Module 5 - Discussion

By Stephen Downes
Nov 12, 2021
Transcript of Module 5 - Discussion

Unedited audio transcription from Google Recorder

Okay, so here we are for the module five discussion. Oh, my should close that door there. You don't really need the local news in the background. That's perfect. That's we're doing this. It's just local news and still nobody else in me. I have to test up that this week has been a completely disconnected week.

So I would benefit from the discussion but mostly by listening, okay. I think it's been like that for most people and it also includes myself oddly enough. Although I really have basically done nothing this week but work on the course but I don't know if I'm really getting the outputs to show for that.

But how about you mark? How did this week go for you? Well it was normal. You know slightly connected slightly connected? Okay.

Less than an hour ago, I did. So I part of the this connection was the activity center. Hey yeah, not updated. So there was a couple days. I don't speak busy Steven Steven, you know, and then yesterday, I found two videos more than hour long. So yeah, I watched the long one.

Oh, yeah. Actually, I'm editing the text or the virtue. Oh, cool. Yes, I'll send that to you once I get it can close. It appears at least on YouTube it appears in the chat. Yeah. Right. And you can turn off the timestamp so that's a column of timestamps, which tell I figured out you can turn those off.

It was really, you know. But yeah, they even for a while. So then you get all these lines and during lines but there's no punctuation and, you know, follow up. Yeah. So I'm gonna clean it up into, you know, a readable text and I'll send it to you. Okay.

Now you do know that I'm also putting up transcripts right of these. So there's also that transcript. I don't know if that helps or not. I didn't know that YouTube was cream. Oh. Okay. It appears in the chat. I haven't looked for. Is there a zoom or you're posting the separate transcript from the phone?

Or? Yeah, it's a separate transcript from the phone. It's not taken from the YouTube. So okay, let's interesting because I didn't know YouTube was doing transcripts of them. Yeah. I think it's kind of amazing. Yeah. Yeah it appears in the chat but it's broken into timestamps. Yeah. Every second.

So you can see the the other ones, let me just quickly pop into the course here, and I'll show you when then, of course, showing you, I'm showing everybody who's in the course. So okay, let's just very quickly. I'll just share this screen here if I can remember, how to do it.

There we go. And there we go. All right. So if you're in the course you can see all the presentations simming the page updates. And again. Yeah, hasn't updated the last two. Let me just do that manually like I say I'm having update issues with my pages and it's been and since the beginning of the course and it has been incredibly annoying, so list the page.

And so I have to update everything manually and you can see there's lots of pages and it's just just something I haven't been able to get up on. So this is course videos, right? Yeah, which is really the list, of course presentation. So I'll just publish that page. There we go.

Now, there we go. So, if we look at the one on duty, which by the way, I was pretty pleased with. But so here's the presentation. You see the video here. The slides are here, you can just step through the slide. You can also listen to the alternative audio recording here and then down here are the, the links to the slides audio video and also to the transcript is right, and just jump to YouTube.

So yeah, exactly. Okay. Yeah, so I don't know if that's better or worse than the YouTube transcript. They're both being produced by the same company, which is Google, but, you know, one is by YouTube and the others from Google recorder and who knows? And did you know that friends?

No, haven't had time to do any of that at all. So it comes with paragraphs and capitalization, it comes with paragraphs and capitalization though. Yeah, it's not perfect by any stretch but you know, having something like that is better way. Better than nothing. Yeah, this is way better than the YouTube.

Okay. Obviously cleaning up that text and I'll send it to you on emergency. Oh yeah, absolutely. That would be so cool because you know what I'm planning to do? I've been creating transcripts of all the sessions that I've been doing and including our sessions with uni Sharita. And anyone else who joins us?

Hello. It's been basically you guys. And after the course, my plan is to clean up all the transcriptions, assemble them into a single large document. It'll be basically book length document and then that's the outcome of this course for me anyways. But you know I mean you guys are obviously valuable participants in the creation of that so because you know we have these interactive sessions and then I'm thinking about these interactive sessions when I'm doing those other videos.

So there's this thing, this bouncing back and forth of ideas has happening. So, you know, it's, it's like keep feeling guilty because, you know, just recording videos. Is it really anyone's idea of a good online course but it's been such a valuable activity for me that I've been spending a lot of time doing that.

So I hope you're getting something out of it. Well, yeah. And the thinking are a loud part. Yeah, no darling. I mean, yeah. So like you should Ford around that when you assemble, whatever the outpour is. Yeah, the order on that. And that's what we're doing on Mondays and Fridays.

We're looking out loud. Absolutely. Yeah.

And then there's the other the 11 finding this interesting to the mook aspect to it now because it's a connectivist moot. It's this distributed associative system, which emulates a lot of the properties of artificial intelligence. It isn't AI, of course. But but, you know, we're sort of modeling the way a lot of AI would work and then that's feeding back into it as well.

Which was my other big thing that I did this week. And I'm not sure if you saw it because it only showed up in the newsletter on Wednesday. It's like test page. But let me show you what I've done. So, here we go. Go back in. Well, I remember seeing that.

I don't know if I clicked on it. Yeah. No you probably didn't because I haven't highlighted it and it's you know test page isn't exactly and inviting thing but it always hides the top of my browser window when I do that. Okay, so here's the page that I created.

So remember before we were, we were doing graphs like this right now, imagine two rows of graphs, right? So on the left hand side, here, you'd have all the different codes. And then on the right hand side here, you'd have all the different values. Now the way we were doing it, we draw lines awkward cumbersome.

Really visually nice but awkward and cumbersome to do so here as the train goes by instead of drawing on, you just check a box. So if you check a box here, for example, that's like drawing a line between the eye tripoli code of ethics and care. And this is more like how people think of AI that are in a, i right, you have this matrix, these things here are vectors, you can have the vectors down and the vectors across and then they do a whole bunch of, you know, if you you look at the the courses and whatever on our artificial intelligence, they do a whole lot of manipulation of these matrices and a lot of AI constitutes matrix mathematics because really like, we've got one matrix in between each of these, each of these sets, right?

You know, like autonomy. Well, each of these values such as values in each of these codes. But in AI system, you might have two, three, four, whatever interconnected matrices. So you have the code at the one end and then couple layers of these things, all densely connected and then the value at the other end and the AI basically fills in all of those squares and they do that using matrix matrix mathematics for lack of a better term.

Do I understand all of the matrix mathematics that they do know because and it's I find this really interesting. You know, I took all the mathematics courses from kindergarten all the way through and into university first year university matrix functions never came up. So I never did addition of matrices multiplication and matrices etc.

Never did anything like that but there's this whole branch of mathematics that addresses it. So for me to understand that for most people will understand, you know, the actual details of artificial intelligence. You have to go back and cover that mathematics that your your foundational education, covering all the basics never covered.

So to me, that's an interesting thing. But anyhow, the idea is right, you click on these boxes. Now I want to do so a little helpful display so that if you click on the name of the code, you get and display of the code here. Or if you click on the name of the value, you get a display of the value.

So you know what? You connecting. But anyhow, once you do that, I also want to have ways of subdividing, these lists because it's a bit hard to work with. But anyhow, you put go down to the bottom. You submit your graph values. And so here we're the grass elements that I submitted.

And now that actually is now stored in the database and that actually works. But I don't have a mechanism for displaying that yet but that's that's like a short job to do assuming it actually went into the graph, the way it should. So that's the other thing I did, what I wanted that to be, was a task for this week, but it took me more than one day to write what you've seen so far.

And so I wasn't able to finish that task, it's been that kind of weak a good week because, you know, I think that was interesting but still. So, what do you think about that? Or do you have thoughts?

I like the way you were saying that. How you might work this? Even further out would be to have, you know, the particular code if you click on, you know, I don't know the code of psychology. Yeah. Right. And then you could read that and then you could click along the top in terms of exactly the meanings.

And that would be an interesting way. If I had it like that, you know, begin to click. We don't have it like that. However, I do have multiple monitors. Yes, what I'm gonna do. Yeah, right. And, and take a look at that. That's, that's should be funny. It probably will not make me as frustrated as drawing the line that I was really, you know.

However melture, what's his first name? My tires Matthias mentioned. Yeah, he did, you know, message me and I actually found not the demo but the actual thing. So I beginning to play a bit with that. Yeah, his. His actual thing is very powerful, it's in Java. So yeah, and and my hatred of Java is legendary love JavaScript, but Java is just, it's like using a battle ship to, you know, whatever go fishing.

So, but yeah, and he's been thinking about this for a very long time and he's out. He's also helped me out. I've been a communication with him on how to use this. And I think that it would, you know, what I want to do. And again, this is not too far away.

As far as actual work goes. Now, you know, once you've done all of those squares, you can display the result using his system. And it will look pretty cool because you'll see all the lines, right? Because it's all the same data. It's just two different ways representing that data and then putting that data.

But the other thing too, is it doing it? This way allows me to ask some interesting questions. For example, one question might be, you know, how would I come up with the things to actually connect? Because that's a relevant question, right? Like I've connected codes and values with this one, I'm connected other things with some of the other ones, right?

But codes and values. Well, how am I looking the codes? Are fairly obvious, you know, there's an actual document, but the values are kind of nebulous. How am I picking out, and naming these values and defining these values. What's that process? Right? Because that's something that's actually being, that's an input to our system, but it's just me making it up, right?

No, not really making it up, but so, but but you know what I mean. So, we we, you know, now the the way this was done in the field paper that I distributed to you guys in the newsletter was that they actually studied the ethical codes, identified, the values that were mentioned in the codes enlisted, those.

And that in fact is the same process that I used. And it makes sense because, you know, we're mapping values to codes. We might as well extract the values from the codes in order to create that matching. It's probably a bit easier that way, but you know, does the list of ethical codes comprehend the list of all possible values, you know and and again what terminology terminology should I use?

I mean these are interesting questions you know and suppose this is a different question suppose we were a machine or maybe designing machine that actually indicates whether we should connect a code with a value. What process should we use, you know, and think about how we do it. Well, I mean we're gonna have to, as you said, should we go back and look at the original, which is why so useful to have it sitting right there.

So, the codes says this, the value is this and we're doing some kind of mental matching task. But what exactly is that process? You know, if I was if I wanted to write a piece of software to do that matching instead of do it personally because it's going to take a long time.

How would I write it? I can't just do simple keyword matching. Yeah. But that presumes that everybody writing, all of these ethical codes, use the same vocabulary and mean the same words in the same way and what are the odds of that, you know, maybe within the discipline of ethics in AI?

Sure. But, you know, I got codes from y'all from legal professions, teaching journalism, psychology medicine, this odds seem pretty slim there. So there's, you know, there's quite a bit of interpretation happening for me to pick the labels in the first place and then quite a bit of interpretation happening in doing the associations.

So simple keyword mapping isn't really going to do the job and so one of the one of almost the fundamental question of AI certainly approaching it this way. Is what would do the job and that is a core question. And I'm not going to try to offer an answer to that here because facts like all of AI.

Yeah, I'm like she read his idea of having the text and I applied really great that you meant that you could have attacked. And then clicked the value category so that you don't have to scroll. Yeah, right that's okay. Yeah that's that's what I heard. That would be very useful and then I also have you know gave up mathematics after somehow yeah.

Oh did I? Yeah well it's it's not necessary to go farther in most. Yeah, and certainly I use now in my through it but what I had wasn't anyway, but I am interested in graphing and what is not going on in this world issue, I want to bring up is the weights or the definitions of the connection.

Yeah, that's another level and that, and that's factors. And, you know, it's one of the original parts of graphene. Is that the connections have you themselves? And I'm just wondering, first of all does AI is, that is AI, do that? And can we eat that short answer to both questions as yes?

AI definitely does that. Yeah, yeah. I'm that's a big part of what these algorithms do, is, adjust the weights of the connections. Like, in this graph it's just offered right? Which is nice but again, it's to it's too ham fisted really, you know, one thing might talk about a one code, might talk about a value a lot and really depend on it.

Another code might just sort of mention it in passing, you don't want to give the both a weight of one. And so part of the processing that I may, I does is do, is to adjust these weights and that, that is, you know, that's one of the things that's made possible in part by having multiple layers in between the codes and the values, right?

If you have multiple layers, multiple connected layers, you can really do some fine, tuning of those weights, really find tuning of those weights and get very good results. And and again, that's what AI does. What I've done is in my code, each one of these values, each one of these associations does have a weight.

So I have like the table, the first element, right? So code number 17 and then the second element, value, number five. And then the type of connection between them and I'm just assigning the name user, right? And then the weight and what I'm doing is each time. A user adds that connection.

I increment the weight by one so that way I can use the same graph with a large number of people. And if a lot of people pick that particular square, the number is higher and the third more, it's weighted more. Yeah. And then you represent that like in the grid thing I can represent it with smaller or bigger dots or in the the graphical representation.

I could represent it with thinner or fatter lines although I don't know how to do that with Matthias's system, I'm not sure if it's possible. It is because, you know, I mean, the the, the it's he is for the JavaScript, anyways. He uses an element called canvas in HTML as a very flexible piece of of HTML.

Most people don't even know what exists but it does exist and you can draw all kinds of things with canvas, so I'm sure I could widen or narrow. The way the the lines based on the weight so I was also wondering about some man picked weights. Yeah that's a another order of complexity that I'm sure.

Okay. What do you mean by semantic? Weights. Well, so, well, one of them is my, is the name using right here collecting user, which is not, well, here you're counting it. So, you're turning it automatical property. Yeah, but it's an average property, the name of the user and then, again, I'm picking, in fact, the first, I don't agree about crafts, right?

Yeah. But there could be semantic connections. Also you know there could be semantic representation of the rates you know could be more of an accountant. That's strong or leave. Well, it could also be comes from the church comes from law, right? How from ethical code again, those kinds, you know, so you can see it's in fact, but yeah.

Somehow, it relates AI, right? That theoretically, the computer can handle that. And then again, it's how how you make those smoke would almost make it. Even. So part of what I'm doing that addresses, this is naming the type of connection, right? So for take any two entities, we draw an association between them that association has a weight, but I also give it a name and I think the name maybe captures in part the semantic aspect of what you're thinking so I could have.

Now I'm just calling these user connections because they were created by users but you know if I'm using a different analytical approach, I could say, you know, I could do semantically like something is a part of something or belongs to, or is it type of and there's the whole range of relations that are defined in the semantic web in the, the world of RDF it's called resource description framework and they allow for all of these different types of connections to form, you know, semantic.

Meaning I don't go full-blown RDF. Although I could, I think, I think the RDF carrot, you know, a label set is limited because I think the relations are not simply semantical and they can be almost anything, right? And the way I've built grasshoppers, you can name the connection, whatever you want, which makes it a lot more flexible.

That's something you just don't see anywhere, but it's in there and it's off. Yeah. Well it's it exists and is very they're inside the software. It's just it just never comes up in any of the applications of it because it's hard to do. The other aspect is that these codes are sorry.

Let's pick the values. The values aren't just related to the codes. They could be related to other things. So like, for example this week, we've been looking at the the different ethical theories. Not finished. That by any stretch and I'm thinking maybe I'll delay the week, but that's a separate issue.

So we have, you know, we have duty-based theories and there's a much of a we have virtue base theories and there's a bunch of them so we could do the same exercise. We could look at these values and ask ourselves, well, what ethical theory or what ethical tradition are they associated with?

And now we've got something kind of interesting because we have you know via this two step linking a way of thinking about how the ethical codes are related to the ethical theory but not just directly related but via the values represented. And so let's kind of away of thinking about it, we could also directly relate them right?

But I'm not sure that would be as easy.

And there's one more thing I wanted to bring up all the support about connection. Yeah. Yeah. And GS is system. There's no undo button and so I know I introduced an error but because of where the labels. Yeah. Watch. Are you understand? Yeah, so so that just, you know, that brings up the issue of error and I introduced some areas because I could not undo.

Yeah, realizing my mistake, I could only add another connection. Yeah, and that, you know, the ones in space cleaned up. Yeah, yeah, then that that goes away but there are reasons reduced, but there's other things here. So yeah, there's always going to be ever. I agree. I'd rather the you know the the way I'd like it all to work is the ever stays over there in on your website.

And but no, there's always going to be ever one of the great things about this kind of approach. So is that the error almost factors out? It creates noise in the system, but the system can stand noise. So, let's take a good example. Suppose you created an ethical code by accident.

It's not really an ethical code, but you put in something that you thought you're putting in the right thing, but it went into ethical codes, it's not a code, but there it is in the database. You can't remove it. Well, okay, it's not like people aren't going to check the box related to that, ethical code because they're not going to find a relation between what you wrote and the value, right?

So already we're beginning to see it abstracted out a bit as well. Each of these things, right? Each value each ethical code, each ethical theory can also have a weight, just like the connections, the individual entities can have weights and over time, the weight of the one that you made as a mistake, would be very low but the other things which get mentioned a lot and used a lot and connected.

A lot those weights will be higher. So again, even though you made a mistake and even though you can't remove it from the system, it's and so even though it's noise in the data in the grand scheme of thing, it doesn't carry very much weight and it's mostly just ignored by the whole rest of the system.

Yeah, this is the, we just need more data approach. It's, you know, we just need more data. And yes, so often and you know, and there you just point out. Yeah. Why. That true statement. Yeah, and I'm tempted, you know like we we don't have thousands of people in the course but that's okay.

I'm tempted to just take, you know, one of these grids once it's done and has all the little displays popped out on Twitter and see if people just feeling that grid, they might, they might not, you know, you know, it's I have a limited following on Twitter, you know, I'm not not Twitter famous.

So, you know, I'm not gonna get hundreds of responses for sure, but probably get more than one or two. I think I know almost double your days which would double maintain its head. Yeah. So can I ask a question? Yeah, as I keep coming back to the entity that develops the develops.

I'm going to say code, but I really think the bias of the entity that tries to put this all together. Yeah, now most of the codes that are reading really are, you know, in some ways based on Western thought. Yeah, so what happens if a that entity that is going to program this or whatever sets it up?

Not to wait things as a West as Western coast, as a Western thought, a western epic. Yeah. So what happens then? Well that's im interesting question. I mean it depends on how you go about doing that. I think like how would you go about setting it up? Not to wait, things based on Western values, you know?

Now what a lot of people are talking about is well you just get opinions from different populations in the data. So the equivalent in our system would be I take my grid. Yeah right and I don't just show it to Western academics or there you know first year psychology students which is what we usually do in education.

I take that grid and I make it available specifically to people in different cultures and indeed if I wanted to be really careful when people feeling that grid, they would also record. Maybe what their language is religion culture background etc, which would feed in as well to the system, so that when I calculated the weights of these connections, I could instead of just taking a numerical average, I can take a representative average where each religion is waited equally or so I could do that and and that's the sort of thing that people do okay but isn't making it grid a Western approach.

Our definition and showing the grid in a lot of confidence that would be like, I don't know if the meaningless almost, okay. No. Ways of looking at that. But I mean, that's a great question, right? The first way is okay, maybe it's just an interface problem, right? Grids are a pretty square, pretty Western, maybe if you had something less with fewer pointy corners, the crafts sort of help with that, you know, or, you know, you could sort of represent things as gradients.

And, you know, you can imagine all kinds of different interfaces where pardon emojis instead of ants emojis and set of taxed, you know, just all kinds of things like that. Yeah, I'm just picturing in my mind, like, a completely analogue interface where people are just using their fingers on, on a touch screen, you know?

And that that might be more appropriate for some culture. I know idea what kind of interface is appropriate for what kind of culture. There wasn't an RC study years ago, that try to associate different colored pallets with different cultures. And I'm not sure what came of that. I'm not sure if you can.

But, but certainly, we can imagine that intuitively. And we know that different colors, represent, different things in different cultures, you know, like, the color of white, for example, means something very different nerd culture that it would say in Chinese culture. So, you know, the other part is, what if the whole idea of breaking things down into parts and associating them is.

It's self a Western idea. It is reductions. All right. And you know, and you know, you look at how I've structured this. Entire course here, right? And like, only an old English guy from an analytical tradition would take a subject like ethics and analytics, and break it all down, into the little pieces, the way I've done it.

All right. Who would do that other than someone like me and there's a point to that. I agree. There's a point to that. And, and then the question and it's not just an ethical question, but it is an ethical question, but it's also a methodological question. What do you do instead?

Yeah, all right. And I think that's a good question.

Where I hit up against it again because my culture is kind of different from the classic analytical Western philosopher. Researcher kind of mold is to distinguish between symbolic elements and sub-symbolic elements. A symbolic element is a word, right? I might include an emoji, but it's a representation system of the word, the title I've given to a value or to an ethical code or to whatever stands for something.

So basically we're manipulating symbol systems. I think words are blunt instruments completely unsuited to the task because you just capture too much with the word. Take a, take a pick any words, you want a cup, there's a cup right? Our understanding of a cup is much more nuanced than the word could possibly date.

And if we look at our our minds, how we have the concept cup in our brain? Yeah, we do have, you know, the physical symbol and the audio symbol cup that we can use, right? We can hear somebody. Say cup and we can say the word cup but we don't have the word cup in our brain anywhere, and we don't even have a physical image of a cup in our brain.

But what we have is several thousand neurons that go to whenever we see one of these and that prompts us to associate this with the word cup or it does for me, might not for you. The unfortunate attributes of dangerous languages expressibility. And so, and you picked the word cut, not me.

Yeah. But so I can say, wow, I hope that football player was wearing his cup when he took that hit. Yep. I don't mean one of these. I did not. Yeah, so yeah, I mean it's it's marketing. We're swimming around on the murky. Beat them together. Yeah. Trying to make meaning and this is, that's what we're doing.

I think. And, and discussing, how machines can make meaning, my skepticism has grown, as we've done this, but that's not going to stop people regarding well, so, shouldn't stop. You from trying doesn't stop me from trying. I'm believe me. I got more skepticism than you can shake a stick at I love metaphors.

Metaphors adds a whole other dimension to this. Yeah. So if we're trying to do a I and we're trying to avoid avoid bias, we're kind of stuck with the fact that any system of language any labeling system is going to have some kind of bias.

So for me what that says is, well, ultimately we should probably illum a eliminate that layer from AI but that's kind of hard. Irrigation was the word I use? I think it was one day whenever we were talking about pizza. Yeah, not a coincidence. Yeah, yeah. And so, and I somewhere along here, you can convince me that elimination is impossible.

So, that's why I reached for mitigation. And so, then we would have to, I would want actually make a person, I would want absolutely transparency, so open source. So that the code could be critique from multiple points of view. And then mitigation is, would be part of the, you know, ongoing discussion as because and the stuff evolves, as another issue with stuff, involves the words designed.

I mean, yes. So so this mitigation problem is going to increase as it evolves, right? The idea of mitigations but they're also grow and grow more complex as the AI grows and grows more complex. Yeah, it's going to be so problem. Yeah is it? But also too, I mean let's bring this back to education for a second especially since we're running to the end of our hour but a lot a lot of educational processes and a lot of educational theory is inherently based around the idea of labeling think about how many theories are tax on these of this.

And categorizations of that right there all based on labeling or alternatively think about the work that's going on right now to develop systems for recognizing and teaching skills and competencies. Where what they're trying to do is create this competency definition, taxonomy with the idea that everybody would use the same tax economy, it's a labeling thing, right?

What you're doing is taking all of this raw data, which is sub symbolic and very nuanced and out there in the world and partially in our own brains and trying to associate it with a bunch of words. And then later on, we'll try to use these words to say, evaluated a test or a performance or something like that.

So we've built in this structural bias into our system that maybe we could have avoided by skipping the part where we categorize everything. Maybe, of course, the problem is, how would we understand what we've done without language, to describe it and interpretive dance. Only goes so far. Yeah, exactly.

You know, at some point somebody's gonna have to sit down and say, well, when we see this motion, what the dancer means is that the moon is full and she is filled by the moon, we bring it back into language. But of course nobody listening to that would say oh the moon she's filled with the moon but she's way too small, the moon is huge.

So of course, now we're in metaphor land, one of the issues I run into all the time in higher education, the for some reason. Well, it's part of part of the system this name and reaching hello and so the one thing I wanted to bring up or bring up again is also beyond the cultural lens and you know that we've been discussing.

It's it's just been a really been struck with this last couple weeks in this class and in my regular college online class. And there's that work class, you know, use multiple ways. Okay. So all of this AI not only is it, you know, Western educated mostly men probably mostly white.

Well, all about the coding part is spreading around. Not only is it, you know, that particular point of view but also class comes into this. Yep. That this is all created and and embedded inside of capitalism. So it's all taking place inside the capitalism. It's all done by a particular class and it's being paid for by the upper portions of the class.

And in my opinion is being used again, the lower classes however you wanted to find. Yep. And that's that's an argument that can certainly be made and, you know, I mean we we might call it say critical class theory and it's analogous to critical race theory you know. And and you do see that kind of argument made by people like marks and angles 150 years ago very similar sorts of arguments very higher education what depends on where you are you know it really does depend on where you are.

I mean there's no shortage of class-based thought out there in the world. Unfortunately a lot of it is propagated by people who are in the upper class and so there's a certain skewed perspective of it, but it exists. But let me challenge you with this. Now that also is a label, all right?

And we you know, you know, I mean it's it's not for no reason that people like marks used basically what we're almost artificial words like bourgeoisie and proletariat to describe to describe what he's talking about because you know is just hard to express those concepts and the language that was in common use and even today's kind, you know, I mean we talk about the working class, right?

But we include in the working class, a lot of people who are aren't strictly speaking working, you know, if they're unemployed for example, what we mean by the working class is probably delineated by their circumstances by where they live, what size house. If they have a house or apartment, they live, what they're is.

And as you alluded to earlier, also their values, their beliefs set etc. And it would be interesting. I think to draw one of these maps that we've been drawing from values to different classes of people. And to be interesting to see, not just an overall graph but how people in their different classes drew that graph.

But again, how do you label it? You know, there are almost certainly values. As we've been using the term that exists in the working class that other classes might not even have a word for, you know, think about these and the word righteous and motorcycle clubs, totally different, meaning from the word righteous among people in the British peerage system, right?

Just totally different, you know. So there is, but the idea here, I don't know if it flies but the idea here is if you get enough input and enough of these graphs and you think of these words, these symbols just as placeholders, but actually not actually as representations of anything.

They're just the actual symbols people use and you get enough data. Then maybe you can get past the bias of labeling. Maybe see, I bring this up again. I'm the American here. Yeah. And this is a very important issue but for the reason spending the time, you know, these weeks is this is such an important issue in American politics because of American corporations influence, the people had built AI Google Facebook.

In the largest application. They are having really military and consequences in this culture or across these cultures in this territory. If we want to use that work and so that's why I bring these things up. It's because it's actually a pressing issue. That's not getting installed. It's not going to be solved by the next selection, which is in one year from now.

And these are the issues that should be discussed in newspaper every day and on the radio every day and they're not. Yeah. And just to give you an example of that in the thing that I did on duty, which I just did yesterday, I'm very proud of it. Maybe proud is the wrong word, but please with it.

There's a section in it on autonomy, which is what you need in order to make ethical decisions as a free agent. And the question is just how autonomous are we. In fact, you know, especially when you think that, you know, all of these labels, all these words that we use do have ability and bias and then all of the media that we're exposed to that, you know, underlined and enforce that by bias.

When we as a free agent, expressing an ethical opinion. Are we actually doing so autonomously? Or are we in a very serious way? Simply reflecting back. The bias that's been fed into us, and that's a core question. You know, that's that's one of these questions and that's why the ethics of artificial intelligence is so important because it also comes back to the ethics of complex systems like society, and you know, that we think we think about biased input for an AI, but we've got biased input for a society as well.

But how do you tag that bias? How do you put a finger on that bias when the very use of words themselves, incorporates bias? And that's a hard problem. Like the really super duper hard problem. And, and here's ours. I'm sorry. You know, George Lakeoff? Yeah so he's been working on this.

You know it's kind of his career. Yeah and I also think there's I don't speak French jar elu. Yeah propaganda. The formation of okay. Men smart formation attitudes and that's what you're talking about. Yep. And that also isn't with the problem. All right. I'm the only person. I know that what I left home at age.

17. I never owned a television. Mm-hmm, and I'd say often people. You shouldn't you should stop watching television. Okay. You should not yeah because what I hear what exactly? He said. I hear something coming out of their mouth and I know they heard about it. Yeah. Or I'm created and so I ask people is, is that your thought?

What why do you think that, you know, and but people don't like it when you do that. So I went for many many years with our channel without a television. Like from the day I left home to. I don't know, maybe 30 something like that. Forget how old I was when I finally got one.

But yeah I mean and he so you can look at people who watch television and see the influence and so you know there they're being I don't want to say programs because that's not the right word trained. It was maybe a better word. And the same way we train AIs, we television to train people and that speaks directly to the subject of learning and development, and courses, and schools.

And that, you know, people try to say, well you can't teach people using video but tell advertisers that. Yeah, anyhow I think we're gonna leave it here because it is one o'clock and I don't want to go over time but this was a an interesting discussion and and definitely related to this week's topic because you know this does tie back into these ethical theories but I'll continue to developing that grid and you know it's generalized like I can use it with some of the other data sets that we have and if you think of and this is useful different sets of labels we can use to describe whatever.

Let me know because we we can one of the things that my system has that pretty much no other system does is we can build these on the fly. We can build them from scratch, doesn't matter what they are. You know, if we want to build a category of items called foufas, we can and nothing prevents us from building them and then linking them into the entire entire grid graph network.

That is this course, I don't know how full should be useful, but that's why I'm asking for your input, you know, maybe there are things that, you know, sets of entities, that we should be thinking about that aren't here yet. Okay? So, till next time, thanks for for popping by.

I hope you're finding this interesting still and I know it's completely wrapping me up. So, and, and to the point where I'm expecting fall phone calls, you know, since they'll come back to her, anyhow, talk to you all later have good again. Thanks you too. So that was Mark, Wilson's reader, Ryan and Steven Downs in conversation for ethics analytics and the duty of care that was for the people listening on audio by everyone.

Force:yes