Computational Conversations EP#2 | Why are people afraid of the big bad AI?

There's a lot of fear around AI. Is this warranted or not? And should we all be panicking? And in what order? So, welcome to our podcast, part of our series on computational conversations, and I think we're entitling this one, who's afraid of the Big Bad AI? Related to who's afraid of the Big Bad Wolf in folklore, so to speak, or fairly recent folklore with the added amusement, of course, that we're called Wolfram. So hopefully not who's afraid of the Big Bad Wolfram AI. But what I think we really want to discuss is there's a lot of fear around AI. Is this warranted or not? And should we all be panicking and in what order? I don't know whether you want to kick that off, John.

Yeah, sure. Well, let's start with the thing that's behind it all and put that out of the way, which is the emotional side. I think we have to accept that if you look at our whole history, and maybe your folklore example is already an example of this, is that we've always been afraid of things that are human, like, but not quite. Ghosts, zombies, psychopaths, you know, that being almost human but not quite is a trope that just goes right back to our roots. So I think AI totally plugs into that. And our fear of killer robots that have been part of cinema for a couple of decades at least, is all past that. But I think behind on top of that, a lot of people have semi rational or rational reasons also to be afraid, which might be informed by the emotional side. But, you know, I think a lot of people at least believe that there's a rationality to think that there is an existential danger, because AI could be intelligent and do bad things to us. It could decide we're no longer needed, jobs will be destroyed, and all of these things. And there's some proof to some of those things, but they're blown out.

So your point? I'm just trying to think through your point about the closer it is to being a human. I mean, because of course there are things that are close to being humans that seem very positive for us, but it's close and destructive in some way. It's close and maybe it's also unknown in its style. So all the things you mention, problem with killer robot, they're unpredictable, like the worst.

The lack of empathy, they're not human enough that we can plead with them or reason with them. They cold and doing their own thing, I think is.

So I'm trying to understand also to what extent has this been true of previous industrial revolutions? As in when suddenly we had big machinery that came along that did the work of many humans. Was that the same feeling, do we believe? Or is it. I think that ties into it with this one.

Yeah, it's part of it because it ties into the rational end of it that I don't think anyone felt the machines were somehow human like and scary, but they could see the practical loss of jobs, that when the mills came along and weavers weren't needed, that their purpose in life was finished. But it's tough to say it's somebody who's losing their job. But if we look at it from a population point of view, destroying jobs is a good thing. We've been doing it for thousands of years. That every advance has been either enabling us to do something new or freeing us time from having to do something before. You know, the first time somebody put a fence around a cow and planted some seeds. The hunter gatherer said, but what do I do now? So history has shown us so far that we've always found better things to do with the time. Nobody says, I want to go back to the days where you had to put your washing through a mangle by hand and scrub it on a board. We love the fact that our jobs of doing washing from for an hour a day has been swept aside. Of course the danger here, the real danger is that there might be a lot of jobs destroyed simultaneously. And then just like when, you know, in the UK when we wiped out the mining industry, you have a whole generation who are skilled in the same way competing for.

Yeah, it's clearly very bad for some people that there's a change, but I don't think that's specific to AI. I think the scary thing with AI, it seems to me legitimately to some extent is the totally unknown nature of what it might do and the scope for that. And to me, when you have anything new, there's a kind of iteration between kind of good things and then bad criminals get hold of things, whatever. There's a new thing and it takes time to work out how to defeat the criminals on the new thing. Because there's a sort of power race between bad use and good use. And I suppose the question is in the end net, is it good? And there's another question which is, I mean, I always look at these. There's a transitionary phase, if you're a physics person, what happens? There's a transitionary phase and there's a steady state at some extent. I mean, there's never quite steady state. It's like you can blow the electric circuit by turning it on. Even if the steady state calculation is fine. If you have too big a back EMF or something, you can blow a circuit in various ways. So there's some times when you can have a transitionary effect that blows what might have been a perfectly good steady state. And with AI, I think that's absolutely possible that it does so much harm in the transition before we get some sort of more steady state as to how we're handling it that the people end up in. I mean, both it gets disparaged and or it takes over in a way it shouldn't.

Yeah, but you've got to separate, I think the harm that is done as a tool by people who want to do harm anyway. And every technology is used by criminals. We don't regret paper because people have been sent out.

Right. But the broader the technology, the broader the technology, the more scope there is for badness and the more unknowable the badness is.

Well, we're not ready for the ways that it's going to be used in a bad way. But I don't know that this. It has any more scope than the written word was a technology that has been used by bad people. We wouldn't undo that. But we know how to deal with human written words at this point.

That's right.

And then there's the intrinsic. What if the AI itself wants to do bad? And there's a power thing here. People are afraid that higher intelligence is more powerful. But if you think about the dangers in the world today, when you hear you're walking late at night and you've got footsteps on the dark alley behind you, you don't think, oh no, I hope that guy's not smarter than me. It's not the intelligence that is the danger here. It's the intent. And the problem with humans is that we're all born with basically animal intent. I want to have possessions. I want to have power over you. I want to be in charge of everything. And it's those intrinsic deep feelings that we have to be educated out of through being civilized that drive all the bad people to do bad things. The AI has got, at least the way we design it at the moment has no intent. It doesn't care if you turn it off. It doesn't need to live, it doesn't need possessions. It's just doing what it's guided to do, which might be guided by bad humans who have those desires and intents. But the AI itself is not intended.

It's hard to imbue either characteristics or no characteristics to it somehow. I mean, it sort of looks like it that way because they're fairly basic still. But it's quite hard to know whether those characteristics apply to animals in the animal kingdom. Non human animals, if you put it that way. Well, clearly there are these characteristics that we see in humans. I mean, there may be different levels of them. I mean, how would you.

It's in our firmware through million years of evolution. So when we're born, we have a certain amount of coding already in us. That is the bottom of the software layers.

Yeah, that's right. So when you think of. I mean. So I suppose I look at it several ways. There's a. There's a steady state. It'll never be steady state, but quasi steady state, net gain or loss. And then there's the transition and there's potentially. I mean, you could talk about a net gain or loss, but you can also think about individuals or individual groups where there's a huge gain or a huge loss. So it's all very well to average it. Yeah, if you look at the. I mean, I think most, many, there are some real dooms, you know, doomsayers, if that's the right word. But I think that most people would say, I suppose that they sort of have a. They think that AI might be quite good in the steady state in the long term, on average, but they're pretty nervous about where it goes. Some people wouldn't say it. I don't know. Do you think they would?

Well, a big, you know, the, the transition things are kind of obvious. There's a potential for a big wave of unemployment for people whose whole class of jobs are being wiped out. And that's a realistic fear that we have to think about how as a society we retrain and repurpose those people. When it's done slowly, it doesn't matter. They retire and they don't get replaced and nobody suffers. And there's the criminality that there's going to be a rush of great ideas of how to take our money off us using AI and we're going to have to learn pretty fast how to.

Fight us more seriously even than that.

Well, I mean, yeah, the fighting is kind of an interesting one, I think, because we've had autonomous weapons for 150, maybe it's 200 years now. And in many scenarios they're banned, but they're not banned because they're autonomous. They're landmines, they're banned because they're not good enough at being autonomous, that they're indiscriminate. And landmines would probably be less likely to be banned if they didn't blow up local children who are running, running around in the fields in between battles, but only killed combatant soldiers. So while AI weapons are perhaps one of the scariest things, again, it's not the intelligence, it's the application. If this made war pointless because no one could.

Right. But it's also who ends up with the power. It's the fact that there's a certain power structure in the world which is already in flux and that potentially I could go far more in flux. And I think there's also the feeling that correctly or incorrectly, depending on your views, in different parts of the world, but that sort of rich, stable, large countries and organizations on the whole at least don't do such completely random things as a small player with huge power. So in a sense there's a feeling that we may have very small players who end up wielding undue power because they suddenly advanced in a way that, or used the advance in a way that was unseen by large powers. Again, in the end, I think appropriate application of overwhelming capability in a sense should solve this to a large extent. But how do you actually do that?

And there's different slices where there's. Where it's quite democratized and other slices where it isn't. So if you're talking about kind of autonomous weapons, then if you're a government distrusting survivalist type who thinks their government's out to get them, ultimately that's pretty scary because you don't have that power to build your own autonomous weapons to keep them off your land or whatever your viewpoint is, whereas things like optimization of business decisions, you know, you can, at that kind of scale, everyone has access to AI now it's quite. Well, people like us, of course, are helping to drive that forward by making it easier for people who are non technical to be able to take their data and make their own AI to fight their version of the business battle of keeping their share of the riches of the world against the big players and the governments. So I think.

So what do we do to optimize this? I mean, I think one thing that, as you know, I've been pushing a lot is I think this computational literacy for everyone is critical. Now, it may seem a bit far out, but I think the problem is that if you have some understanding for how to think in an ordered way with the use of sort of the computational Process that essentially is being used not only by AIs but many humans making the most important decisions right now. That helps society get to grips with this better, helps to put the human on top. So that seems one important.

You are focused on that democratization end of things, which is how can we give everyone the power to judge when they're being manipulated or to do their own computational work and have equal power.

And I think the other thing is how do we supply the best platform for people to use AI? Or I would say, I mean, I think when we talk about this, right, we often talk about computation being the bigger picture actually. I mean AI in a sense is an application of computation. And by the way, people often muddle up specific techniques of AI like machine learning with the aim of AI, as in artificial intelligence. Of course. We've been trying to, for 35 years or so build the ultimate sort of platform, encompassing all of that, all the possible algorithms and ways to do it, including things that you probably wouldn't classify as AI. But so how do we, you know, by giving a sort of maximum scope a bit like, you know, how do we. That potentially allows better decisions to get taken, but seems quite a long way away from these, these concerns.

Yeah, and I think you've got to separate again, there's different slices of this. There is pure consumers of AI and those people, you know, the platform's too broad a term for it. It's going to be embedded in everything. Your camera will take better pictures, already does because it has AI, much better. Your washing machine will do better washing. And you're not thinking in terms of using AI. It just is just part of the fabric of society. And then there's the open ended problems and the creative and the creating side of that. And that's where we need to come in and figure out what those platforms are. Now I think our end of things is to be more focused on the slightly higher end of that consumerization. The people who understand data. If they've been educated in the way you'd like to see them educated, there'll be more of those people who can do come in with a totally open ended question. How do I make something better using AI and be able to wield that tool as a much more general tool and be the makers. And that I think is. Well for where we are. Of course it is. The most exciting challenge is what we think about a lot is how do we make the makers. And then there's going to be the super high end, which is what the researchers and the universities should be focused on, which is inventing new tools that we put in the hands of.

Of people to wrap up. I mean, I was curious about the brand AI. Is that part of the problem here? Right. It seems general, confused. It's exciting as it is scary.

Yes. And it's meaningless. And we have to admit that we're all part of the hype that talks about this, that we've been doing AI since before any of this generative stuff and chatter existed in. You know, when I started with you, back in 30 years ago, the ability to solve mathematical equations and do integrals in calculus on a computer was called AI because, you know, people would gasp when we showed them and that the computer could do something.

Doing an integral computer seemed like AI at the time.

Yeah, yeah. Because they thought that was the preserve of human intellect. And we showed them it was just another thing, automatable thing. And that what is AI has changed every year to be whatever is at the current leading edge.

But you like the term. Well, I mean, you don't like the term, but you have to admit it's been rather successful at pushing. I mean, because it makes a big difference. Right. I mean, in a sense, huge amounts of money are now attached to AI. Now, that's partly to do with the reality of what's been done, and it's partly to do with the fact that it's a catchy, you know, artificial intelligence, sounds very intellectual and clever. And then it's great because you can shorten it to AI, which just is very easy and seems.

Yeah. And it's put it on a lot of people's agenda. So in that sense, it's been a great term to have a flag to rally around and say, the world has changed. I need to get in on this. And it's driven a lot of conversations I have in our consulting team with clients who come and say we need to use AI. What does it mean? And how do I do it? And we then kind of turn that into real work. But when I'm actually talking about how we solve problems, I don't use the term because there's so many different kinds of AI I want to focus on for this kind of problem. We're going to do this. And this is 20 different things that sit under the AI umbrella, of which LLMs happens to be the latest thing. But it's not the only game in town. And it's quite bad for some things, which is another unspoken thing, that our AI is not as I as people think. It's giving the appearance of intelligence. But actually there's the current round of AI. Yeah. The current round is being given much more credit than it deserves because it sounds human. Yeah.

And interact in the right way.

We think it understands the last few.

Percent will take a huge amount more time to make it really feel more human, so to speak. But it is amazing nevertheless, compared to what we saw a few years ago.

It's great for what it does, but it doesn't do what people think it does to the level.

On that note, we should probably let people go on this round, and thanks for joining us and hope to see you again for the next computational conversation.