Latest / Ethical Machines / Season Finale: AI Paradoxes

Transcript

Season Finale: AI Paradoxes · Ethical Machines

0:01
[SPEAKER_01]: Okay, here are three claims I'm going to give you.
0:02
[SPEAKER_01]: Ready?
0:03
[SPEAKER_01]: These are three, what you might call paradoxes.
0:05
[SPEAKER_01]: Paradox number one.
0:07
[SPEAKER_01]: The more AI can do, the more it highlights the irreplaceable nature of human intelligence.
0:12
[SPEAKER_01]: That's number one, the more AI can do, the more it highlights the irreplaceable nature of human intelligence.
0:18
[SPEAKER_01]: Okay, that's paradox number one, paradox number two.
0:20
[SPEAKER_01]: It's called the justice paradox.
0:22
[SPEAKER_01]: Less bias doesn't always mean more justice, maybe more specifically less bias in AI doesn't always mean more justice.
0:29
[SPEAKER_01]: That's number two.
0:30
[SPEAKER_01]: The third and final paradox.
0:32
[SPEAKER_01]: call it the super intelligence paradox.
0:35
[SPEAKER_01]: The more we chase artificial super intelligence, the more we realize it consists not in the intelligence of the super intelligence, it's just consistent, the intelligence of individual human beings, but rather in human beings collective intelligence.
0:51
[SPEAKER_01]: So the more we look after for we go look for super intelligence, the more we find it not in AI or some future version of AI, but rather we find it in
1:02
[SPEAKER_01]: across the centuries across millennia.
1:04
[SPEAKER_01]: Okay, those three paradoxes are three of the seven paradoxes that we are going to discuss today.
1:11
[SPEAKER_01]: They're three of the seven because there are seven in the book called the AI paradox, how to make sense of a complex future.
1:18
[SPEAKER_01]: The author is my guest, of course, her name is Virginia Dignum.
1:22
[SPEAKER_01]: She's a professor computer science at Umea University in Sweden.
1:26
[SPEAKER_01]: She's a professional responsible AI and the director of their AI policy lab.
1:29
[SPEAKER_01]: She's also a member of the UN High Level Advisory Board.
1:33
[SPEAKER_01]: So this conversation is, of course, about each one of those things.
1:36
[SPEAKER_01]: It's about each one of those paradoxes and why she thinks that human intelligence is not going to get replaced by AI in short.
1:45
[SPEAKER_01]: It's far too rich, complex of a thing to be, if you'll just recreate it by an AI or frankly anything like an AI.
1:53
[SPEAKER_01]: She thinks that when it comes to talking about bias and AI, we really should be focused more on something that's upstream of that, which is the kinds of things that led to our data being what it is, which then in turn leads to bias AI.
2:06
[SPEAKER_01]: So we should be paying more attention to broader social political issues that we should about bias and AI.
2:13
[SPEAKER_01]: And again, let's note about what we're this insight about.
2:16
[SPEAKER_01]: super real super intelligence consists in the collective intelligence of humanity.
2:21
[SPEAKER_01]: So we're going to talk about all that and more in this season finale.
2:25
[SPEAKER_01]: Don't worry, no other seasons coming back in, I don't know, let's call it mid-September, but this is the last episode for this season and I think we're ending on a very high note.
2:34
[SPEAKER_01]: Before we get into that, of course, while we're on break, you could read my book.
2:38
[SPEAKER_01]: The Ethical Nightmare Challenge, find it wherever you buy books.
2:41
[SPEAKER_01]: And you've got suggestions, comments, criticism, or whatever, email me at EM at readblackman.com.
2:48
[SPEAKER_01]: And that's it.
2:48
[SPEAKER_01]: Go read the Ethical Nightmare Challenge.
2:50
[SPEAKER_01]: Email me if you got things to say.
2:52
[SPEAKER_01]: And now let's go listen to Virginia.
3:08
[SPEAKER_01]: So Virginia, you recently wrote this book, it's already out called the AI Paradox.
3:13
[SPEAKER_01]: Each chapter is another paradox.
3:15
[SPEAKER_01]: And we're not going to explore every paradox that you do in that book, but let's discuss at least some of them.
3:20
[SPEAKER_01]: So let's start with this one.
3:22
[SPEAKER_01]: Is there a first one?
3:23
[SPEAKER_01]: The more AI can do, the more I like the irreplaceable nature of human intelligence.
3:30
[SPEAKER_01]: Of course, isn't as against the basic line of thought by the technology company, the more AI we have, it becomes smarter than people.
3:37
[SPEAKER_01]: They replace people on a job, etc.
3:39
[SPEAKER_01]: But you think it actually highlights the irreplaceable, what is it?
3:42
[SPEAKER_01]: The irreplaceable nature of human intelligence.
3:44
[SPEAKER_01]: So, yeah, explain the paradox.
3:47
[SPEAKER_00]: Maybe it's starting a little bit before that.
3:49
[SPEAKER_00]: I wrote a book because exactly I was getting a bit fed up with the way that AI has been explained and has been sent down our throat by all kinds of ways.
4:01
[SPEAKER_00]: And I thought I was time to explain things of it in a different way and also putting a couple of,
4:07
[SPEAKER_00]: Maybe not completely paradox, but a kind of challenges and controversies highlighting those.
4:12
[SPEAKER_00]: So it starts indeed with the idea that the more AI can do, the more we'll see what exactly AI is not doing yet.
4:19
[SPEAKER_00]: There is a world definition of AI, which already from the 70s I believe, which says that AI is what AI cannot do yet.
4:28
[SPEAKER_00]: And that means exactly once we solve the challenge, then it becomes just computing.
4:34
[SPEAKER_00]: It's not anymore, something people in AI concerned about.
4:39
[SPEAKER_00]: And we have seen that happening from the challenge around playing chess, which was in the 80s and the 70s, a huge singing AI.
4:46
[SPEAKER_00]: I know there is no one, it's really concerned with the playing chess.
4:50
[SPEAKER_00]: Manning and mapping and searching was another issue.
4:53
[SPEAKER_00]: which now any of us use Google Maps or whatever with not even thinking that we are using AI.
5:00
[SPEAKER_00]: And the point is that all time, and that is the explanation for the paradox, we are building systems that are quite good and very advanced in some type of cognitive capability.
5:15
[SPEAKER_00]: Human intelligence is much more multidimensional than just cognition.
5:20
[SPEAKER_00]: We are intelligent in many different ways, especially musically, socially, emotionally, you name it, there are a huge broadness to human intelligence, but we are building systems in one way only, and we are building these systems to be
5:39
[SPEAKER_00]: more able cognitively, more able to solve some cognitive problems.
5:45
[SPEAKER_00]: That is not all what our intelligence is.
5:48
[SPEAKER_00]: And therefore, the more we are solving those type of issues, the more we are realizing that our intelligence includes other capabilities, other dimensions that are not being covered by the systems.
6:01
[SPEAKER_00]: I also I liked there another paradox, the more a deaf paradox,
6:05
[SPEAKER_00]: which says that basically while we are making systems better and better at cognitive capabilities, they are still very bad at very simple things like pick up a cup or turn around or sit on a chair, so all kinds of very simple and almost automatic things that we do all the time without really sinking even at its intelligent, that's where the systems are still very far behind.
6:30
[SPEAKER_01]: One, let me echo at the point about our eyes, whatever we computers can't do yet.
6:34
[SPEAKER_01]: My favorite example is actually, Dave Carter, the philosopher and I, Dave Carter, who hundreds of years ago said, you could never have a machine engaging in mathematical calculations.
6:42
[SPEAKER_01]: Just can't, that's reserved for the human mind alone.
6:46
[SPEAKER_01]: Probably divine mind as well for him.
6:47
[SPEAKER_01]: But then of course, we've got calculators.
6:49
[SPEAKER_01]: No one's like calculators are so amazing.
6:51
[SPEAKER_01]: So contrary to surprising human cognition.
6:54
[SPEAKER_01]: So the more you spoke, the more it felt like,
6:57
[SPEAKER_01]: I wonder if you can put the paradox slightly different, if I rephrase it only slightly with one award whether you agree with the, if I'm getting your point.
7:04
[SPEAKER_01]: So the word you put it is, the more AI can do is the more highlights the irreplaceable nature of human intelligence.
7:09
[SPEAKER_01]: I think I kind of want to say that what you might mean,
7:13
[SPEAKER_01]: And so there's something different, the more highlights and the breadth or the multi-dimensionality.
7:17
[SPEAKER_01]: I'll tell you.
7:18
[SPEAKER_00]: Yes, but I do believe that it is also why I like the kind of things that the machines cannot do yet.
7:24
[SPEAKER_00]: And I do believe that there is a limit to what we can do with them.
7:28
[SPEAKER_00]: It's not only believe, there is...
7:30
[SPEAKER_00]: There are many formal studies that show that there is a need to computation.
7:34
[SPEAKER_00]: There is only so much that can be computed.
7:37
[SPEAKER_00]: And in a mathematics, it's very clear what is a computable function and what is functions that are not computable.
7:45
[SPEAKER_00]: A lot of what we do, and let's say those lower or less, less cognitive parts of our intelligence, which are still intelligence, nevertheless, are exactly in this type of non-computable aspect.
8:00
[SPEAKER_00]: And therefore, I would still maintain that those are irreplaceable, at least with the tools and the type of computation that we are using nowadays.
8:09
[SPEAKER_00]: And the more we see that the system is doing some stuff, the more we'll find out, oh yeah, this system cannot do it.
8:17
[SPEAKER_00]: We will continue finding those forever.
8:21
[SPEAKER_01]: So when you say you're a placeable, you mean something like you're a placeable given to the given the technology and the data right now.
8:28
[SPEAKER_01]: But you don't mean in principle, you're a placeable by, I was not as opposed to the work where our brains are going to be.
8:35
[SPEAKER_00]: I would maintain still that it's irreplaceable by principle.
8:40
[SPEAKER_00]: What do with the technology that we have?
8:44
[SPEAKER_00]: It's always a model of what we have.
8:47
[SPEAKER_00]: It's a representation of the reality, a representation of our capabilities, and the representation the map is never the territory.
8:57
[SPEAKER_00]: And there is more to the territory, to reality, to human capabilities than what can be represented.
9:05
[SPEAKER_00]: So I think that it's a principle question here.
9:08
[SPEAKER_00]: He's not just, of course, we change the technology we can do a little bit more in one way or another.
9:13
[SPEAKER_00]: But the principle remains the same.
9:16
[SPEAKER_01]: I wanted to give you what I think their pushback is going to be.
9:20
[SPEAKER_01]: I think I'm broadly simple about it to review, but here's the technologist, right?
9:25
[SPEAKER_01]: Here's the, here's Sam Altman or somebody like that.
9:28
[SPEAKER_01]: Look, it brings a computer.
9:30
[SPEAKER_01]: It's engaged in calculations, and the mind is some software, and it breaks the hardware at the mind as a software, so to speak, or something along those lines, then we can create, let's say, functionally identical systems of different kinds of hardware and different kinds of software, that can pull off the same kind of thing as humans.
9:48
[SPEAKER_00]: Yes, I assume I believe very much that Sam Altman or someone like that would say that, the brain is where and whatever is software, but that actually is not only one who has said that, already since plateau, philosophers and technologies have been comparing the human mind to the most advanced technology that we had.
10:10
[SPEAKER_00]: We have been comparing it to hydraulic machines, we
10:14
[SPEAKER_00]: we're compared it to computers and now we're compared it to artificial intelligence.
10:20
[SPEAKER_00]: So we take now can you say more about that?
10:24
[SPEAKER_01]: Can you say a little bit about how we've out historically we've done that?
10:27
[SPEAKER_01]: We said, oh, this new technology, whether it's hydrology or the clock, whatever it is really just how the brain works or how the mind works, can you just tell us?
10:34
[SPEAKER_00]: I'm not really sure if I can get the correct examples now, I wouldn't mind it, but in in the times in which
10:44
[SPEAKER_00]: That was the most advanced, the most complex type of technology that we could do.
10:49
[SPEAKER_00]: And therefore we start comparing our brain to that.
10:53
[SPEAKER_00]: And even recently we were comparing our brain to computers, not to artificial intelligence, but to computers itself.
11:00
[SPEAKER_00]: We still use that one.
11:01
[SPEAKER_00]: Yeah.
11:02
[SPEAKER_00]: But the point is that it is not.
11:04
[SPEAKER_00]: The technology that we build again is possible representation of our brains, of our capabilities, of our world, but it's never the full thing.
11:17
[SPEAKER_00]: And we can compare it to clocks, or twider all of machines, or to steam machines, whatever else it has been done in the past.
11:26
[SPEAKER_00]: Now it sounds ridiculous, but at that time that was the most
11:33
[SPEAKER_00]: we still don't understand well our brain's work and there is a lot there that it still needs to be identified and it's a what they are that we don't really are able to place exactly on how the machines work and we try to dismiss those parts because that doesn't fit our picture.
11:51
[SPEAKER_00]: But at the end of the day, a mild man is not saying anything different than whoever was comparing brains to clocks or to hydraulic machines.
12:01
[SPEAKER_01]: Yeah, yeah, I know friend who is better known as a philosopher and talking to scientists named Eric Manelbaum who has a great paper or he says something like no or our minds don't work like computer programs do aside from neurons fire, you know, there are patterns of neural fires, there's also neurotransmitters, where's that, where does that show up in in computer science or data science or AI, there are hormones.
12:27
[SPEAKER_01]: right, there's estrogen and testosterone slushin around in the air.
12:30
[SPEAKER_01]: And so the idea that something like our minds would just as intelligent as we are now, if you like suck the all the neurotransmitters and all the hormones out of our bodies, just seems it's an experiment.
12:42
[SPEAKER_01]: I don't think it's going to
12:47
[SPEAKER_01]: take all that stuff out of our brains, out of our bodies, and our minds will work exactly as they do now.
12:52
[SPEAKER_01]: Seems crazy.
12:53
[SPEAKER_01]: And they don't have any of these analogs.
12:56
[SPEAKER_00]: And I think that's the right way to describe it.
12:59
[SPEAKER_00]: And there is a lot that we know about our brains, that we know that it's not representable with the current technology.
13:07
[SPEAKER_00]: And there is a lot, we don't know about our brains, and that is even more difficult to connect to.
13:13
[SPEAKER_00]: It is, in a sense, a lot, we always are able to represent more than before.
13:19
[SPEAKER_00]: The system represents better than a calculator and much broader in the type of problems that it handles.
13:26
[SPEAKER_00]: But why is very bad at doing what a calculator does, at least with the
13:30
[SPEAKER_00]: Gen AI, that we have seen nowadays, I still respect better my calculator, then jet-chipit calculator, two plus two, even.
13:38
[SPEAKER_00]: But it's still, again, there is a lot in our brains that we don't, not even understand the function.
13:55
[SPEAKER_01]: There's a bunch of articles that have been published lately about how different companies are trying to do things to drain the intelligence out of current experts.
14:03
[SPEAKER_01]: So there's a company for what it's called, but there's a company that's hiring PhDs and various areas and trying to download their brains, get all their insights, get all their information, their knowledge, so that they can model it.
14:19
[SPEAKER_01]: meta is doing this now too.
14:21
[SPEAKER_01]: I just read something like this.
14:22
[SPEAKER_01]: I think it's in a called mistreat.com.
14:24
[SPEAKER_01]: There was an article that I read yesterday about meta saying they're going to track everything that their employees do.
14:33
[SPEAKER_01]: Not for the purposes of surveilling them for the sake of promotion and firing and that sort of thing, but so that they can take that data
14:48
[SPEAKER_01]: one reaction is repulsion, but maybe that's not the right attitude given what you've said about the irreplaceability of human intelligence.
14:56
[SPEAKER_01]: The map is not the terrain, I realize that metaphor.
14:59
[SPEAKER_01]: Maybe we should laugh at their sort of, I don't know, this is, we see what you're trying to do when it's ridiculous.
15:05
[SPEAKER_01]: So we don't need to get morally outraged by it because it's such a improbable.
15:08
[SPEAKER_00]: Well, I don't think it's probably just wrong.
15:10
[SPEAKER_00]: Data is the result of whatever our brain's hard to link.
15:14
[SPEAKER_00]: Data is output, expecting that looking at output, we are able to reconstruct all what comes behind.
15:23
[SPEAKER_00]: It is proven even with any type of technology that's not the way to go.
15:29
[SPEAKER_00]: So this insistent that we have nowadays,
15:33
[SPEAKER_00]: on using data as the let's say ground matter to develop AI is by definition already taking a wrong turn somewhere because the data is what is the result of the process of thinking of observing or feeling of whatever.
15:50
[SPEAKER_00]: It's not the process itself and we are very far away to understand how those processes work and even more what is the
15:59
[SPEAKER_00]: Let's say the starting point for all of that.
16:02
[SPEAKER_00]: Another issue with data is that data is at end of the day measurement.
16:08
[SPEAKER_00]: It's something that we extract from whatever from the terrain, from the territory,
16:14
[SPEAKER_00]: based on the tools that we have.
16:18
[SPEAKER_00]: If I cannot extract the smell, I will never have data about the smell of that environment.
16:24
[SPEAKER_00]: If I cannot have sensors for temperature, I will not have the temperature.
16:29
[SPEAKER_00]: So it depends on our sensors on the capabilities that are there, and human sensors are quite different from the mechanical sensors that we are building.
16:39
[SPEAKER_00]: But also, even more important,
16:44
[SPEAKER_00]: Data doesn't exist as a natural resource.
16:48
[SPEAKER_00]: Data is constructed and is constructed based on what we think is relevant to keep.
16:54
[SPEAKER_00]: If meta is now looking at tracing all the behavior of their employees just to be able to create a dump of the human mind or whatever it is, again they are making a decision on what kind of things they are going to do.
17:13
[SPEAKER_00]: They are not tracking other ones somewhere else, other types of people in other situations outside of the office.
17:20
[SPEAKER_00]: So all of those are choice and the choice is that we make on how we are measuring our environment and what are we measuring, where we are measuring, that all bias the data by definition.
17:34
[SPEAKER_00]: Data is never the full picture.
17:37
[SPEAKER_01]: So a theme weren't in through what you're saying right now, which is something like to put it a bit poetically or something like that, that everything they're doing, everything they're dealing with is just an echo of intelligence and not intelligence.
17:50
[SPEAKER_01]: The data as you put it is the results of.
17:53
[SPEAKER_01]: intelligence doing its thing, thinking, planning, dealing, whatever.
17:57
[SPEAKER_01]: And then, as you said earlier, whatever they wind up with, they take that data, they make a model, but the model is not, or the math is not the terrain.
18:04
[SPEAKER_01]: And the whole thing is just reminding me, you've worked up Plato, so now we've got Plato on the mind.
18:08
[SPEAKER_01]: And Plato is whole thing, and the Republic is,
18:12
[SPEAKER_01]: He's indicated with the people, it's an allegory that people are in the cave, and there's a fire in front of them.
18:17
[SPEAKER_01]: Sorry, there's a fire behind the people, and the leaders are holding puppets in front of the shot.
18:22
[SPEAKER_01]: And front of the fires was cast the shadow on the people, sorry, cast the shadow on the wall before the people.
18:27
[SPEAKER_01]: And they must take these shadows for the real thing.
18:31
[SPEAKER_01]: but they're really just like echoes of the puppets, they're not.
18:34
[SPEAKER_01]: And as you look, you're saying it to some of the time situation here.
18:37
[SPEAKER_01]: Where the technologists are the ones chained to the wall watching the shadows thinking that it's the real thing.
18:43
[SPEAKER_01]: And then they try to measure it.
18:45
[SPEAKER_01]: They try to capture it.
18:46
[SPEAKER_01]: They try to recreate it.
18:47
[SPEAKER_01]: But it's dooms of failure because they're dealing with.
18:50
[SPEAKER_01]: Yes.
18:50
[SPEAKER_00]: I think that's a very good one.
18:52
[SPEAKER_00]: And actually, we are all the ones watching the puppets.
18:55
[SPEAKER_00]: And someone is deciding what puppets are put in front of us.
19:00
[SPEAKER_00]: Someone is deciding what those puppets are doing, where they are, and so on.
19:04
[SPEAKER_00]: So there is, again, another thing that I try to say, which is AI doesn't happen to.
19:10
[SPEAKER_00]: as we people make AI happen and someone is making the choice, someone is making the option, taking the priorities on what we are looking at or not.
19:20
[SPEAKER_00]: And actually this method for us, I wish I had thought about that one before I wrote a book because I think it's a very
19:29
[SPEAKER_01]: The other thing that occurs to me as you talk about the stuff is something like, I wanted to paraphrase what you were saying or translate what you were saying and to talk about something like breaths or multi, and you use the phrase multi dimensionality, and tell you the multi-dimensional.
19:43
[SPEAKER_01]: I actually didn't, maybe that's not even the right word either, because
19:47
[SPEAKER_01]: When we ask, okay, so we're to the dimensions.
19:49
[SPEAKER_01]: And then you start saying things like, there's mathematical intelligence, there's the emotional intelligence, there's social intelligence, and let me get into this sort of category listing thing.
19:57
[SPEAKER_01]: And then we're going to wind up with five, six, seven categories wherever it is.
20:01
[SPEAKER_01]: But I actually think what you're a point is not that there are all these categories where AI can't do anything.
20:06
[SPEAKER_01]: It's that the sort of inarticulated and maybe inarticula bowl,
20:11
[SPEAKER_01]: richness, a fine detail was in each one of those categories.
20:15
[SPEAKER_01]: So, you can say social tells, but the amount of stuff that's in there, or if you take the worker at meta, who exhibits some behavior that is then measured, the quantity of stuff that the fine green detail that went into, what made that person do that, is lost.
20:30
[SPEAKER_01]: It can't it's
20:32
[SPEAKER_01]: seemingly uncapturable.
20:34
[SPEAKER_01]: And so it's not just if you like the number of dimensions to intelligence, but it's the richness of each dimension that is probably a problem.
20:42
[SPEAKER_00]: And it's not only, so you're completely right in terms of the categories.
20:46
[SPEAKER_00]: Once we start to categorize, we start finding out that there is more to categorize than what you can put, but it is a way to try to explain that there is more to it than just the cognition.
20:57
[SPEAKER_00]: Yeah, but you are completely right.
20:58
[SPEAKER_00]: Each one of them is extremely complex and it's rich in it breath in ways that we are only now trying to identify, but it's not only that, it's also the combination of those categories together, the combination of those types of intelligence together.
21:14
[SPEAKER_00]: They are not standing by, it's not isolated silos, here is my social intelligence, here is my
21:22
[SPEAKER_00]: all the interactions between all those things that create even more complex where think to describe or even to aim to fully categorize, fully represent and fully models.
21:35
[SPEAKER_01]: But let's pivot a little bit.
21:37
[SPEAKER_01]: We talked a lot about intelligence.
21:39
[SPEAKER_01]: I wanted to go on to another paradox.
21:40
[SPEAKER_01]: I mean, that was only one paradox.
21:41
[SPEAKER_01]: So I was just, there's a lot, just speaking of richness.
21:44
[SPEAKER_01]: There's a, there must be a lot on your book because there's a lot on one paradox and there's more to say.
21:48
[SPEAKER_01]: Okay, but let's go to a different one.
21:51
[SPEAKER_01]: This is, I think, chapter four.
21:53
[SPEAKER_01]: in the justice paradox.
21:56
[SPEAKER_01]: In a lot of AI circles, especially AI ethics circles, we talk a lot about fairness and justice.
22:02
[SPEAKER_01]: That language quickly gets translated to talk about bias and bias and education.
22:07
[SPEAKER_01]: And the paradox that you'll highlight in this chapter is less bias is not always more justice.
22:14
[SPEAKER_00]: So again, bias, the way we are working on it in AI is a lot to do with the bias on the data.
22:22
[SPEAKER_00]: Like I said, the data is already by itself a product of choice that we're made before.
22:28
[SPEAKER_00]: So if we are trying to unbiased or to solve the bias on the data,
22:33
[SPEAKER_00]: we are definitely not yet addressing the reasons why that data was different in the first place.
22:41
[SPEAKER_00]: Why are women not the same type of salaries as men or why are less women working in a either men?
22:48
[SPEAKER_00]: It is not just about trying to fix it at the level of, okay, I don't know, I'm going to start hiring more women or give higher salaries to women.
22:57
[SPEAKER_00]: but the justice is starch much before that.
23:00
[SPEAKER_00]: Where does it start?
23:02
[SPEAKER_00]: What kind of conditions did men and women head to that led them to the place in which more men study STEM than women or whatever other type of is just to give a simple one that people understand quickly?
23:17
[SPEAKER_00]: So the bias is again the result of a very long process and this is even more than individual life process because it's something which transfers from generations to generations in which by addressing the symptoms the bias that we see on the data we are definitely not yet addressing the cause of those symptoms and just this is there.
23:42
[SPEAKER_01]: Get a little bit of you slightly different
23:45
[SPEAKER_01]: I'm in an extreme way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way, way,
24:11
[SPEAKER_01]: All this focus on we're going to make, I'm going to blow up the point of what sort of big one.
24:14
[SPEAKER_01]: All this focus on AI is going to bring equality.
24:18
[SPEAKER_01]: It's going to democratize access to knowledge and everyone's going to have access to their own AI agent and it's going to spread the power.
24:26
[SPEAKER_01]: And because we're going to make sure that their fair AI systems are going to correct for the biases and society by making sure we have unbiased AI, blah, blah, blah.
24:34
[SPEAKER_01]: And that sort of,
24:37
[SPEAKER_01]: completely overestimates something like the power of AI relative to larger social political movement.
24:46
[SPEAKER_00]: In a sense, it's also a kind of, I don't know, you call it, it's a distraction.
24:50
[SPEAKER_00]: Because as long as we are discussed, the bias in AI and the complexity of addressing bias in AI, we are not addressing the real problems.
24:59
[SPEAKER_00]: It's great to do that AI is not, it's not bias, that AI is going to democratize access to knowledge and to whatever, but if you're not even ever obsessed with electricity, forget it.
25:12
[SPEAKER_00]: And if we're fixing the AI, we are not really taking the opportunities and the investment that we need to make sure that clean water and electricity reaches everyone.
25:23
[SPEAKER_00]: And investment, that's again, one of the level of government all over the world at the moment, is that they are over-focused on the investments on AI, on the deciding their own AI strategies, if the building, they, for their own, is right in AI and so on, but that is all investments that are not going somewhere else.
25:45
[SPEAKER_00]: What, and those choice are political choice, are not technological choice, are political choice.
25:51
[SPEAKER_00]: As long as you are wheeled your great data center to have your own sovereign AI and you are investing there what you are not investing in clean water for the population, you are making a choice and those choice needs to be pointed out much more than just looking okay good great this guy, this country is having some sovereign AI which is even wheeled with data great but it's not enough.
26:16
[SPEAKER_01]: This recently been, maybe the past from now until the past year, Ash, maybe more, has been debate within something like the AI ethics community or the AI community.
26:26
[SPEAKER_01]: You've got people who are screaming about the existential risks that AI is going to take over, let's put a pause on the research.
26:35
[SPEAKER_01]: Humans are probably doomed.
26:36
[SPEAKER_01]: Everyone's talking about stupid p-dume, and the probability that I'm now saying is for you.
26:40
[SPEAKER_01]: But for listeners, p-dume is something like the probability that humans are screwed because AI is going to take over
26:46
[SPEAKER_01]: And then you have these other people who are a bit more, they would probably describe themselves as more grounded.
26:52
[SPEAKER_01]: And they might say something like, or they do say things like, could you stop talking about the existential risks today?
26:59
[SPEAKER_01]: I not only is it bullshit because AI is not going to do that kind of thing, relax yourselves.
27:04
[SPEAKER_01]: But more importantly, it's a distraction.
27:06
[SPEAKER_01]: from the very real issues that we face today, not tomorrow, or some day that's 1,000 years ago, 1,000 years from now.
27:14
[SPEAKER_01]: We have faced real problems with privacy in AI, and bias in AI, and lack of transparency in AI, and lack of accountability in AI.
27:24
[SPEAKER_01]: And it's just to me like you're coming into this debate between these existentialist people, and these sort of quotidian everyday AI ethics people.
27:32
[SPEAKER_01]: And you're thinking then, no, you're creating the distraction, you would think it's all about fairness and AI or unbiased AI and privacy and AI and transparency and AI.
27:43
[SPEAKER_01]: But that's a distraction from the actual larger social political issues of fairness and privacy and transparency.
27:50
[SPEAKER_01]: So, you're the one who's distracting us from.
27:53
[SPEAKER_01]: What we should really be paying attention to, which is the larger social relationships.
27:56
[SPEAKER_00]: And both sides are a distraction, because both sides, and more that we have, and I fully acknowledge and recognize these two sides of to come to that you describe.
28:06
[SPEAKER_00]: Yes.
28:06
[SPEAKER_00]: They are a little further and they are fighting each other in any conference or in any podcast going around.
28:12
[SPEAKER_00]: But in it, the Europe's salvation is exactly to the point.
28:16
[SPEAKER_00]: We are distracting ourselves with this discussion and this, yes, no, yes, no, sides, while the world is burning around us.
28:25
[SPEAKER_00]: Literally, it's burning too, but look at what, and for the people who are really suffering from what is happening in the world at this moment, either from the wars or for displacement or from climate change or from whatever else,
28:39
[SPEAKER_00]: whether AI is taking over existentially in a couple of years, or we are concerned about the privacy of the data that we are using.
28:48
[SPEAKER_00]: None of that is a point.
28:50
[SPEAKER_00]: It's all wood.
28:51
[SPEAKER_00]: We do need to ground ourselves in what's happening and
28:55
[SPEAKER_00]: Hey, it can be definitely a support for building better societies, creating a more democratic access to all kinds of resources.
29:05
[SPEAKER_00]: And at the same time, it can be exactly the opposite and creating even more.
29:10
[SPEAKER_00]: The version and more and equal distribution of resources, both of them are okay, but at the end of the day is one of the many types of technologies of many types of issues that we have.
29:24
[SPEAKER_00]: And again, the issue is not the territory, let's go back and see what's happening and look outside.
29:31
[SPEAKER_01]: Yeah, it's totally unborn.
29:32
[SPEAKER_01]: That's why when everybody says, oh, we're going to use AI to make the world a better place.
29:35
[SPEAKER_01]: I'm like, no, you're, no, you're not.
29:37
[SPEAKER_01]: You've got to fix.
29:38
[SPEAKER_01]: There's like, all these broken social political systems that are in shatters if they were ever put together whole.
29:43
[SPEAKER_01]: And you're trying to polish this little piece over here, but it's surrounded by, yeah, anyway, so totally agree with you.
29:50
[SPEAKER_01]: And then the other thing that I find frustrating, though, is that for those people who do get to this, they say,
29:58
[SPEAKER_01]: It's not AI as such as the problem.
30:00
[SPEAKER_01]: It's capital is not the problem.
30:02
[SPEAKER_01]: But that's the entirety of their critique.
30:04
[SPEAKER_01]: It's capitalism bad.
30:05
[SPEAKER_01]: Okay, that's that particularly helpful.
30:07
[SPEAKER_01]: Okay, so you want to overstar capitalism, what would you like exactly next?
30:11
[SPEAKER_01]: What does that look like?
30:12
[SPEAKER_01]: And so, you, they fall into these almost flat tunes.
30:15
[SPEAKER_01]: Capital must bad, greed is bad, massive wealth and equality is bad.
30:19
[SPEAKER_01]: And then there's nothing like a gesture towards a legitimate solution.
30:23
[SPEAKER_01]: There's very little by way of action.
30:25
[SPEAKER_01]: and so people get to feel good about themselves by saying capitals and bad, but that's that actually addressing social political issues.
30:31
[SPEAKER_01]: You actually have to do some stuff.
30:32
[SPEAKER_00]: I do a completely right and I fully agree with you, the observations are made.
30:36
[SPEAKER_00]: In a sense, we are trying to find, let's say, the solution to all our problems.
30:42
[SPEAKER_00]: And that's all it, it either is capital news, more is not capitalism, it's AI or whatever.
30:47
[SPEAKER_00]: of this feeling that somehow we should find this thing that is going to solve all our problems.
30:54
[SPEAKER_00]: In religion they do the same, but the end of the day, we'd not even start to understand what are our problems.
31:03
[SPEAKER_00]: We definitely don't agree what are our problems.
31:05
[SPEAKER_00]: But also the idea that there will be one way to address everything, it's also extremely nice because we do need to have much broader understanding.
31:15
[SPEAKER_00]: And it starts with work that I'm doing recently, and you talked about the card a couple of minutes ago, all our Western knowledge has been extremely grounded on the Cartesian view of the world.
31:33
[SPEAKER_00]: I think, therefore, I have a little extremely individualistic view of intelligence.
31:39
[SPEAKER_00]: There is one thing that it's intelligent and that thing exists by the fact of being intelligent.
31:45
[SPEAKER_00]: If you start looking at African or Asian or even native cultures,
31:51
[SPEAKER_00]: Philosophies is very much more, and I'm just quoting Ubuntu, which is one of the Philosophies in Africa, which doesn't say at all, I think, therefore, I am, but it says, I am because we are, that is the ground of it.
32:06
[SPEAKER_00]: And so much more collective and by necessity, joint type of understanding than what we have with the Cartesian view, that it's this individual intelligence that is going to solve it all.
32:21
[SPEAKER_00]: So we do need start looking at many different, first understanding what our problems and also understanding that we need a toolbox with much more than just different sizes of screwdriver.
32:35
[SPEAKER_01]: Yeah, I might disagree with you about what they Cartman, but I think, before I don't think he was engaged in the epistemological project of trying to get, is it can just figure out something that I definitely know for sure without a doubt as opposed to articulating what it is to be intelligence or the primary city of the individual.
32:51
[SPEAKER_01]: But that's how to get your point that there's a way of thinking about intelligence as manifested in discrete individual entities, each individual human beings, as well as being some collective thing, which brings us
33:03
[SPEAKER_01]: to chapter seven of your book.
33:05
[SPEAKER_01]: I think the super intelligence paradox and the paradoxes and this is just really asking you to expound upon the bulldozing and surrounding thoughts.
33:14
[SPEAKER_01]: But the paradoxes you wrote here is the more we chase AGI artificial general intelligence, we discover that true super intelligence lies in human cooperation.
33:23
[SPEAKER_00]: OK, so that's one.
33:25
[SPEAKER_00]: What I mean by it.
33:26
[SPEAKER_00]: Let's look at the definition that most people are using or understanding what is super intelligence, which is a system that is more intelligence than person, that people.
33:36
[SPEAKER_00]: So the mind that is bigger than our mind.
33:39
[SPEAKER_00]: Sure.
33:39
[SPEAKER_00]: The two of us together and we have the most advantage already now.
33:43
[SPEAKER_00]: The two of us together are more intelligent than each of us.
33:48
[SPEAKER_00]: So if we take the idea of collective intelligence as it really means, so the combination of our capability to make something that it's more than each one of us, collective intelligence meets the definition of superintelligence.
34:03
[SPEAKER_00]: And maybe we need to focus much, and this is a kind of work game, and we can discuss the dinner that I'm trying to make here, but I do believe that this pursuit of some external intelligence was the idea of building a God that it's going to solve all our problems.
34:23
[SPEAKER_00]: is again a distraction from the opportunity that we would have to really try to work together and together addressing those problems and those issues that are affecting all of us.
34:36
[SPEAKER_00]: So I'm not, again, it is a gimmick.
34:39
[SPEAKER_00]: But it is really trying to focus on the idea of intelligence and what we do need is much more than just one way, one size fits all type of intelligence, something which is outside of us.
34:55
[SPEAKER_00]: And it is going to do it all for us.
34:57
[SPEAKER_00]: It's an extremely lazy approach.
34:59
[SPEAKER_00]: Put it on this quote that we are going to build and then we can expect it either to solve our problems or we blame the God to not doing it.
35:06
[SPEAKER_00]: And that is distracting us from the communities to really understand each other and look together.
35:16
[SPEAKER_01]: So let me try to put it in a way that isn't, isn't gimmicky.
35:20
[SPEAKER_01]: I don't think, I don't think, I know what you mean when you say it's gimmicky.
35:22
[SPEAKER_01]: But there's a way that it's just not.
35:23
[SPEAKER_01]: Because here's how it goes on the AI side.
35:25
[SPEAKER_01]: If you're a semi-alternant type, we're going to create an AI, and it's going to endlessly aerate on itself, becoming more and more intelligent, until you have
35:36
[SPEAKER_01]: artificial super intelligence of this sort of insanely intelligent and any way more intelligent than any human being that's ever existed, look at it glowing, and that it will rule us all, or solve all of our problems, or cyber-to-calisol or cyber-to-calisol or cyber-to-calisol, it's worth depending on your temperament.
35:51
[SPEAKER_01]: But I think your point is the wrong benchmark here for super intelligence is not an individual human being, even if it's like an Einstein or something like that.
35:59
[SPEAKER_01]: It's really good physics.
36:00
[SPEAKER_01]: I'm talking about, I don't know how multi-dimensional
36:04
[SPEAKER_01]: Your points, I'm like, look, the human right, so the human being as a species has created spaceships and commercial airlines and educational systems and political systems and philosophy and art and blah blah blah and so the relevant benchmark is something like the set of capacities instantiated by
36:29
[SPEAKER_01]: the set of human beings, something like that, the entire set of human beings, including past, present, and future.
36:35
[SPEAKER_01]: And compared to that, you're not gonna get some individual single AI that has the kinds of capabilities that the human species has as a whole.
36:47
[SPEAKER_01]: That's okay, that's what I'm talking about.
36:49
[SPEAKER_00]: Also, like you said, if we are trying to create this super intelligence by wielding systems, that it iterate on themselves.
36:58
[SPEAKER_00]: And we are to get out what we put in.
37:02
[SPEAKER_00]: And what we put in is this data that we are just discussing before, which is just a reflection or a marginal understanding of reality.
37:11
[SPEAKER_00]: So, we are going to create something which will be able to solve some type of issues, but we'll more and more redefine what really matters, because only what is inside the data that the machine is using becomes what is relevant.
37:28
[SPEAKER_00]: Not all ourselves much, poor in terms of capabilities and understanding, because we are going to be limited to whatever that machine is going to address.
37:42
[SPEAKER_01]: a little race of them.
37:43
[SPEAKER_01]: I know the outlines of your spot, but I'll make sure you get the specifics.
37:46
[SPEAKER_01]: One of the things that's been gone around lately in the AI space is oh yeah, actually Virginia's right, you know, if they don't name you, they're sometimes they don't sure.
37:56
[SPEAKER_01]: But right, it's a communal thing.
37:58
[SPEAKER_01]: And so what we need are community of agents, communities of AI is talking to each other and learning from each other.
38:05
[SPEAKER_01]: So we need to mimic echo, whatever is the way that
38:11
[SPEAKER_01]: the human, the human, the human species that we've evolved by them working together.
38:16
[SPEAKER_01]: But we do that at the AI level, so you get like whatever, a lot of bought, mall, buck, whatever the health.
38:20
[SPEAKER_01]: And so that's how we're going to get the kind of super intelligence that we're looking for.
38:24
[SPEAKER_01]: It's through AI talking to lots of other thousands, tens of thousands of other AI.
38:29
[SPEAKER_01]: So we're
38:30
[SPEAKER_00]: that we're in for for in your point to what I'm doing so we're in the right I'm putting or trying to combine the capabilities of different AI systems is definitely or most likely better than just hoping that one thing is going to do it all but in this type of approach we are still ignoring a huge opportunity which is to incorporate people as well.
38:54
[SPEAKER_00]: We still have other types of intelligence, other types of capabilities, then whatever system we are wielding can do.
39:02
[SPEAKER_00]: Even if we put all those systems together, we are limited to what those systems can perceive or understand or define in their space, which is much
39:18
[SPEAKER_00]: much narrower and much less rich than what we can do.
39:23
[SPEAKER_00]: So probably the wife forward again is a much more collective type of intelligence in which we, yes, we can use all the definitely all this AI systems that can do, I don't know, diagnostic of cancer images much better than what we can.
39:39
[SPEAKER_00]: Calculators that calculate the square root much better than what we can do ourselves.
39:44
[SPEAKER_00]: But let's put it all together.
39:45
[SPEAKER_00]: Don't expect that the tools by themselves we're going to decide what to, what house to build.
39:55
[SPEAKER_01]: So yeah, there's no my last question then.
39:57
[SPEAKER_01]: I'm going to do a lot more to cover.
39:58
[SPEAKER_01]: That's why I think to go buy your box.
40:00
[SPEAKER_01]: But it's my question is something like, what's the upshot?
40:05
[SPEAKER_01]: We're not going to buy your lights and I'm on the same page.
40:07
[SPEAKER_01]: We're not going to build an AI that
40:15
[SPEAKER_01]: If we get a collection of humans and agents or AIs, talk to each other, we're still providing it to get there because of the richness and depth, et cetera, et cetera, of our intelligence, the multi-dimensionity out of everything we talked about.
40:26
[SPEAKER_01]: So what's this sort of suggestion that you have or do you have a suggestion for moving forward?
40:31
[SPEAKER_01]: Yeah, I'll stop there.
40:32
[SPEAKER_01]: Is there a suggestion about something maybe around reframing or how we should be taking what AI or how AI people should continue to develop?
40:41
[SPEAKER_00]: Again, I think that we need to stop demusing that there is one way forward.
40:46
[SPEAKER_00]: There are many ways forward.
40:47
[SPEAKER_00]: And the race in which we are all rushing each other down the same road.
40:53
[SPEAKER_00]: It's going to end wherever that road ends.
40:56
[SPEAKER_00]: And we need a much more explorative approach.
40:59
[SPEAKER_00]: And this exploration probably will many times hit a wall or whatever, but to expect that with one type of technology, one type of understanding intelligence or understanding the world, or understanding whatever, we are going to do it.
41:16
[SPEAKER_00]: It's extremely naive, and it's also extremely limited, limited to what we can do.
41:22
[SPEAKER_00]: So exploring different alternatives, seeing what
41:26
[SPEAKER_00]: What can be represented and why that should be represented and what for?
41:31
[SPEAKER_00]: And what type of options do we have to based our decisions on our own capabilities, or any capabilities of the machine or a combination of all these things?
41:41
[SPEAKER_00]: That's where we need.
41:42
[SPEAKER_00]: And we are just scratching this.
41:44
[SPEAKER_00]: We don't have a theory or any type of understanding of how to combine and addressing this evaluating this type of explorative approaches.
41:55
[SPEAKER_00]: So that's where we have to go and it's a lot to be done there.
41:58
[SPEAKER_00]: So there is a, I'm not at all negative for trying to tell something to say, but let's do it in a much broader and richer way.
42:10
[SPEAKER_01]: So then this is my last question, hence.
42:12
[SPEAKER_01]: That's the fact that I'll be answered, but maybe not.
42:15
[SPEAKER_01]: You want people to engage in all those exploring different AI different ways of making AI, blah, blah, blah.
42:22
[SPEAKER_01]: I guess I want to know, explore into what end or to what ends.
42:25
[SPEAKER_00]: Again, there, I think the end needs to be the purpose of improving human, and that's any technology.
42:34
[SPEAKER_00]: Is he a exists to somehow improve some type of our human capabilities, human existence, human know, well-being?
42:42
[SPEAKER_00]: or a not-on-human, a planetarium, or a global well-being and the capabilities.
42:48
[SPEAKER_00]: That needs to be the end.
42:49
[SPEAKER_00]: So the improvement of, and again, this improvement can be very different from different people, different situations, different times.
42:58
[SPEAKER_00]: But also there, we need to be more pluralistic than expecting one way forward than one measure by which everything can be compared together.
43:11
[SPEAKER_01]: Virginia, this has been great and it's really fun.
43:14
[SPEAKER_01]: So thank you so much for coming on.
43:16
[SPEAKER_00]: Thank you very much.
43:16
[SPEAKER_00]: What's fun?
43:19
[SPEAKER_01]: Thanks everyone for listening.
43:21
[SPEAKER_01]: I hope you enjoyed the episode.
43:22
[SPEAKER_01]: And I want to thank a few people who make this all possible.
43:25
[SPEAKER_01]: Delano Alexander, he's my producer.
43:26
[SPEAKER_01]: He does such an amazing job on everything that he touches.
43:29
[SPEAKER_01]: Garrett Lane is the sound engineer.
43:31
[SPEAKER_01]: That's honestly, he's just, he's too good for me.
43:33
[SPEAKER_01]: Authamia Jairam has joined us regionally.
43:35
[SPEAKER_01]: He's helping us find the best guests to bring on to this podcast.
43:39
[SPEAKER_01]: And Ashley Wasselenko helps us organize all this.
43:41
[SPEAKER_01]: Get people on the schedule, get people's bios, headshots, and really keeps it straight on the track.
43:46
[SPEAKER_01]: So thanks guys, really appreciate all the hard work you do.
43:59
[UNKNOWN]: Thank you.