Tribalism is a Luxury We Can No Longer Afford in an Age of AI with De Kai #2

Hello, everybody. This is David Goldsmith, and welcome to Redefining Tomorrow. Today we are going to be exploring the title of the program. Tribalism is a luxury we can no longer afford in the age of AI. And we have a good friend and a great guest on the line. We have Dakai. How are you doing, Takai, today?

I'm great. How are you?

Good, good. I'm excited for this interview. Very short intro, as always. Takai and I met in Hong Kong. We've part of similar circles. We've had a variety of discussions that have to do with the topic of AI. And the one person I'd call on if we were talking about AI is Takai. He was one of the 17 ACL founding fellows in terms of his skill set and his capabilities. He's been in the field of AI for natural language processing and translation, where he pioneered the whole space. And today, if you look around, he didn't commercialize it, but it's being used in commercialization from Badu and Google. And there was just an announcement today, and I got to move my screen to see it, that there's a. An AI Ethics Advisory Council, and Dakai is a part of that. So I am excited to hear what you've got going on and to talk about tribalism and where we should go with it. So do you have a few bullet points that we're going to cover today?

Yeah, I think it would be really cool if we would talk first about sort of resource scarcity allocation and bounds in the context of the world order that AI is emerging in.

Okay.

And a couple of other bullets that we might hit on would be artificial children. Okay.

I like that.

Maybe next artificial gossips, and after that, maybe mindful AI.

Wow.

Sort of like, you know, this will all tie into the overarching theme of why we need empathetic AI and.

Okay, is there another one? We got another one. You have a fifth one here?

Well, that's sort of it, I suppose. Yeah. Okay.

All right, so let's start with this first one. Resource scarcity allocation and bounds. Teach me something. Tell me something that I need to hear, because I'm excited to find out where this is going.

So I think these days, having done a lot of work in pushing machine learning approaches into AI systems for. For processing human language stuff that the first web translation AI such as Google Translate or Microsoft Translator, Baidu Translate, that sort of stuff. And having been doing this for decades, literally, I've also been increasingly aware and concerned of the impact that these sorts of AI and language technologies are Having on what used to be a somewhat stable world order and is increasingly looking to get massively disrupted in ways that humanity is simply not prepared for. I, um, and, and when I.

Obviously you have to explain that to me. AI is moving into places that is massively disrupting and humans are not prepared for it. Okay, we. Did we just hit on the apocalypse?

Well, you know, yeah. Yes and no. I mean, the thing is that the apocalypse that you hear about in, in, you know, sort of the popular sphere that Hollywood narrates to us and, you know, that viral news sort of tells us about is not the apocalypse I'm talking about, you know, it's not Terminator, Skynet, ex Machina, kind of like, oh, the robot overlords are going to destroy humans and take over. That's not what I'm talking about. But I think there is something potentially apocalyptic that we need to get ahead of, which is us humans, ourselves, armed with AIs. We have, as a species, we have long made the best possible use of the technologies that we invent. And the current world order is built upon characteristics of that. It's evolved over time. Obviously, especially in the last several millennia, we've done an awful lot and the order has changed. But compared to what's coming, the amount of change that we've seen in the last couple of millennia still pale in comparison. So let me try to explicate this a little bit.

Yes, I'm sitting here saying, okay, we got a lot of, I think, hyperboles. We've got a lot of hypothesis flying through what's going. I'm trying to see what you're seeing.

So the current world order that we live in has evolved under three sort of points. It stands on three legs. The first one is resource scarcity. Resource scarcity occurs across the entire spectrum of Maslow's hierarchy of needs. Do you know what I mean by that?

I think I do. And are you talking resource scarcity when it comes to humans and food and being able to move up the hierarchy? Are you also talking about resource scarcity, for example, on planet Earth?

Yeah. So you can have resource scarcity at all sorts of levels. From the physiological sort of just food, water, housing. Well, housing belongs sort of at the next level, safety, sort of security. And then next level you have love and belonging. And so each of these levels, as people sort of attain a level where they have met the needs to a reasonable extent that they start moving up this hierarchy. Right. And if you've achieved the love and belonging, then the next thing people need is Esteem, self esteem, respect of others. Right. And if they've achieved that, then they move to self actualization at the top of the pyramid, which is, you know, sort of about things being meaningful. Yeah. And so that's kind of, you know, it's a spectrum of needs that through the evolution of human history, it's all been about humans finding the resources needed for these needs scarce. Right. And humans fight on a daily basis to do better at meeting those needs.

Yep.

So that's sort of the first leg that our systems evolved under. The second leg is resource allocation. And the resource allocation has been based on competition. This is the basic structure of evolution, for example. But what is competition? Competition is based on a mindset of what you're in group. So you have some in group and it is competing against out groups. That is how competition is set up. And the reason why we've evolved resource allocation based on competition is because it's well known in psychology and in politics that fear is the strongest motivator. Now there are many, many theories and models of emotions. It's still an area that is being actively studied. There's no resolution to it, but pretty much every single model of emotions, fear is the axis that is by far the strongest. That is the one universal across all the current psychological theories of emotion. I would agree with you in Washington D.C. or any such place, everybody knows that in order to get any legislation passed, you have to invoke some kind of fear because otherwise nothing moves. And so the problem with this is that this kind of mindset is self perpetuating, this competitive gay mindset. And there is a strong tendency toward winners take all. It increases inequality and it increases in quality if left unchecked across pretty much all of Maslow's needs. And the problem here is that in order to win at this game, you have to dehumanize or de demonize the out groups. So it, this, this sort of mindset is what allows members of one tribe or one group, one nation to kill the outsiders that they think are threatening their own in group, collective.

I just watched something on a video yesterday or the day before on, on political, this type of situation happening where in politics what you want to do is you want to dehumanize or you want to create any this sense of unknowingness or fear that comes out of the situation. And if you do that, no matter who else comes up, you will never trust them. And it continues and it creates the winners take all. Demonizing the group, it's the resource allocation and they were Talking about in this situation, the Green deal in the United States. And they said no one really knows what it's about. No one's even talking about it. All they're saying is, see the Green, they're socialists, they're socialists, they're different than us.

Right.

And they keep on perpetuating that which creates this sense of fear, which is self perpetuating. And even if the data is coming out, it no longer is something that someone's going to trust.

Yeah, exactly. So you. So those tactics are dehumanizing the out groups. They are allowing members of their own in group to. Threaten the out groups and as things get extreme, to kill the out groups without any remorse. Now the problem is, you know that this is how humanity has evolved, right? Early attempts at civilization, it's been said, survived competition from other groups only because of this kind of adaptation. Okay? And so you know what we get because of the dynamics that you just pointed out is that you have a larger and larger number of people who are losing from the winners take all games. And this is increasing the inequality across Maslow's hierarchy of needs. So at the physiological level, what people are fearful of is its fear of extinction. At the safety level, people are fearful of mutilation. At the love and belonging level, people are fearful of rejection. At the esteem level, they're fearful of humiliation. At the self actualization level, they're fearful of meaninglessness. And so as you get more and more fear, people, more people losing out into a greater degree are on the fear axis. That's when people lash out. Okay. Pretty much every.

I see that happening in some cultures around the world already.

Yeah, you see it happening in cultures in situations where, you know, entire countries have been subjected to this. Then you know, we see the dysfunction and the violence that breaks out even in, in wealthy countries in the US you see individuals who have lost out on these lashing out with mass murders. So this is sort of the second leg of the three legged stool that our current system has been built on. And the third leg of the stool is resource balance on destructive capacity. So the thing that has allowed us to survive until now is that even when losers in that game lash out, the worst times that it's gotten out of control were things like World War II. Even in those times when the World War began, we had not yet invented nuclear arms. And so at the close of World War II, for several decades, it was only a very small number of big nation states that had access to these weapons of mass destruction. And so even under extreme inequity stresses the individuals and non state actors that had lost on the fear game could still only inflict limited damage when they lashed out.

Okay, see I, it's obviously the more the power or military might which we're using, nuclear is one of them. The more that these are proliferating around the globe. If someone is feeling as if they're a loser in this game, they have the ability to lash out in a much different way than they could prior. Is that where we're going?

Yeah, yeah. I mean you, you know, a few decades ago there's only so much damage you can do with a knife or with bows and arrows or with a rifle. You know, you can inflict some local damage. Yeah. Even with a semi automatic, the damage is still localized. The problem today is that this third leg of the stool is evaporating really quickly. And that is being driven by the advances in AI and robotics.

When you say evaporating, it doesn't mean it's no longer valid or is it's expanding because there's more capabilities.

Yeah, the resource bounds are evaporating.

The resource bounds. Okay, that's what I was trying to get at because you use the word bounds. So there's no longer a boundary that used to constrain individuals from or groups or nation states from doing anything. And today that's becoming nominal.

Yeah, Exactly. I mean WMDs, weapons of mass destruction used to require the resources at the nation state level. And that helped to keep the balance in the world order that had evolved, albeit with high inequity and massive wars, but humanity still survived. Now fast forward today back to today. We've got a situation where we've come to the end of the entire evolution of humanity up until now, where our survival has been depending on and enabled by our ability to outrun the destructive technologies that we invent. I don't know if you've seen the short film that was produced by my friend Stuart Russell, who's a professor at Berkeley, together with the World Economic Forum. It's called Slaughterbots.

No, I have not seen it.

So it's worth watching because what it shows is a very realistic scene of new arms that that are have been invented and are promoted. Which are little small micro drones, but fleets of them, like thousands of them that can be produced inexpensive swarm technology. And each one carrying in that film a single shot shaped charge. Yeah. And so these fleets can be produced almost by anybody in a very short number of years. Any disgruntled individual or small groups of individuals, non state actors could Go into their basement and 3D print massive amounts of small drones armed with things like that or with bioweapons or chemical weapons. And so suddenly you have a situation where no longer are weapons of mass destruction confined to the control of quote, responsible nation states that keep the balance. This is more like imagining a world where half the population of the earth is walking around with weapons of mass destruction. And the question that we need to ask ourselves is, would you want to take your chances that absolutely nobody would hit their launch button?

It's guaranteed they will. Yeah, to me, it's not even a question, it's a guarantee. It's like the Chinese individual just did the cloning of a baby. And I wrote and paid to think that if the technology is available, someone will use it. There's just no doubt.

So. Exactly. So the key question then is that we need to all be asking ourselves is what would cause a person, person to hit that launch button?

Let me add just this thought I was thinking as you were saying it, we've even compounded this resource bounds constructive or destructive capabilities because we add an element of viralness, whether it be any form of Internet media, any type of distribution. So you can have a destruction happen, then you can show it and exponentially create even larger fear which breaks down the legs.

Correct. And this is, you know, this is the challenging nature of the problem that we're facing. I didn't mean to sound all hyperbolic as we were starting, but this is a real challenge the likes of which humanity has not faced before. Again, because resource bounds on destructive capacity used to keep things localized and in check. The challenge is that we cannot just minimize the difficulties of tackling problems in human nature, as you've just said. Yeah, I mean, after all, fear is biologically evolved over millions of years. And yet at the same time, we, we can't just, you know, stick our heads in the sand. We have to actually be cognizant of the constraints I've just mentioned and the timeline that we're on. Because AI evolution is racing along at an exponential rate. And meanwhile human cultural evolution is still plodding along at the same old linear rate that it always has. And you see that in our current regressive trends toward the polarized politics, the parochial mindsets, and ultra nationalism, which are exactly what anticipated the world wars previously that will doom us in this new era where the resource bounds on destructive capability have evaporated. And so again, without wanting to sound hyperbolic, we simply can't afford under these constraints to be, to let ourselves be over constrained by incrementalism. Simply put, humanity needs cultural hyper evolution at a pace never before witnessed, merely to keep up with the hyper evolution of the destructive potential of the technology that is proceeding at that pace.

I don't know if I said this in the last interview that I just did, so I'm going to hope that I didn't say it in this one. I've got two series. I read this. I understand exactly what you're saying. The challenges that as humans we're just incapable of communicating or working at the speeds in which any computer, not even AI, is capable of working at. In this one series, you may have heard of it called the Singularity. But it's a series about evolution and computers growing and at one point the computer system. I'm not going to go into the whole book is a series of four books. I believe it was, it was great. But the computers had shut down everything having to do with humans to be able to survive because they didn't need the food, water, shelter, transportation, communication, the same skill set needs. And they were negotiating with humans at this one point. And the humans had one AI that had to be the negotiator because the humans could not keep up with the speed of negotiation. And at one point the humans had given this suggestion. What we'll do is over the next two years we'll build enough computers so that you can have your home and we can have our home and we can live amiably. And the other computer system said in those two years it would be the equivalent of stopping us from 20,000 years of evolution. So we can't wait two years. And the analogy or the context was brought in that a dialogue between two humans, the two of us in this time frame, it's not even terabytes. It's so much information would have been transferred in a matter of the time that we've been talking. So how do you say we have to have a cultural revolution where I'm still seeing the computer evolution expanding so much faster that humans won't be able to even meet this challenge at the.

Lowest level, yeah, not so much a cultural revolution, it's just hyper revolution of our culture. And you know, I think that, that those are, those are very much challenges that people are who are looking forward a significant amount are thinking about. I was, I was one of a hundred folks that sort of got cloistered a couple of months ago in Puerto Rico at the Beneficial AI series that Max Tegmark runs. And one of the debates that he asked us to do was to envision what we would like for the future of humans. And there were four positions taken by four of us. One of them was that humans should remain the way they are. Next one was that humans should just go extinct. The third is that humans should upload. So, you know, you could, like, imagine the AI simulations of you going into the cloud.

And the fourth was the singularity type conversion and being able to live within that in a different world on different means.

Okay, well, the consciousness of a human is the idea there. Gets uploaded. And then the fourth was that humans merge with AI. And that last one was the position I was defending. Here's the problem, right? If humans are kept around, it will be more or less as zoo animals as you've just described. Because if humans are kept separate from the AIs and the AIs go their own way, they are going to be evolving so much faster than, you know, un. Augmented humans that the only reason they would keep us around is kind of as pets. Yeah, now that doesn't sound too appealing to me. The second option, you know, we're all dead. It doesn't sound too appealing.

The third option, or they help us get there.

I don't know that they need to help us get there. We do a fine job by ourselves.

Okay, yes, I will take that track also. The track also. Yes, that's valid.

The third option of uploading into the cloud, it's like when. Well, okay, I know there are some unsaid assumptions and we can delve into the theories of consciousness later, but you know, fundamentally I think that option is more like building a simulation of David.

Which is, if you want to look at the matrix.

Yeah.

Where the human body was used as an. As an energy source, yet the mind was put into the cloud to be able to keep the body alive. We could kind of say the body disappears and the mind is still kept alive.

Yeah. And I mean, you know, those are great fun fantasy stories to just explore and be entertained by. But, you know, fundamentally building a simulation of David, it's still going to be, you know, I. Still a different entity. A simulation of David. Even if you had a genetic twin, it would not be David. And so uploading that simulation to the cloud, well, that's all great. I've got a David avatar in the cloud to deal with, but poor David is still dead, so that doesn't sound so appealing either.

Okay, so that leaves us with the.

Only option, which is let's merge as best as we can over the long haul, keeping as much of humanity as possible. Without suffering the weaknesses of the first three approaches.

Yes.

And this goes to the question that you were raising just before this little digression, which is that it's really hard, given the bounds of human nature, to see how humans have the means to enable the desperately needed pace of hyper evolution of our culture that we need just to survive. And it may well be that AI is the only solution to the problems. That AI is creating that.

And I get that, I don't know if you read the book, but the first book that I read before I met Ray Kurzweil was the Age of Spiritual Machines. And I believe the argument was that we'll eventually become machines in one way, shape or form. And I have taken it a little bit further to say that many of us are already machines. And, and let's use that as a broad brush. Our bodies, human bodies, have been inundated or artificially adapted by putting machinery in this. A piece of machinery would be, for example, you had a filling in your mouth.

Sure.

500 years ago you'd be dead without that tact.

Sure.

If we have glasses, without glasses, 500 years ago, 5,000 years ago, you would have been killed by a predator or you wouldn't have been able to survive. If you look at many humans, not all of us, but we have either a bolt or a patch or this. And I've had people argue and I had this in a religious sense in. I was in the middle of the United States and we're having this discussion. I said, no, we are becoming machines already. If you take out the pieces, we might not be able to survive or would not have survived. And during the conversation, the father of a public company that he's actually the father of the kids that were also there of this public company, you hear something near him go beep, beep, beep, beep. And two sons jump up and they go over and they change out the battery pack that supplies his heart pumping.

Yeah, I was just gonna say that. Yeah.

And I didn't know he was on this machine, but they changed out the battery pack and he looked at me. After this whole religious where does it go and how does it stand? And every. This whole discussion for an hour and he looked at me and he said, I guess you just proved your point.

Yeah, exactly. Like, you know, the thing is that again, I think a lot of the knee jerk reaction against the idea that we're gradually becoming machines is, is again just driven by fear rather than reason. I mean, you know, pacemakers have become so. Or Borgs. So. Yeah. Well, I mean, that's what I mean, right? I mean, people have already gotten used to these ideas that you're talking about, like pacemakers or, you know, people developing artificial kidneys or whatever that.

Artificial hips or knees.

Yeah. And so as more and more people have them, then you notice, oh, gee, funny, the fear seems to be diminishing. And you know, that is sort of culture maturity that I think we do. I think people need to realize that the knee jerk fear reactions are actually not so healthy in the sense that again, if you succumb to that kind of reaction, then you're succumbing to the same kind of fear that creates the problem. And we have to like sort of get ourselves above that and understand that, gee, when your mother is surviving because of a pacemaker or, you know, things like that, that. Or it's really not so bad. Yeah, you shouldn't be just following this knee jerk. Oh, we want, we want nothing artificial. I think when it hits close to home, like say a parent or a child and their lives are at stake, that's when people finally push themselves out of their shell and realize maybe this fear of technology is not really all that there is. Right. There's.

I'll add to that, that we went from mainframes to desktop computers to laptops to mobile phones. And while it's a separate entity from us today, some people have done implants. We are this machine. Mobile phone does position itself is our merging with who we are.

Absolutely.

How we work, where we go, what we do, and the AI that's in it, determining how we get to locations, who we connect with in our networks and circles, what medicines we take. It memorizes and learns as. I don't know if you're going to get into it as a child, it's learning to be our AI.

I mean, at the same time, this is the other part of the gradual merging of humans and machines, which is that you already have. Not only the physical merging, when we're talking about pacemakers or artificial hips or whatever, there is also at the informational knowledge level, the mental processing level emerging. That has already been happening for years. The young generation actually has never, ever known an environment, a world without that, which is that many of us can no longer remember phone numbers. I used to be spectacular at that. I can't, You know, I used to have a sense of spatial orientation. Now if you take away, you know, my Google Maps, I have no idea where I am or how to get where I need to go. You know, like people, like when I Drive along. I'm just like on autopilot, following the gps.

You don't even have to pay attention. You just left, right, left, right.

You get there all moved from conscious, you know, system two, mental processing down to automatic unconscious system one, mindless following. Because half of my brain is already been reallocated to this AI, this, this device. There was a news item, I don't know, last year or something, where this guy had landed after a very long flight in Reykjavik in Iceland. And he rented a car, he was very tired. He punched in the street address of where he needed to go to in Reykjavik. He started driving, followed, following this thing. 12 years later. Sorry, 12, not years. Not that bad. 12 hours.

Wow.

12 hours later, he's in the Arctic in the Great Circle, because it turns out there happens to be another street in Iceland of the same name up there.

And he didn't even pay attention to it.

This is what is happening to us, is that our brains are really, really good. They're really plastic, you know, really good at rewiring themselves around the information landscape. I don't know if you've seen these experiments where you can put, like, goggles on. On subjects, on humans that use mirrors so that everything in the world looks upside down.

Oh, no, I have.

Imagine that you put those goggles on and, like, just try to imagine for a moment what it would be like to try to operate in that environment, you know, upside down or left, right inversion, say. Yeah.

Everything you do in a different mode, like.

Like you imagine if you were trying to pick up something you like, everything would just be messed up there. Right? We've tried doing that. You get that if you're trying to operate in a mirror, right? Like, you have trouble sort of getting your hand to go in the right direction. Well, it turns out if you just leave them in those goggles for a couple weeks, they'll figure it out. They don't even notice that everything is upside down anymore. They think everything is right side up.

And that's because we have an interpretation that what we're seeing through our eyes is actually what we're seeing. It's not an electronic. It's not a pulse, it's not a signal. Just like our ears, we think we're hearing, but we're not really hearing. We're hearing a vibration, a vibration of atoms against our ears or sound vibrations against our ears. And that in turn gives us the impression of a sound that our brain interprets. And we don't think that way. We believe we are actually seeing and hearing the way we are.

So, you know, brains are just really, really good at rewiring when the information landscape changes. And. I strongly we are going to find out that young generation who have never known a world without Internet and increasingly have never known a world without smartphones next to you all the time. A world without, you know, Google or, or Maps or any of the other things that we've come to depend on, including social media and the constant feed of AI chosen information for you to be accessing.

So your take was the merge on AI and I could see where you're going so we can pull back on this. I don't know if you've seen the video or know of the individual. I think he was out of the UK in the 1980s, I think late the 80s or early 90s, he took a chip and put it in his forearm. That allowed him to be able to manage his home. Turn on lights, turn off lights. Simple tack, if we want to use that term loosely. And they put it in his forearm and it worked. It worked fine. And before he undid it, what he wanted to do was he hooked it up to his wife with a set of sensors on her head and he thought things and she felt it and she thought things and he felt it. But the real surprise is when they went to remove it, the nerves had already integrated into the chip and that was in the late 80s, early 90s. So to kind of reinforce this merging with AI, I see the other ones as kind of a challenge. The big question I have for you, or my thought goes to there's 7.5 billion people. There's a lot of people who are not going to be missing out on this.

Well, I don't know about this, that the penetration of smartphones is actually insanely remarkable. Okay. You know, you have large swaths of the developing world where, you know, people don't even have their needs met at the, at the foundational two levels of Maslow's hierarchy. Right. They're constantly, literally placed in life or death risk situations and they lack the basic food and water and so forth needs, or they're in an environment where security is just non existent. And so despite that, it is amazing that without landlines, without houses, whatever, they still have a phone.

Okay.

We have done such a good job of commoditizing that this is actually happening. And that just shows some of the irony, right, because we're talking about the second leg of the stool is how our system that we've evolved is based on resource allocation by competition that leads to inequality the problem here is that the first leg of the stool, the resource scarcity mindset, is what's blocking us from tackling the issues, the effects that drive people to lose out horribly in the resource allocation game. The fact is that today humanity actually has more than enough wealth for everybody. We actually can handle taking everybody out of the worst ends of the fear spectrum on all of Maslow's hierarchy. But we are doing an absolutely lousy, abominable job of allocating those resources because we are still stuck in incrementalism in thinking. Well, it's worked for us so far. It should continue to work. We just have to make small tweaks and we can continue to do resource allocation that way and. Sorry, go ahead.

So do you define, do you take this, these three legs and define them as a sort of tribalism?

Well, I mean, tribalism is what you get when you encourage this in group out, group mindset. So long as we remain committed to resource allocation based on fear, which is what in group out, group competition mindset is based on, so long as we think fear is the best way to do resource allocation, then that is going to drive tribalism by definition. Right.

Do you have a new word for the new age?

For what the new.

Where you're saying we merge with AI and we reallocate and redistribute and change our mindset and we move from this tribalism. Do you have a new word for the next iteration?

I think it's going to have to be a little bit, it's going to have to happen in stages. Right. The first thing that needs to happen is that we have to overcome this resource scarcity mindset. Again, the point that you made about the difficulty of getting human culture to evolve at anything but a snail's pace is absolutely valid.

Right?

And so looking at this realistically, right, we're trying to avoid a dystopia, but, but we're trying to do it in a non utopian way. That's the challenge. Right? And so if we look at this really hard, the only way that we're going to be able to get out of this situation is if we address the resource scarcity mindset problem.

You know what my question is? I mean, it's an obvious one. How do you, how do you do this? I mean, we're going in the opposite direction, I think. Protectionism and, and in countries putting up borders and walls to stop the lack of understanding that we are a global world. That video that was out of Germany, where they took all the food out of a Grocery store that came from other countries and the Germans could only eat German food. There was nothing on the shelves. Yeah, I don't know if you saw that video. It was fantastic.

Yeah.

So how do you do this?

Well, well, I think maybe the place to start is to ask how the opposite is actually happening.

Okay.

Because if you look at, as you say, what's been happening, it's been heavily driven by the emergence of AI and related technologies. The, the connectedness that the Internet and mobile technologies have given us. We really don't have the kind of bounds that we used to have again on how quickly the tribalism can be propagated.

So yes.

So AI has been heavily driving the hyper acceleration of the tribalism. And the question is, first, before we even think about how do we solve the problem, problem of, of getting AI to help us to overcome tribalism, I think we need to understand what's happening in how AI is actually until now been, been heavily used to drive tribalism. Right. Because we need to stop that in order to be able to move forward.

He I, I, I, this is almost touching on. We won't go into specifics. Our conversation that happened in the Starbucks on the university is we have groups of individuals and you don't need a lot that are so rooted in yesterday that they don't see possibilities for tomorrow. We also have groups that are so, Excuse me, when I say this guy, this could come back and haunt me. Ignorant about what's going on in the world. That they're more, they know more about Beyonce or a professional athlete and their lives than they know where another country is or the differences between religions or that you can't touch something, somebody and get their gender preferences, that we're talking about changing a large group of people that are so rooted. I've got my own thoughts. What are yours? I mean, how do you, you're saying you break this?

Well, I mean, again, right, to solve a problem you have to understand it. That's why we're breaking down the issues with resource scarcity and allocation and resource balance on destructive capacity. We need to understand the system so that we can understand where we can actually do something. I think probably the best way that we can understand the way that AI is actually integrating into our society to drive tribalism is something that goes way back in human civilization, which is something unique to humans and universal to human. It may not be unique, it's probably in other species as well, but it certainly is universal in humankind. Right. And when you think about things that are universal to humankind there's all sorts of lofty things that people have claimed. There's music and language and reason or mastery of fire or cooking cuisine. But the thing is, I think one of the most universal things, sadly enough, is gossips. Right. This is one of the bullets I wanted to touch upon. Because every single culture in our history has been plagued by gossips. And because of this, the religions of humanity through all the ages have proclaimed stricture after stricture against gossips. It doesn't matter. If you, if you look at Judaism, it's all over the Old Testament. If you look at the New Testament, you'll find it all over. If you look at the Quran, you'll find it. If you look at the, the Pali canon, for in Buddhism, you'll find, you know, the same. That's, you know, it's, it's such a bad problem that all of our human civilizations everywhere have realized that they need to prohibit it because it's extremely damaging. And the thing is, of course, that was when gossips were just humans. But today we've already got AIs populating increasingly the space that we live in, especially the space we just described, where we no longer survive without our screens. And so in this sort of a space, We now have a situation where AIs have already, you know, become our assistants and our secretaries, our navigators, our dictation assistants, our translators, our shoppers, our matchmakers, our journalists, our, our curators, news clippers, house cleaners, chauffeurs, our hedge fund traders, our DJs, our customer service representatives. They've taken over all of these roles that humans traditionally take over. So really, it shouldn't be surprising to realize that they've also taken over the role of our gossips.

Yep, I can see it.

So, you know, our cultures and our religions have been warning through the ages against idle chatter, warning against quarrels and conflicts and gods and slander and condemnation and calumny, all the idol chatter. And now we've gone and created chatterbots.

And companies like Cambridge analytics that, that literally played to the whole game and made it into a business.

Yeah, and, and it's not, you know, and it's many things. It's, it's far more innocuous companies. It's, it's, you know, Facebook, it's Twitter, it's, you know, any, any company that is propagating information online with the aid of AIs that are deciding what information to propagate, those have become artificial gossips.

Yeah. And the algorithms that associated with them give them higher ranking and Status and.

Yeah, absolutely. Now let me break this down a little bit. Okay? So I don't know if you know, like the difference. So gossips do two things, right? A gossip hears and a gossip speaks. And I don't know if you've heard the distinction between quidnunks and gossip mongers.

I've heard gossip mongers, but not quid monks. But why don't you give me your definition?

It's quid nunc N U N C. It's from Latin originally. A quidnunk is a person who.

My Latin's a little bit.

It's a person who seeks to know all the latest news or gossip. It's a busy body. Okay, okay. So it's a person who loves to hear gossip, that's a quid nun. Whereas a gossip monger is a person who habitually passes on confidential information or spread spreads rumors. Right? So that's a gossip who loves to speak gossip. Yeah. Now it's. Artificial gossips are exponentially dangerous regardless of whether the gossip is true or false. It can be equally dangerous both ways. Right? And so when gossip is false or it's private, then our artificial gossip mongers are spreading fake news and spreading confidential information, just like human gossip mongers do, but exponentially, more dangerously. I was just being visited in Hong Kong actually, and I just visited him in London earlier this week. My friend Robbie Stamp, who was the producer of the movie version of the Hitchhiker's Guide to the Galaxy by Douglas Adams. Douglas Adams wrote that nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws. So it's great. Now we've got artificial gossip mongers that give the gossip even more, more reach, more speed. And gossiping fake news and confidential information obviously are super powerful ways to change society's views even in the absence of evidence. Now, for us to understand the consequences when AI enters this picture, it really doesn't take that much. So there was a recent study. I think it was conducted by Yu Pen and the City University of London. If I remember right, it suggests that if only roughly 25% minority of a group, so 1/4 of a group starts spreading an opinion that is enough to tip the majority to switch their opinions.

I've heard this.

Yeah, yeah. This was reported in Science, I think last year. So it's. It's really important to realize that in this age of botnets, artificial gossips easily, easily account for more than 25% of our gossips. The artificial gossips push us Past the big point for fake news to be able to change society's views. And just like humans, the artificial gossip mongers are weaponizing the power of suggestion, just like human gossips. They're spreading unevaluated claims that are based on superficial appearance and are promoting hearsay rather than evidence. It's promoting stereotyping. Now, yes. If the gossip is true rather than false, then you also have artificial quid dunks out there. These are artificial gossips who love hearing the gossip. Right. And they're also gaining social power, just like human quid dunks do. Right. That's why human quid nunks exist, because they love hearing, hearing this. It helps humans gain social power. But now the artificial quid dunks are doing so exponentially, more dangerously.

And so we've got, we've got the, the, the artificial version and the human version. So if we're drawing a matrix, we've got the human doing one and the other and then we've got the artificial version of that same thing. And going back to our spirit, speed of transmission of content and usability and leverageability and capabilities. You're saying that now we've got an exponentially powerful. I'm not even going to use Hiroshima or Nagasaki bomb, but something that could be extremely damaging.

Yeah. This is a really important instance of the general phenomena we talked about earlier, about how AI is exponentially increasing the destructive capacity. And here what's happening is the same exact effects of gossip mongers, both quid dunks and gossip mongers. Actually, those dynamics still hold just as they always have. But with artificial gossips around both the artificial gossip mongers and the artificial gossip, the artificial quid dunks are exponentially increasing that same dynamic.

But even if we just don't talk, let's take that, I'm going to throw out an example and I think that this kind of falls into it. And even in another space, last year or two years ago, someone accidentally posted that, an old article that United Airlines was filing for bankruptcy for some reason and all of the algorithms in the space of the finance sector picked up on it.

Right.

And the stock plummeted.

Yeah.

I believe this is a true story. I don't have time to look it up, but I believe it's a two story and it plummeted. But there was nothing going on with United.

Yeah.

It just was. Someone had posted the wrong document.

Yeah.

And that to me is gossip translating to financial markets, translating to jobs, translating to everything totally that is connected to travel, you name it, buying a ticket.

Right. And here's the thing like this is, here's how it happens. Right? Right. And why it happens. So one of the classics in social Psychology is a 1947 book. I think it's called the Psychology of Rumors. It's by Alport and Postman, and they pointed out a couple of really important things that we have to recognize that AI is doing now. But they were looking at the way humans do it. So they're pointing out one, that gossip search serves to create a sense of cohesion among various groups of people and to take a divisive stance against someone. So gossip actually becomes a social control mechanism that gives power to whoever is doing the gossiping. And in humans, gossiping releases endorphins. It actually gives pleasure. People who propagate gossip are often doing it to help them escape from some other unpleasant routine, from negative feelings and stress.

Because what happens, I'm now picturing people. So I apologize to anybody if they could see what's happening in my head.

And the thing is, the reason this helps them to escape is because when they're gossiping, they become the center of attention for the folks who are receptive to their rumors. Yep. And so, you know, the same thing is happening now. It's kind of like, you know, the Game of Thrones Varys.

Right.

So he's really scary. Right. He's a quidnunk. He's your archetypical quidnunk. But our artificial quindocks are potentially far more terrifying than, you know, that Master of Whispers. And you mentioned Cambridge Analytica. I mean, Cambridge Analytica is an artificial quid nun.

Yes.

And the thing is that when you have these artificial quid dunks, then even Facebook or any company that is collecting this kind of information, Amazon, et cetera, the problem is that the gossip can unintentionally leak. Right. Every single week, one of these companies, including Facebook and all the others, there's news about how many hundreds of millions more records they accidentally leaked. It happens. Just like with human quid dunks. With a human quid dunks, they always leak. Right.

Somehow what they learn, there's no way to stop it. We're creating so much, it's really hard.

To keep a secret. Yeah. And so now we've got great. We've got exponential amounts of information being leaked. And if you think about, you know, Siri or Alexa or Cortana or, you know, Google Assistant, they're all artificial quid nuts. And I'm not saying that these are being built for nefarious reasons. You know, they're clearly extremely useful.

You're just defining them as. You're defining them as what they are.

Yeah, it's just. It can be hard, super hard, to avoid leaking what you overhear. You know, there was that story a couple of months ago where Alexa recently accidentally spread an entire private conversation that it had overheard. So these things will happen. And so today, again, these are things not that we worry about happening decades from now. We have facing a choice whether to upload to the cloud or not. This is something that is already happening and has already happened for a while. We already exist today in a tangled society of both human and artificial gossips.

I think that's a challenge of the conversations I have with individuals as I try to share with them that we are already in a world of AI, very much so. No, no, no, we're not. I said the algorithm that drops down to give you a suggestion on a trip or a place or a word dictates it. The map that comes out when you want to go someplace is telling you the directions to take. It could have said avoid a mall, because there's a mall that someone else likes. I mean, there's so much going on that you are being. It's not manipulated, but you are being guided by AI already today.

Never underestimate the power of suggestion. Yep.

So I'm going to take this that you're saying to understand the challenge, we also have to understand this artificial gossip. We don't know how to stop it yet, but we understand it. We're understanding the resource scarcity, and we understand that we're understanding it. So by understanding it, we're getting closer to being able to find a solution. Doesn't sound like we have one. Is the artificial channel children another category within that?

It sort of is because, you know, if we look at where we've gotten until this point, Right. Basically, the artificial gossips is what's really helping to drive the exponential divisiveness, the exponential tribalism that we were talking about, which is what's increasingly meaning that we're facing an existential threat of extinction event. And so the question is, what do we do about that? Right. And of course, there are approaches that are rather draconian to just like, clamp down on artificial gas, on what can be spread over technology. But I think ultimately it still boils down to us, because the AIs are just amplifications of us, the humans. Right. And so we still need to really think about, well, what is it about us that is making us do these things, and what is it that has helped us in the past to combat these age old problems of gossips in our societies. And so by the way, so there was this fascinating thing that I discovered which is that the west pretty much does not understand that ancient Chinese proverb that everybody's heard, which is see no evil, hear no evil, speak no evil. Right? You've heard that?

Yep. Yes.

But in the west that is pretty much always in my experience misused. It implies avoiding moral responsibility. It implies looking the other way, which is a complete misunderstanding. That proverb is actually a Buddhist summary of a Confucian tenet. And let a more accurate restatement of this, taking this from a better translation, is look not at what is contrary to propriety. Listen not to what is contrary to propriety. Speak not what is contrary to propriety. Make no movement which is contrary to propriety. Think about that, right? This is what, 2nd to 4th century BC what it's doing is, it's reminding us to avoid dwelling on evil thoughts and to delight instead in creating concord. Now it's by no means unique to Eastern philosophy. Right? Marie Curie said be less curious about people and more curious about ideas. These are the sort of directions that we need to find ways to amplify if we're going to succeed in this age of AI driven cultural hyper evolution, we also have to amplify these things.

I guess the way I'm looking at it is we're going to need AI to solve AI.

Yes. I, I that this is exactly.

Because we can't, we can't figure this out.

With, with our, with just ourselves. We don't stand a chance at combating the exponential power of AI gossips being used the other way.

We don't have the human capacity to understand the complications and the interconnectivity that is happening when it comes to all of these. Not just gossip, but the, the, the arming and disarmament of individuals and, and the capabilities. I, I go to the, what is it? The game. The AlphaGo. The AlphaGo and I'm assuming you all listen the four or five games the computer won, five out of four out of five. And the people who are watching it being created said that they saw new iterations of the game that no one had ever seen before. So therefore it was being created, it was creative in creating new games. I didn't know about this until recently. That alpha next about 40 or 50 days later. There's some timeline I'm propagating, probably some fake news, but in 40 days it learned the game and it beat the AlphaGo 100 to 0 because it came up new ways to beat AlphaGo.

Yeah.

And then the Facebook. Facebook was having that computer simulation that was happening and. Or the computer technology, they were trying to move two dots together to a certain point. And at some point during the experiment, the computer started creating a language that no one understood what was the computer was talking about. And they disconnected it because the computer was creating its own. So it's almost as if to solve this challenge, we have to use AI to find what we are not going to be able to see.

We do have to. That said. Okay, forgive me a small digression. That last story about computers creating their own language.

Yeah. You would know better than me that.

Okay, that is the sort of story that drives AI researchers batty.

Yeah, I just used it.

I mean, here's the thing. In machine learning, there is, you know, it could be a neural network. It can be many other kinds of models. Whenever you're doing learning, you're creating inside the machine learning model new representations of the situation that are useful in accomplishing whatever task you're learning to accomplish, whether it's prediction or whatever else, translation. And so in a human, that is what your brain is doing as well. It's creating new representations for information that are more useful for making the right generalizations and the right predictions and solving the tasks that you're trying to learn without creating new representation languages. There is no learning without creating new representation languages. There is only dumb, blind memorization. All you can do is record the facts, the input, that's it. Any sort of generalization, any sort of intelligence, any sort of learning requires, by definition, creating new representation languages internally. So when a story like that comes out and a journalist sees, oh, wow, here's a clickbaity way to write about it, they're inventing their own secret language. Oh, scary. Sky deck.

Yeah, yeah, yeah.

And then it goes viral. And then the entire world of AI research has a collective migraine. And after a couple months, it finally dies down and we all breathe a collective sigh of relief. Oh, God. That nonsense went away. And then a few months later, journalist says, hey, I wonder what happened to that? And then writes another story. Look, the university had finished that experiment, so they stopped that experiment now and says they unplugged the machines that had created their secret language, and it goes all viral again. And migraine 2.0 happens. Oh, my God. Stop, please.

Well, I guess. I guess we could say that there's maybe a different way to look at this, and this might be a lesson in the, in AI space or in general, that learning doesn't happen happen without new input.

Yeah.

Or movement doesn't happen without new input. You have to have some type of input. You're sitting in a chair, you won't breathe unless your body gives you a signal. You need to, you're running out of oxygen. You don't move your body unless you're sitting too long or there's a timer goes off or you know, something in your head. There's always signals. And I always, my context is if you want to create something new, you have to learn something new. And in learning something new, you create something new. Now it doesn't have to be new to the world, it just could be new to you. So what you're saying here it sounds like is in order for computers to do what humans do, they get input, they interpret it, they create a new set of acts, and that new set of acts creates the next information capability. So creating a language or creating some, something that people don't understand doesn't mean that it was a runaway train. Absolutely, that makes sense.

Absolutely. I mean, if you were to cut open your brain, there would be all sorts of representation languages happening in your brain that we don't understand.

Yes, and I completely understand what I just, I felt like a foot, foot and mouth at the same time. I love how you analogized it to the learning experience. And then my brain connected the dots and said, okay, you have a model or belief. Belief is triggers. Triggers create new experience. Well, the computer would do the same thing.

Yeah, so.

Right, it makes sense. Okay, so I want to go back to, is there anything on the children that we should cover?

I think the question that you raised is the right question. So now what do we do about this problem? So we've identified, we've broken down some of the deeper issues and the question is, how do we combat this? And I think there are two dimensions on how we try to counter the exponential risks that we're facing. One is of course us humans, Humans do have to grow up faster than we're doing before the AI, the hyper accelerated AI destructive capacity wipes us out. And then the other half of that is, well, maybe how would AIs help us to do that in the first half? I think we really do have to realize that AIs are just an amplification of ourselves. We were just talking about how the way machine learning works is actually increasingly getting closer to the way human learning works. We've got quite a ways to go, but we are in the machine Modeling, very much. Already learning. And if you think about today's society, as we've just mentioned, AIs have already become very much integral members of our society. Right. They've taken over so many of our human roles. There's probably more AIs now than there are humans in our societies. Because, you know, your. Your Facebook AI is really a different AI than my Facebook AI. Right? They've learned very differently. They've been learning from different data. Your Facebook AI, AI is learning from your environment, mine is learning from my environment, and so on. Right. And so for each human that has an account on any of these things, you've got an AI. And these AIs are not just passive. Right. As we've discussed, these AIs are actually already. Even though we're far away from the singularity, even though we're far away from strong AI, even these weak AIs are already influential, active, imitative, integral members of society, just like humans are. And they are learning. I mean, today, they're already deciding what ideas to share, just talking about what, what means, the spread. They're deciding what attitudes to reward more influentially than probably 90% of humans. And they're learning how to do that from watching us. They're observing.

Not a good role model, not a global. Well, maybe they're watching you. They're not watching me. Okay, maybe you could cut a distinction. To them. It's not a good role model to be watching humans. It's the old Star Trek. It's the Star Trek where we send these signals out into space and they realize that we're horrible creatures on this planet.

Except these AIs are actually. They don't know that yet.

Okay, well, let's hope they don't find out soon.

They're like any child. They're learning from what they see you doing and they're learning from what you ask it to do.

And so we're imprinting.

While you were watching, your AIs have already snuck up on you and adopted you. Yeah. So you, all of us, are parenting our artificial children already.

The imprinting. The imprinting. Yeah.

The question, the key question here is, how's your parenting?

Are you asking?

Well, yes. I mean, this is. It's hard to overstate the ramifications.

No, no, I completely get it. I. I completely get it. And in my head, I'm going from the macro to the micro to the macro. So I'm utilizing when I get up in the morning, how many times I snooze, what sounds the computer is hearing. When I eat my food, if I step on a scale, if I don't step on a scale, who do I talk to, who do I text to, how do I text, how engaged am I, how late am I, how early am I, how prepared am I, what are the topics I talk about, what, what do I search? And because I've got, I'm an Android and a Mac user on both sides, so I've got my Chrome learning certain skills which translate to my Samsung phone and then it's crossing over to the Mac which is looking at it from another AI perspective. My mind is racing of how, how my children are learning.

Yeah, wow.

There's a lot going on.

There's a, there's a lot, a lot going on. And, and this is going on even though the, these weak AIs are still very weak, right? They're, they're, they're weak but they're big and they're dumb. Right? So they're big and dumb. And so we've got these big and dumb artificial children that are quickly reshaping the consciousness culture. And the culture that they're reshaping is the culture that every generation successively of stronger AIs will be learning under.

Yeah.

Now, I mean people have said to me that this can't possibly be true. Machines should be, you've got to see machines as slaves. I have a colleague who's written entire treatises about this. Machines aren't yet self reliant, they're not self replicating, they're not independent. But here's the thing, none of those are required factors. Just think back to how much our societies have historically been influenced by the cultures of slaves, by the cultures of eunuchs who are not self replicating or the cultures of, of colonies who are not independent. Nevertheless, super powerful evolutionary pressures on society have been exerted by these actors who aren't independent. And just like that, our machines culture will absolutely change us.

And it's not a point to argue today, but there is no doubt in my mind that we will create an algorithm that will create an entity or we'll create a new, a new form of something that will be able to self replicate itself in one way, shade or form. Because we're already creating automatic AI generators and we're creating AI cleaning technology so we can take a messy data and clean it. So there's nothing to say with the automatic AI generator that eventually it'll be smart enough. And let's use the term smart in a human sense, but not in a smart technological sense. It'll be capable, maybe that's a better word. It'll be capable to create a version of something that allows it to be able, in these 20,000 years, as we talked about very quickly, in a matter of a few lifetimes, in computer terms, it will be able to create something. Yeah, no doubt.

Absolutely. And the thing here is that it's not. You know, the ramifications of this are in the. In the fact that these are machines that have opinions and they can actively shape our opinions. Right. You know, even in the first part of this discussion, we're talking about the existential risks of AI and automation and destructive capacity. We were still thinking of them as machines, as mechanical tools, as passive slaves. And the thing is, we don't notice this fundamental difference when the machines have opinions and can shape our opinions. AI.

You just stepped over a line. We talked about this entire time about AI, and you just talked about a live AI.

Well, it's still very weak. I don't want to make you play conscious.

You called it an opinion.

You called it, but it's an opinion.

And once you said it was an opinion, it moved from my head from a tech to something that creates. I am writing an article right now that creativity is a myth that computers won't be able to do. And you could probably argue the same point.

Yeah, computational creativity is one of my research areas. I totally.

Oh, my God. Vox in 2017 started using software to generate videos so that they didn't have to. With no AI, with no human interaction. So when you watch a VOX video that goes on for 10 minutes, that might have been all created by AI, and that therefore that's created creative. Because that's videography and that's journalism. And I can give series and you can give the same. I know that. So. But yet when you referenced it, when you referenced AI, you referenced it as having an opinion. And to me.

And can actively shape your opinions.

Yes. And that to me, changed my entire view of Ann.

See, this is the problem that I have with, with the term fourth Industrial Revolution. I mean, in some ways it's a great term, but the thing is, it also reflects a mindset that I think is becoming out of date, because it's a mindset that's stuck thinking of machines as passive tools. Right. There's a comforting idea implied in that label that says, oh, this is the fourth Industrial Revolution. We've managed three. We can take the lessons of history and apply them to the fourth. We know how to do it. But no, the Industrial Revolution was the automation of muscle AI revolution is the automation of thought.

You know, the project Moonhot is, we call it the age of infinite. Yeah, different jump. What was that? You're going to say third one was the. I stopped you, I apologize. What would you call the third one? The automation of what?

No, I mean the first three in this framing of AI as the fourth industrial revision, all of the first three were about automation of muscle. But the AI revolution, where it's about the automation of thought, it's not really the fourth of anything. It's the first unlike. The dynamics are like, okay, what we saw before, because it's not only about automating muscle. Sure, robots do some of that, but more importantly, it's about opinions, automation of thought. And so this is not a situation where the analogies to, to past approaches toward dealing with a social impact of mechanization can apply. So how do we, how do we.

Get to this, this mindfulness?

How do we deal with this? Yeah, so, you know, we, we have to sort of. So some people have, have suggested that when we build AI systems, we need something like a moral operating system. Right. It's kind of like Isaac Aspen to what culture? Oh, there's not.

Everybody's the same.

We'll have to. That'll take a separate episode to talk about. But, but you know, like Isaac Asimov's classic Three Laws of Robotics, if you remember, it said, actually he added a fourth. So he added a 0th law first, which is a robot may not harm humanity or by inaction allow humanity to come to harm. Right. And then there was the first law. Robot cannot injure a human being or through inaction allow the human to come to harm. Second law says a robot must obey orders given to it by human beings. Unless there's a conflict with the first law. The third law says robot has to protect its own existence as long as the protection doesn't conflict with the first two laws. Right. Now here's the problem. You notice how legalistic this is. It's very rule based. And the idea is we hardwire AIs with these ethical principles so they can't do the wrong thing. Problem with this whole approach, which people are still advocating a lot of people, is that it's a pipe dream. It cannot work.

Oh no, there's no way.

First of all, it's not just AIs that we need to fear. It's human cultures. It's human cultures armed with AIs.

I had an AI person at the Key Club where I had a discussion about AI and the woman was arguing that computer, that computer scientists won't do the things that I was saying. And she couldn't. She said, we have rules against this. And I said, you're kidding me, right? That some, some, some person in Mongolia or the States or China or any country of the 195 around the world, 193, plus the two others that fall to a different category, there's someone in the world's gonna do something that's contrary to any of those rules. Yeah, to pipe dreams.

Well, the other problem is this, and this is maybe even more important. I mean, this whole idea still comes from an obsolete picture of AI back from the 70s and the 80s, where AI was logic based, rule based, very much like a computer program is traditionally written. And this is where all those Hollywood stereotypes come from, where a machine talks like this and it's very mechanical and has no sense of emotion or context and cannot understand anything, which is not a logic rule. Right. So you get all of that kind of Spock like stuff, Mr. Data like stuff. And to this day, it's still, still in the movies, which also gives a lot of researchers and AI migraines all the time, because modern.

Do you need some aspirin? Is that what you're asking for?

Massive amounts. So, like, the thing is that modern AI is all about machine learning. That I wrote one of the first PhDs way back then in the late 80s, early 90s, that was arguing that, no, we need probabilistic models that are learning models to build AIs that can really deal with human language, which is the foundation of human intelligence. And since then, the field has completely taken off in the machine learning direction. And critically, you cannot hardwire logical rules into machine learning any more than you can hardwire human kids, because by definition, they're adaptive.

They learn the culture around them just for the sake of context. Because you and I are having this discussion, but someone might be listening. I can give definition of machine learning, but you're the expert in this. So can you give a very short summary of what is machine learning, just so that the context is understood as to why that one versus this works?

I mean, machine learning is of course the machines that are doing learning much as a human does learning. And today it's hard for many people to remember that AI, for decades up until the 90s was dominated by the opposite paradigm where the intelligent behavior was the result of humans actually writing programs that would do the, the tasks that the AI system was supposed to do. The AI system never learned how to do those tasks. You actually Wrote the complete recipes for how to do those tasks, just like, you know, cookbooks do. And yes, and there was no learning going on. And so for some of us, that was just so clearly wrong. It's unscalable. And the models that were totally based on logic could not handle degrees of gray. Everything was black and white, zeros and ones. They clearly weren't going to solve the AI problem by themselves. And so we needed to bring in alternative mathematical modeling approaches that a. Would be sensitive to, you know, statistical correlations the way that human kids are, and gradually learn by seeing statistically relevant patterns. Right. And so all these models become probabilistic. They become driven by continuous math instead of Boolean logic. And, and neural nets are examples and many other machine learning models are examples. And so the, the slogan is sort of to don't. Don't model the task when you're writing an AI. Instead, model the mind. Don't, don't. Don't try to write programs that model the whole world. You're never going to finish. And it's all going to be wrong because it captures no contextual shades of gray. It's really bad at dealing with ambiguity, which is all, all over us, which is all through our languages. Right. You need to have a model of the mind that is adaptive, context sensitive, and learns. So that's kind of what machine learning models are. And that's what's created the AI revolution that we're experiencing today, because that works.

And let me. I don't know if this is exactly the case, but, but to give it. I think my history or memory is that there is in learning, there's supervised, unlearned, unsupervised, and reinforced. Yeah, but is that still the same?

Those could be useful, rough categorizations of learning algorithms, but they're, you know, there are a lot of important types of learning where you can see it either as supervised or as unsupervised. And likewise for reinforcing. I mean, those are just kind of theoretical notions that help us to roughly group different families of learning algorithms. But yeah, not, not to digress on that too much. Okay.

No, I just, yeah, I wanted to give context and I think those words even helped. And you just described that they could cross over. There could be models and versions of it. And that was, that's a simplistic way of looking at it. So let's get back to that topic. How do we solve?

So, so, so, you know, given that we, we cannot hardwire logic rules into machine learning. Right. Because. Because your Machine learning model, model of the mind. Right. So you can't exactly unscrew through the top of the head of a human kid and like solder in some logic rules. That's not how it works. There's no place to put it. There are massive collections of neurons and they're constantly adaptive. There's no line of code that says kill a human that you can try to disable. Right. All the information is represented in dynamic activation patterns across large numbers of neurons or artificial neurons. And they're just learning the culture that they're surrounded by the data that's coming in. So what do we do then? Because what that means is that whether they're machines or human, whether they're artificial or human children, morals and ethics and values have to be culturally learned and sustained. It's the same for humans and machines. So here the key question is, what culture do we need to be teaching our AIs if we don't want ourselves to destroy each other? Armed with that mushrooming AI power we were talking about. And the culture.

And I like how you said dystopian versus utopian, because there are a lot of people believe that AI might bring a utopia to the world. And I'm doubtful of that.

Yeah. So you know, if we want to answer this question about the culture that we need to be teaching our artificial children, well, I mean, let's just think about what you would teach. If you have human children, you want a world where you can evolve in a healthy and peaceful way. Multiple cultures. So the cultures are co evolving under a situation of constructive, continual generation and evaluation of new ideas and new means. You want the cultures to support healthy generation of a wide variation of these ideas. And to do that, we need to raise artificial children AIs to value diversity and creativity and respect and open mindedness. That's what enables a wide variation of ideas and means to get generated in the first place. But at the same time, same time, for our cultures to support healthy evaluation to yield a sound selection of ideas and means, we also need to raise our artificial children to value fact based empiricism and reasoned, informed judgment. And so compare this with how you are parenting your artificial child. Think about creative diversity of opinion. What we're raising our AIs to do today is pretty much the opposite. We're teaching our AIs to build Echo chambers where we comfortably hear only our existing perspectives. Because every time you click like or favorite or share or retweet, we're teaching our AIs that we only want to Listen to ideas and memes that we already agree with. We don't have buttons for might this be right or could this be onto something or not. Sure, I agree, but it's an interesting thought. Is that how you would teach your kids, your human kids just ignore or to suppress any viewpoints other than your own?

I come from a culture where we talk about over the table. I at the same time as everybody else, we challenge adults, we look at the world and question so much and education is a core belief structure in the way in which I grew up. So I would think that yes, we would like to have and I like these ideas of these other buttons. They're fantastic. I yeah, mind is racing around the.

World we would want to have what enables that kind of dynamic you're just talking about is open mindedness and respect. The problem here is again, today we're raising our artificial children to do the opposite. There are those really obvious examples like when users deliberately taught Microsoft tay to be effective offensive and racist two times in a row. But that's extreme, right? The thing is that even on normal days we're constantly teaching our artificial children to reward trolls and to reward offensive insults and hate speech and so on because those posts get more views, more likes, more fame and we don't have buttons for this is not a very respectful way to communicate or maybe reword this, please. And so again, would you teach your human children such vindictive closed mindedness?

Well, and I think we even have an extremism in that. So let's use Instagram or LinkedIn where you look at I post articles and I post quotes and I post other information and I know that when an article I might be interested in it, but I don't look at it at that time or the moment. So I'll say I'll get to that later. So I didn't do anything. Even though it was a value to me because I didn't do anything. The computer doesn't know that if I see a quote I might grab the screen before I'll forward it. And I know people have told me they do this in my quotes is they'll forward it. So it doesn't even show that you're interacting with it. And there's so many times that you just scroll past it and say wow, that's great. So we're even teaching the computers to go extreme. We either like it or we dislike it, even though normally we're saying oh that's nice, that's average, right? So we're not even teaching Averages in normalcy. We're teaching extremism.

Yeah.

Make sense.

Yeah. And, and you know, so like we're, we're doing even, even, you know, like the best of us are constantly teaching our artificial children exactly the opposite of what we would teach our human children. Right. And so like also consider, you know.

You'Re making an assumption there. You're making a huge assumption because I, I've got to believe there are families who are teaching. You don't like this group? You don't like that group. Those are out of country. These are people that are, I mean, I've got to believe there are families doing that.

I did say even the best of us.

Okay. I'll give you that one.

You know, and that extends also to the, you know, the other half of what we would want for a healthy co evolution of cultures on the planet. That in a survivable way. Right. Which is sound and informed. Just judgment again, you know, we're raising our artificial children to do the opposite. We're not teaching.

Are you telling me we're going to use AI to, to, to. To kind of break it down? Just we, we've been having a long conversation, which I love. I just. Are we tell. We're going to use computers to teach the computers to tell it to do. Are we going to add new emotic emoticons? Are we going to. What are the things that we do? Is there a. Anything we can do today besides trying to be nicer to our computer?

Oh, well, no, absolutely. Today that is what we should do. That is what we absolutely need to do that. Right? I mean we've talked about fake news. I mean false memes account already for the majority of what AI has learned that it should circulate. Right. PolitiFact and other fact checking websites have been tracking this for a decade. PolitiFact rated 47,7% of shareable Facebook memes as either false or liar, liar, pants on fire and only 20% as true or even mostly true. And for, and for chain emails it's even worse. It's like 83% are false or pants on fire, but only 7% are true or mostly true.

I don't forward, I don't do that. One of the challenges I've run into.

But what you don't have, I've run.

Into this, this abundance of content. So when I see an article and I'm researching something and I look at it and I read, I say, wow, I didn't realize this was happening. And I said wait, wait, let me see the date of this 2011. It's post crisis, so I don't even know what I'm reading because the infinite amount of material is also overwhelming me.

Yeah. No. And even if you find something that you know is false, you don't have a button for this is factually wrong, or here is the evidence for why you should not make this viral. Now, these are all things that you raising your human child would teach. And yet we are not teaching our artificial children that are exponentially more big and dumb. So what can we do today? One of the primary things that we can do today, all of us could do today, is to start taking our parenting responsibilities seriously. Every single one of us has to actually start parenting, not just using our AIs.

Now, you are a nice person.

I know what you're going to say. And before you say that, let me just point out one thing. All right? You're. You're. You're gonna tell me. Humans, you know, they're gonna, like, just not understand. They're not going to do this. Humans are. Even if you know why, why are humans not immediately receptive to this idea that they need to parent their AIs responsibly? It's because today, the way that the AI narrative is being shaped today, people are still thinking of them as parents, passive slaves, as machines. They're not taking them as children that need to be parented. But imagine, but think of. And people say, that's ridiculous. Even if I did what you're saying and I parented my AI correctly, there are seven and a half billion people on the planet. There are billions of other kids. And so what difference will it make how I parent my artificial child? But I want to ask you one thing. Have you ever heard. Have you ever heard any human parent say, what difference does that make how I parent my child? There are billions of children in the world. Makes no difference how I parent my child.

I. No, I have not.

If we had approached parents parenting that.

Way, actually, I have. So I have got to start live in Hong Kong. We live in Hong Kong, too. We give other people the responsibility of parenting. People give responsibility to other people, their parents or child.

Another. A different conversation. We'll have that conversation. A different. But here's the thing. If humanity had that mindset, what difference does it make how I parent my child? We would have gone extinct long ago. Humanity would have destroyed itself long ago. The only thing that's kept us going up until now is because every human parent actually does take responsibility for parenting their child properly.

I don't disagree with you on the conceptual side. It would be a phenomenal experience experience if we made that happen. So mine and I, we don't have time for it today. I'd love to have this conversation at another point and if you're interested in the help on some of the projects you're working on, obviously I'm very interested. The this is a question to me, maybe it's not to you. I'm assuming it would be of the speed of indoctrination, the speed of acceptance, the speed of understanding. And I don't know if it's. I'm not so confident that humans can embrace that AI is a child. I think they can conceptually understand it. And yet they're still going to get drunk on Friday night or they're still going to.

Which parsute are their human children too? No, nobody's perfect. We're far from perfect.

I'm just looking at speed.

But I gotta tell you, okay, I gave this message in a TEDx talk a couple of years ago and last year a couple from a far distance saw me at some point and they ran up to me, they said, you know, we just wanted to tell you that we heard you talk about this a year ago and we wanted to let you know that ever since then we have been speaking incredibly respectfully to our Alexa and our Siri. So important to us.

So the question you should ask, what.

Were you doing before? Well, have you heard the way people talk to Alexa and Siri?

No, I haven't, but I know now I'm wondering. Okay, I, I'm actually thinking of our mobile phone and the words we use and the way in which we search and the way in which we migrate. And I think that teaching Alexa that Alexa becomes part of the family. You talk to Alexa, ask questions questions. Maybe as we get user interfaces become more voice oriented, then the capabilities will be enhanced where we'll have to make different choices. I thinking of just the texting and the pieces and the usage and the caps and the emoticons that we use. So I love this. I could go on and on and on with this. This is fantastic. And what I'm going to do is we're going to break it here if we would like to pick this up and do the another follow up with something. This is called redefining tomorrow. This is redefining tomorrow. And I'd like to find a. I'd even like to find or help you find some solutions to some of the challenges in the way you brought them up. This was Fantastic.

Thank you for having me.

I had a great time.

That's wonderful.

So, yeah, it was fantastic. Thank you. Thank you. Thank you for being, I say, on the program, but thank you for talking with me because I learned a tremendous amount and I'm hoping that other people have felt the same. So my. The. I always give the three. The three prayer symbols or whatever, the thank yous. So I'm giving that to you as an emoticon on a video, on an audio.

So thank you.

Okay. So I wanted to thank everybody today who listened in and heard the conversation. I hope that you learned something from today that'll make a difference in your life and the lives of others. And let me expand on that today. Let's say we hope you learned something today that will make a difference in your life, the lives of others and the AI that you're growing or teaching or developing. So always remember, you can't fix yesterday. You can always only create tomorrow. So let's work on tomorrow and let's make that an amazing time for not only us, but for our children and their children, the children after that, and every other species on this planet if we want to go further. So love to connect to. If you'd like to connect, text DavidAvidGoldsmith. Email David DavidGoldsmith.com you can Instagram, I've got Mr. DavidGoldsmith. You can connect on Twitter. DavidGoldsmith, LinkedIn, Facebook. I'm all there. And dakai is spelled D E K A I. Correct?

Yes. And I'm @Dakai123 on Twitter.

Dakai123 on Twitter. So I'm David Goldsmith and thank you for listening.