes so much more complex and so much smarter than a human, I mean, on the order of potentially thousands or millions of times, that it is no longer comprehensible to a human mind. Once it achieves these levels, an artificial superintelligence will think about problems in ways that we cannot comprehend. And it raises questions about what will it do when given an instruction, how will it interpret the instruction? And there's a. It's a professor, I want to say, at MIT, but I can't recall who, likes to use the example that you could say, I want to solve a really hard math problem, artificial superintelligence. Solve the math problem.  

 
03:12 
Steven Millman 
And the artificial superintelligence decides that the way to solve this problem is to build a massive computer the size of a solar system and kills all of humanity in order to use all of our resources to create the supercomputer. To answer that question, it just takes us beyond where we can imagine. And when you get to that stage, the principal concern is, how do you control an AI that can do that? If you think about all of the species we know, has there ever been an example of a much smarter species treating a less smart species with kindness? It's a big concern, I'm guessing, like.  

 
03:50 
Dom Hawes 
These are quite a way off. We'll talk about when. But Arnold Schwarzenegger is not going to be here to save us from the future.  

 
03:56 
Steven Millman 
He might not be, although I tell you, I was at a conference last week, and someone was talking about what I felt was a fairly naive approach. How do you give these things morality? And he was talking about crowdsourcing morality and handing that off.  

 
04:09 
Dom Hawes 
I was just thinking about that.  

 
04:11 
Steven Millman 
Yeah. I think what I said was, have you seen Twitter? Have you seen what happens when you put AI on Twitter? But in any event, the idea would be then you would hand that to philosophers and ethicists, and then they would use that to construct a form of virtue that you could teach a computer, and that you would give it our sense of morality, and then it would simply be self guided. At that point, I'm going to bring.  

 
04:32 
Dom Hawes 
Us back into the realms of the present day, because I find artificial superintelligence genuinely frightening. But presumably we can actually tell it to build a time machine and then bring Arnold back to sort of problems.  

 
04:44 
Steven Millman 
That's what I was going to say about this conference. I kept expecting a ball, flashing ball of light, and someone to come from the future and shoot the guy. It's like, okay, we're safe now. We're safe now. Don't do that. Just to be clear. Don't do that.  

 
04:57 
Dom Hawes 
And it wouldn't be coming back in a DeLorean, either.  

 
04:59 
Steven Millman 
No, probably not.  

 
05:00 
Dom Hawes 
Artificial general attention, though. That is nothing unfathomable or inconceivable. In fact, quite the opposite. When do you think that we may be able to see something like artificial general intelligence starting to appear? Is it in our lifetimes?  

 
05:16 
Steven Millman 
So I'm 55. I would like to think I've got another solid 25 years, maybe more. Who knows? We'll see what happens. I do drink a lot. I am absolutely convinced this will happen in my lifetime. And, in fact, I think possibly within the next ten years, we'll see artificial general intelligence. I have heard people say that it could be in the next three to five. I think it's a little Pollyanna, but at the rate things are moving, I do think it's entirely likely that we're going to see this sometime in the next ten years.  

 
05:45 
Dom Hawes 
So this generation of workers effectively will have to cope with that transition.  

 
05:51 
Steven Millman 
Yeah.  

 
05:51 
Dom Hawes 
And it's going to be profound, and.  

 
05:54 
Steven Millman 
It'S going to be, I think, a more profound change to the economy and to the way we live our lives, potentially, than the internal gas engine.  

 
06:04 
Dom Hawes 
So when Sam Altman said, artificial intelligence is going to be able to do 90% of what marketing currently does today, that either means he's got a bad experience of marketers and thinks they don't do very much, or it means when artificial general intelligence comes, we're in trouble.  

 
06:19 
Steven Millman 
Or he's got an artificial intelligence company and he's trying to push valuation. Yeah, I mean, all those are possible. Let's also not forget that the team that he put together to save us from the ills of AI, he had this team of people who were guaranteed 20% of the compute time to focus on how to prevent AI from doing terrible things. He just disbanded entirely. The division no longer exists. A bunch of people quit, were very vocal that they were quitting, because they didn't think that they were really working to protect people. All of these large language models, all these AI companies are in absolute stone cold panic that they will not be as fast as the other ones we're not seeing, I think, the kind of safety measures we might want, and today it's not that problematic.  

 
07:07 
Steven Millman 
But if we're still doing this when general intelligence, artificial general intelligence, appears, we might be cooked. I think it was a little over a year ago. We talked, and I said something of the sort that worrying about artificial intelligence taking over the world or killing humanity is a lot like worrying about overpopulation on Mars, and we'll see how I feel in three to five years. Well, you know, a year later, I don't think we're talking about overpopulation on Mars anymore. I'm a little terrified that it's moving this fast.  

 
07:34 
Dom Hawes 
So given that he thinks, from a marketing point of view, 90% of what's currently done today is going to be able to be done by AI. That's kind of. That's not possible clearly today or with today's tools, but artificial general intelligence could lead us in that direction.  

 
07:46 
Steven Millman 
Yeah, and no doubt there'll be a massive decrease, because if you think about what any company does the most of, well, let's not talk about sort of fabrication or companies like that, but professional organisations that focus on the movement of intelligence, which is pretty much what I do and what you do. The vast majority of the time we spend is not spent on creative processes, this conversation is for sure. But how much time do we spend editing? How much time do we spend managing our finances, ordering things we need to make things work? Fixing the mixing board? When you talk about eliminating 90%, I think probably Sam's thinking mostly about those sorts of things and then speeding up the things that we do to assist the creative process. 90% seems high to me, but, I mean, it's not out.  

 
08:34 
Steven Millman 
It's really not out of the ballpark. When we get to a point where we're closer to artificial general intelligence, when we have quantum computing, there's a lot of talk today about providing basic income that so much will be taken out of our hands that we do need to think about if we're generating this much wealth with this little effort of making sure people aren't left out.  

 
08:59 
Dom Hawes 
So how are you finding today's episode? It's fascinating, isn't it? Steven Millman's given us incredible insight into how quantum computing will revolutionise AI, bringing speed, efficiency and real time data processing to the forefront of marketing. Imagine a world where we can model real world scenarios in real time, enabling us to respond instantly to customer actions and needs. We explored the technical marvels and the potential of quantum computing. Now, let's connect this to the practical implications for us as marketers. Real time processing capability that you get with quantum computing will enable hyper personalised marketing campaigns tailored to individual customer preferences and behaviours. And by the way, it does it on the fly. This isn't just a leap forward, it's like a quantum leap forward. It's going to redefine customer engagement completely. Combine this with smart devices and other innovations that we all already use.  

 
09:58 
Dom Hawes 
The Internet of things, NFC, chips, GPs. We could be modelling every single customer in real time, everywhere, all at once. Man, that's making my brain hurt. Coming up, we're going to look at specific applications of these technologies in marketing. We're going to discuss how to implement these advances responsibly, ensuring that we all build lasting and more meaningful connections with our customers. So let's get back to the studio. My mind is cast back to the last time we met, and you and I talked about maybe the disappearance of the current interface that we all use. And I've since started to hear on other podcasts, which I may say even all in, we're discussing this after were airing it, but a lot of our time and attention is given to creating work that is consumed through an interface.  

 
10:48 
Dom Hawes 
And if that presentation layer is taken away, if we do finally get to live the vision of Siri or Alexa, where there's the same as we can now do an awful lot with a text chat prompt, we could be doing an awful lot with voice. That might destroy quite a lot of work.  

 
11:04 
Steven Millman 
Yeah, and we're already starting to see that people are talking about the death of the web becomes concerning, because then where do you create new. What's the incentive for anyone to create new content? But that's definitely, there's no replacement for it yet.  

 
11:17 
Dom Hawes 
There's got to be something. So let's come back to our artificial general intelligence, because that feels more tangible, less frightening than superintelligence. Still a bit concerning. What do you think the key steps and milestones on the path to its development and then release are?  

 
11:32 
Steven Millman 
The first thing is going to be developing computational power that will allow it to exist, and that's quantum computing, more likely than not the next step. Obviously, people need to develop this. It is not going to be the case that large language models are going to be general intelligence. That's not what they are. A large language model is what's called a transformer neural network, and all it's doing is really predicting the next probably right word in a response. So the work that needs to be done to create artificial general intelligence is still very hypothetical. So very smart people are going to need to figure out how you construct a model that will do this kind of thinking. There's been a lot of progress, but that has to be accomplished.  

 
12:16 
Steven Millman 
And then there's going to be the question around how do you make them safe and how do you regulate them? And so they're not like super intelligence, where we don't understand anything that's going on. They're going to be too far ahead of us, but we are still going to need to be able to put fundamental boundaries on them so they don't do bad things. And some of those are going to be easy to do, right? Don't kill anyone. That's a good one. But we are going to need to come to some general agreement and understanding as to what those rules are. And it's not easy. There's a whole literature now on how do you teach an AI human values or morality. Do you say, here are what morals are, and now you figure it out.  

 
12:55 
Steven Millman 
Do you give it a series of instructions that it might misinterpret? It's really not clear what to do. One thing I've been thinking about that's funny is as you look at this literature, if you read through it's kind of fun if you have the opportunity. The kinds of Armageddon that people talk about in the movies, the supercomputers, and oh, we've decided that humans are bad for the plants, we get rid of all the humans. Seems really far fetched. What doesn't seem far fetched? I think the most accurate movie I've seen that describes what AI, general AI would likely do if it was sort of left to its own devices is the movie WAlL E. And it becomes a caretaker and it prevents people from harming each other. So it kind of bubble wraps humanity and takes away agency.  

 
13:36 
Steven Millman 
And I think that's really the thing we need to be cautious of as these things roll out. I thought was interesting that what a lot of people are talking about today are kill switches, is to make sure that there is a incontrovertible way to stop them if they start doing something you don't like today, that's not really a thing because you can't just unplug it. These things are cloud computers.  

 
13:57 
Dom Hawes 
I think the other thing, from a marketer's point of view, then don't take away agency, for God's sake. I won't be able to pay the mortgage. Bad pun there, but from marketers point of view. So we're all wrestling with things like GDPR at the moment still, because it's just dreadful legislation. But privacy is top of mind. Consumer customer privacy particularly. Also, of course, you know, we have quite a lot of each other's data that we're all trying to protect. How do we make sure when we're building systems that have general artificial intelligence, that ethics is one thing, but kind of the rules, I suppose you can teach them the rules of legislation, but given that humans disagree on interpretation of law, will AI help for marketers? Will it make it easier for us to interpret and act within laws?  

 
14:42 
Dom Hawes 
Or will it be just as this.  

 
14:44 
Steven Millman 
Is a terrible thing to say, of course, but I think it's actually going to make it easier for bad actors to defy the law. The biggest challenge with AI from a regulatory standpoint is that they are not transparent. It is really hard to figure out what in the world they're doing. And I think Anthropoc actually did a really good job recently of figuring out how to expose that. And so hopefully anthropic is, are the folks behind large language model called Claude. So I think it's going to enable bad actors more than anything at the beginning. So we'll have the legislation. The legislation takes several years to get through and several years might as well be 50. The way that this stuff is moving.  

 
15:27 
Dom Hawes 
It brings to mind Google and cookies. Yeah, our very first show we talked about.  

 
15:32 
Steven Millman 
That's right.  

 
15:34 
Dom Hawes 
And we're still where were then.  

 
15:36 
Steven Millman 
But hey, if you listeners want to go back and listen to that, you'll know that I predicted that we wouldn't deprecate them by now. And we haven't.  

 
15:44 
Dom Hawes 
You did. What about your business is about understanding human condition, human thought, the likelihood of humans to behave in a certain way, research, how can advanced or much, much more advanced, underpinned by quantum computing, how can that help marketers enhance the work they do?  

 
16:04 
Steven Millman 
Yeah, I've been talking a lot lately in conversations on this about columns, not rows. And what I mean by that is the AI is really spectacularly good at taking rich, complex Personas and making good guesses at what those Personas would say if they were asked a question that has never been asked to them. What they're very bad at today is creating entirely synthetic people. And I think that's going to be very hard for a very long time. I don't know when that becomes a reality. So most of what you see in the marketplace right now falls into the let's create fake people are either being used legitimately for the purposes of simulation, which is very effective, or illegitimately by saying, hey, here's a bunch more people, and now run your stat tests because we've got all these extra people.  

 
16:55 
Steven Millman 
I won't go into why that's a big problem right now unless you ask.  

 
16:59 
Dom Hawes 
We can go back to the previous episode. We did talk about it there.  

 
17:01 
Steven Millman 
We did talk about it there. That's right. But I think where it's really going to be valuable for a company like mine and how we support our clients is it's going to allow us to say, okay, instead of asking an hour long survey, let's ask a 30 minutes survey. And then let's be able to extrapolate the other information that we need on the back of that.  

 
17:20 
Dom Hawes 
So then you get a high response rate to your original survey, high degree of accuracy, and therefore the extrapolation is better.  

 
17:27 
Steven Millman 
So ask the questions. You must get a real answer to ask enough questions to be able to create a Persona that will accurately. They'll have a relatively high accuracy in predicting others. And then it also allows you to go back and say, gosh, you know, I wish I'd asked this question. I wonder what the answer to that question would have been.  

 
17:44 
Dom Hawes 
And then when we get artificial general intelligence, though, you'll be able to synthesise much more accurate results without necessarily having to sacrifice either the length of a survey, because it'll be done almost instantaneously. It'll be a different approach.  

 
18:00 
Steven Millman 
No, I think you still are going to need to do your basic survey work. I think the typical block and tackle is going to need to happen. Artificial general intelligence can do all the things a person can do, but it's also not a mind reader. So it doesn't know, just intrinsically, that more people right now are thinking that they want an electric car or less. In those scenarios, you wouldn't want to ask it to give you 100 survey respondents what you'd want to do in that case, if it were that good, is to say, you know what? I would have asked a survey to find out what people think about electric cars.  

 
18:34 
Steven Millman 
But I'm just going to ask the AI, what do you think the population of people is right now who want to buy an electric car and if it's got access to enough quality data, if the data are. If there's enough recency in the data, and recency is a big problem right now, then it might be able to give you the answer without having to talk to people. But even with artificial general intelligence, we don't know if it's going to really be able to replicate or estimate human emotion, affect the things that it's really bad at today, because it's really important to, I think, emphasise artificial general intelligence is not a conscious person. It is not a self aware computer. This is not Skynet. It is a computer. It is a thing. It's not a he or a she.  

 
19:22 
Steven Millman 
It has no emotions, it has no feelings. They talk about. I've heard people talk about, is AI ethical? No. AI is ethical, nor is it attempting to do harm. It's simply a thing that we've created. And ethics lie in the hands of the people who construct them.  

 
19:42 
Dom Hawes 
Wow. What a journey from the dystopian visions of Skynet in the Terminator movies to the cutting edge realities of quantum computing and AI. We have covered a lot of ground on this podcast. Now, remember the first time you saw AI on the big screen? It was a terrifying, self aware Skynet launching thermonuclear warfare to eliminate humanity. Thankfully, the ANI that you and I know is a lot less menacing. It's a lot more helpful. So breathe easy. The machines do still work for us, even if they're still crap at copywriting. Let's start with a few unexpected insights that emerged from our discussion with Steven Millman. Steven told us that quantum computing isn't just incremental. The improvement is a dramatic leap forward.  

 
20:27 
Dom Hawes 
One of the most mind bending aspects of quantum computing is its ability to process vast amounts of data distributed over miles in real time. And this opens up loads of possibilities for marketers, from real time customer engagement to instant, and I mean instant, data driven decision making. Like, imagine the ability to personalise marketing messages to each customer in the blink of an eye. It's like having a supercharged, turbo boosted marketing engine everywhere, all at once, all at the same time. We also explored how these technology advances are going to reshape predictive analytics. With quantum computing's enhanced AI capabilities, we'll be able to anticipate customer needs with an accuracy that we can't even imagine. And that means we can create marketing campaigns that aren't only reactive, but are also proactive.  

 
21:16 
Dom Hawes 
And I'm going to give you a trite example, because I'm struggling even to envisage what it means. Bye. But, like, if you're running a service or supply business today, you will know in future when each and every one of your customers runs out of each and every one of your products, you'll know their intent, you'll know their market, you'll know their environment, and you'll be able to automate sending replenishments of the right products at the right time to the right place if they need it. It's like having like an everywhere crystal ball, but better because it's based on real and real time data. Another unexpected insight for me was the profound impact of quantum computing on creativity. Now, I'm a bit of a sceptic when it comes to AI's general ability to match humans for commercially effective creativity.  

 
22:03 
Dom Hawes 
Like today, AI can generate ideas and it can assist in the creative process. I think with quantum computing that might change because the scope and speed its capabilities are basically going to skyrocket and that means that as marketers or as creators, we can iterate and refine concepts much faster. We can do much more innovative, much more engaging things. Imagine an AI that not only suggests ideas, but it also creates really compelling content on the fly, tailored to the nuances of each individual person you're trying to reach. Let's catch a breath, though, to remember the ethical considerations, because, as Steven pointed out, with great power comes great responsibility. We're going to have to navigate the fine line between innovation and privacy, ensuring that our innovations, yours and mine, don't damage trust or transparency with our customers.  

 
22:54 
Dom Hawes 
So it's not just about harnessing the technology for better outcomes, it's about doing it in a way that respects and protects our customers. After all, it's all about trust. It's all about experience. So, as I now get ready to go and find a cold Negroni to chase my fears away, I'd like you to picture this. A world where marketing is so seamlessly integrated with AI and quantum computing that it becomes almost invisible. Marketing messages feel like natural conversation. Brands anticipate your needs before even you know you have them. And customer experiences are so personalised, they feel tailor made just for you. This isn't science fiction, it's the future. And it's coming to and for you. You have been listening to unicorny. I'm your host, Dom Hawes.  

 
23:51 
Dom Hawes 
Nichola Fairliey is the series producer, Laura Taylor McAllister is the production assistant, Pete Allen is the editor and Peter Powell is our scriptwriter.