Session 2.6: Will AI DESTROY or SAVE education?

0:01 Hello everyone. Hello, welcome, welcome. Good to see you. Um, welcome and welcome on Instagram as well. Let me welcome everyone. Call your friends, tell them we are here. We are about to start and uh they should all come and join us. Um, we're having an amazing session today um on artificial intelligence. Call everyone to join us. Call everyone to join us. Everyone on Instagram, welcome. Welcome. And call everybody to join us. Call everyone to join us. We are about to start. Uh we're about to start. It's a

0:48 session on artificial intelligence. And um you're all welcome to join us. Call your friends. Good to see you on Instagram. Tell me where you're coming from and and um I will welcome you accordingly. Tell me insh welcome. Welcome. Welcome everyone. We are talking about artificial intelligence. It's amazing. Every week I get I I get to interact with people in the work that I'm doing and you realize how much other people are scared of in artificial intelligence or how much people are just getting into it and they don't want to do any

1:33 thinking anymore or people who are sitting there oblivious of the developments that are taking place and artificial intelligence is busy creeping into their space and they don't understand what's going onelo precious welcome I can see you thank you so much students from universities around the country around the continent and even those in Scotland Dublin Nigeria Nairobi Namibia you're welcome to join us h this is a session for you we're going to be talking about artificial intelligence and we will talk about um how it's affecting students at university today but I'm not I'm not alone today. So students from University of Ptoria and dosh I can see you um from

2:21 Pritoria I'm sure students from University of Pritoria are going to join us as they do every every weekend and because we are I'm having a conversation today h with professor Vosi Mariva I'll introduce him just now tell you a little bit about him so that you know h that we talking to a real expert in this field but let me continue to welcome you stock vermont h thank you very much for for for from from PE. Raymon, thank you for joining us. I see Gamu from um Pritoria in Montana. Thank you very much for joining us. Gamu and and and and people from Polan. I can see Makanu is there from Pouani joining us. Welcome. Welcome everyone. Welcome. I haven't seen my

3:07 fellow Zimbabweans. Where are you? It good to see you. H where are the fellow Zimbos? Uh Zimbos, please indicate. Indicate. I feel safe when you're around. You know what I mean? H fellow Nigas coming in as well. War warm regards from Hungary. Thank you very much N for joining me from Hungary. Where are you in Hungary? In Budapest. Interesting place Hungary is. I mean it's interesting when you come from a country that romanticizes communism and then you go to to to Hungary and you get people two different sides of the story. Very very interesting. Um, Vukosi, South African Institute of Physics. Thank you so much for joining joining me. Vosi, very very welcome. Put

3:53 a letter from Mafik. Call everyone at the University of Northwest that we here. I'm just looking at the numbers. I just want them to get to 100 and then we call in Professor Mariv. Um, tell me where you're coming from. Getting always joins us. Thank you so much for your support. I really appreciate it. When I'm in Nell Spritz in White River, I will definitely give you a shout out and we can have coffee. Okay, we'll definitely do that. I appreciate your support. My friends from Botswana, I can see the flag. Keep flying the flag, the blue flag. Emmanuel, thank you so much for joining me. My fellow Botswanas from Botswana, um, thank you so much. I can't see the comments on YouTube, but I'm sure you

4:39 are commenting. I can't see your comments, but I know you are commenting and and and I I I I appreciate all your comments. Let's see how the numbers are doing. Call your friends. Okay, let me get on and introduce Professor Vugosi Marvat. Professor Vugosi Marivate is of course my homeboy. Everyone from Harangua should be online and joining us. I can see Willow is joining us from Melbourne. Thank you very much for but please if you are somewhere in Ptoria those who are in in Montana and so on call everyone in Rangua Mopani Winterfeld so Shangu and tell them homeboy is in the house today he's going to be teaching the world and we as

5:26 Haranguans are very proud of him are very proud of what he's doing of course the the Marivatas have taken care of us in Haranguan so it's always wonderful to See Marat raising rising up and taking care of us in other ways. Leo from Botswana. Welcome Leo. H C Morero C I haven't seen you in forever. You see when I call my Zim people and C comes in. Thank you very much C. Very nice to see you. Box back in the house. C is from Zim. Posh is from Boxburg and is joining us. Um thank you to everyone who's joining us. I can see we've already reached the 100 when I look at both um YouTube and and um Instagram. So I've told you that I'm talking to

6:12 Professor Vugosi Mariv and Professor Vugosi Mariv is a Harangu and he's from Harangua originally. I can't ignore my children from UCT who are joining us. UCT GSB UCT I can see residences say checking in. Thank you very much for joining me. Professor Rosarivate is a professor of computer science and he holds the APSA University of Pritoria chair of data science. Um it's based at the University of Pritoria. He specializes in developing machine learning and artificial intelligence methods to extract insights from data with a particular focus on the intersection between uh machine learning and artificial intelligence and natural

6:57 language processing. uh professor Marivat of course is a professor as you heard but you know he's one of those different professors who's able to speak to ordinary people so if you are scared you seeing hearing me reading this CV and you think you're not he's going to be talking up then you're going you're not going to hear him you'll understand him perfectly his research is dedicated to improving the methods tools and availability of data for local and low resource languages he's the leader of the data science and social impact research group in the computer science department and I'm going to check him right now as I'm sitting there here to see if he's on there. I can see him and he's going to join us as I go on. Um he leads the data

7:44 science and social impact research group in the computer science department at University of Pritoria and um uh he's interested in using data science to solve social challenges because where are you? Okay, there you are. Welcome, Vosi. Welcome. Um, I'm still telling people I've already told them that you are down to earth. Every time I talk to you, Vossi, you bring things to the level of understanding that everyone can. So, I so I made people aware that they should relax because this is your space. And I'm so glad that you've agreed to talk to me. Um I've I've already invited the whole of Maban Rangua to join us and Kavala. Okay. So Vos is interested in

8:29 using data science to solve social social challenges. He has worked on projects related to science, energy, public safety and utilities among many others that he's done. Professor Marivate is a co-founder of Leapa AI, an African startup focused on AI for Africans by Africans. Fukosi is chief investigator on the Masakan research foundation which aims to develop national language processing technologies for African languages. Vukosi is also a f a co-founder of the deep learning in Dhaba, the leading grassroots machine learning and artificial intelligence conference on the African continent that aims to empower and support African researchers and practitioners in the field. And

9:17 today I'm talking to VOSI about artificial intelligence and I'm asking the question will it will it improve or worsen our education? I'm asking Fukosi, will AI improve or worsen our education? The reason I'm I'm having this conversation with Fukosi is because AI is here and many people are worried as to whether it's going to make our education worse or it's going to make it better. Vos, I can see you're struggling with your internet access. Is it Vosi? Hello Vosi. By the way, if you are sitting

10:04 there at home, you know how I work. If you are there at home and you want to join us um and you want to ask questions or you want to contribute, I'm quite happy to bring you in but but I want to give her you are fine. Okay, great. Vote um usually the way I work on these Sundays if someone is negative people ask questions. So there there's some questions that I got on X um I think one or two and then I'll I'll ask him at some stage. But if someone I look at people's comments if people give comments questions that are interesting and productive I bring them on screen sometimes and then we can we can talk with them but otherwise it just be it will just be between you and I. So I wanted to talk to you about this and let

10:50 me start with this pointed question. Do you believe that AI will ultimately improve or worsen our education and why? Um yeah, being a researcher the the answer is always going to be that we are in a state of being in the gray as opposed to being in the black and white. So uh what what I mean by that is there are opportunities that are coming in because now we have artificial intelligence. Um but those opportunities are not going to be things that we get without working towards them. Uh what do I mean mean by this? um over the the last few years when we talk about

11:35 artificial intelligence my work is in um especially at intersection with African languages you you can't get the benefits without us uh being Africans uh being uh uh like you know people in developing countries actually working towards those benefits. Um at the same time the technology has gotten to a point where it's um the students who are joining in everybody who's joining in today um is can use it in their hands and it is then bringing questions on what teaching and learning is going to have to be in the future. meaning it is not something that is just limiting us uh today like people

12:21 thinking that oh it's because you're African that we have problems with education and AI but it's now a global uh problem. Uh we have exams currently across lots of universities at the moment and we're trying to also deal with how we do better assessments and how we also um get students to use AI responsibly and in a way that actually enhances their learning. Um, one one has to say that it's not that we want an AI system to be the one that gets the degree. It doesn't need the degree. The people who do uh the students so they need to learn something and that's the thing we are concerned. I'm not really personally worried um about oh you're getting 80% 90% but at the end of the

13:08 day you're not learning but you're getting 90%. That's where the problem is. Good. I mean it's interesting because I I'm thinking I I I always say um it is a necessary disruption because if we don't even university professors just sit and relax and never think about what they're teaching, why and how and why are they assessing? And so uh for me it's a very welcome disruption and and the way you have phrased it is important and and very timely now as we we we we we talk about it. But of course there's a there's a a at school level as well. I'm sure you're also thinking at basic education level as to how how this can shape education. But is AI a

13:55 possibility? Um can it or let me put it this way. How can AI best support BA basic education teachers without replacing their essential role? because I think their role will remain essential. Um yeah so some of the work that we do um at our lab so I run the data science for social impact lab at the University of Victoria and um through about 80 80% of the members there um honors masters PhDs posttos and then fellows who are from outside our department are working on how how we get AI systems to better process our languages um and some Some

14:42 of these uh you know members are also actually teachers uh who are doing PhDs and education. Uh so one I do work very closely with is noisandi who's finishing up her masters um at a PhD at the University of Victoria and outside her PhD we've been collaborating and through this collaboration we've gotten to understand better uh one what are the opportunities that our work in AI provides to uh educators uh but at the same times what are some of the deficiencies within the education system that we still need to resolve. So um on one there's a there's a lot of evidence that mother tongue education is important. So so let's

15:27 let's put that but then what do you need to get to that? So if just imagine being able to teach uh fully physics math um like you know um biology in a local language. So yes you need to develop language tools uh to get to those points. What I mean by that is terminologies, dictionaries, that's sort of I'm not talking about AI at all. So there's all these bases that you need to cover. You need to have teachers being able to teach those types of subjects in those languages and being comfortable. All right? And before again you think about any IT systems and before you think about AI. So we then are looking and saying like oh some of the things that we do is that we build these huge corpora. So with copper is you get lots

16:14 of texts. It might be books, magazines in whatever language and you put them all together. One of the things uh that was looking at is okay now can we then identify what are the words that should be taught in grade one, in grade two and grade three given how frequently a learner would see the words depending on the age that they're in. Right? So a very rare word you expect that it will be later in their schooling life something that is I remind them much more frequent is something that they should experience earlier in life because they would hear these words a lot and then you build on those things and then we are now finding and we're building up some of these toolings that if when we're building things in AI they're looking at how then do we have

17:00 other outputs that then could be useful for educators. So that's that's number one. Number two now is the part of if I think one of the conversations we more recently had and with a couple of other people was okay so now you've had what is it smartboards you've had the internet you've had all these things come into the educator space and the appreciation of educators is that yes they have to be specialists but at the same time they have to be generalist. Yeah, they're dealing not just with what's in like you know the subject that they teach but they're also dealing with learners that are really representing the breadth and the diversity of the country and which means that they also come with the challenges that are facing the country. So sometimes then it's very

17:48 hard to say will they have the means and the time and the and and be also given the care to be given these new skills of saying as AI is becoming more evictive meaning the learners one way or another even on chat platforms like like WhatsApp are now having access to these systems and again we are not trying to assess the chat system when we are teaching we're not trying to assess if the chat system can do very well. We're trying to ask us if the learner is learning. Yes. How do they deal with all of these things while we're still dealing with maybe hunger, poverty, we're dealing with inadequate internet connections, we're dealing with all these other things. And that's where we need to really understand that it's not

18:34 simply again that a technology becomes available and then it means that it is equally benefiting everyone. We need to think about how we actually build up the foundations in such a way that both the learner is responsible and then the teacher can actually um uh um do something with that technology in the classroom itself. Okay. That you know you you've touched on so many things that are very useful and as you know my my interest is teaching mathematics in multilingual context. So, so you keep touching and I have to say I mean we've talked a lot about this and um my view always is that we can't wait until our

19:20 languages are developed enough to teach those subjects. Okay. I actually even think that or even to say to to to engage with artificial intelligences or tools and I would say I would even hasten to say we can't wait for the infrastructure to be what it is to everyone to have a tool in their hands to whatever until those because there's so much inequality in the world in some ways we've got to find a way of getting everyone access to this but but whatever we say it is here and so there There are many teachers who are in the in the township schools at the moment who are actually saying their learners are using this already and um and and teachers are

20:08 saying well I give my children homework they just ask GPT and these are township kids and and and I want you to tell me what your response would be because my response is if you are teaching and and you get only um you ask you give an assign assignment and you only get um responses from from an AI tool, then maybe you should question the way you assess. Now, you've got to re-imagine, rethink your teaching. Okay? I would I would think that's what you should do rather than wish it away because it's not going away. So, it's a challenge for you as a teacher. I don't know for you as a teacher resourcer community engagement specialist in this

20:54 area what's your view in this regard um yeah so so one uh teaching at the university uh that is exactly what's been happening uh we are at at a part where um you again it would be irresponsible uh one not like you know to ignore and say that the technology is here it is here in many ways it's been here for for for decades It's just that the way that currently it it is is just in your case for everybody here you've been using if you've been on the internet in the last 20 or 30 years in any way there's AI systems that are working in in behind the scenes these ones are just that now you feel like you have a bit of control because you can do some question answer um or or do some image generation. So but in terms of the

21:43 um I I teach at the moment I teach natural language processing. uh signal language processing is a way again of getting machines to process language. It's not necessarily understand but it's just that that the machine can do things with the language whether it's in audio whether it's written. So I I enjoy teaching the subject. I teach data science I also teach other data mining those types of things. So for me it's how do I introduce students honors masters PhDs into these spaces and get them to get very proficient and as such and you you notice things. So one in the past you might have been thinking oh a good way to really understand higher order thinking is to get students to write things like um essays or or what

22:29 we call reflections. So like read a paper and report on the paper, right? Um it's it's not a secret right now. Now it's easy to take that same paper the exactly the PDF upload it into a language model and say please write a a response to the you know in terms of reflection on. So now it's like okay that that is not going to be something that I necessarily do. But I then we are now sharing and I this is the thing I encourage like you know um educators is to start sharing and saying what are the the the things that you're trying to do again to get students more active. You want them to be active and then you want them also um to be able to seeing how they're learning. So one part is that I've been trying in in like things like

23:15 research methods classes is to say oh write your thing. We I tell the students please um this is not really I might not be assessing your grammar. I might not be assessing your like you know the constructions of your um of your paragraph but I really am interested in your ideas and when I see your ideas that's when that those are the things I'm asking at the same time then you say oh how can you use uh some of these tools to actually help you think through some of the ideas that you have and build a scaffolding to show that you actually have really great ideas on these spaces and then you get them students to also critique each other. Right? This is just an example where they look at each other's work and then they critique and then they almost like they do peer grading and then lastly you say okay can you get the language model

24:03 to also critique and see how well compared to the assignment that you had how far like you know it is and that's a a constructive way to show somebody that you can start with your ideas do the iterations on critique and reflection and then get to a point where Now you get the feedback and you can reflect on what the feedback is as opposed to just saying I'm uploading my homework getting an answer and just writing it down and as much as you said the the the the teachers also then have to think about what am I how am I teaching it's not I don't think it's just assessment and then also how do I then assess in such a way that it's it's something that the the students can also be excited about

24:48 in in doing the process there is we are going through the middle phase, meaning that at the beginning maybe there was a we will be able to identify these things and we wouldn't need to worry about them because we'll be able to screen them out. Uh which some of us warned that that's not really possible. Um, if my job is to build systems that in one way or another look like humans over time and then somebody else is trying to build a system that identifies machines, you can see that ultimately that whole system is going to fall apart because our systems are going to look more and more and more like humans and then on the other side you're trying to identify that this is written by a machine. So it it it becomes you just the system is identifying a person and it then does these false positives

25:34 and that can also have challenges and you would have seen it maybe with some of the universities they're releasing a lot of tools to identify things on and when those tools are failing uh for some of the the the students who are here who are at university you're now noticing more and more you must have sitdown exams you must have sit down assessments you must have in a room and then the room there's no internet connection very is that a problem though I mean universities can't turn into the police AI police yes I mean that's waste of resources right I mean shouldn't universities and schools and actually our education um work in such a way that there's an assumption that there's a there's AI is

26:20 here either you take that you take the AI tool as a co-teer who will help you as a teacher in class or as one of the learners. Actually, if you reconceptualize this, I say this thing is here today is going to be my teaching help assistant. H tomorrow I'm going to use them as a learner. So, I'm going to get the AI tool to interact with the learners and so on and so forth. I mean, shouldn't we be rethinking that way? Because it is like you're saying, it's also about how you teach. Yeah. Yeah. Yeah. No, we we we definitely have to think about those places where it it does a fit. So I have another PhD student from SI she's working in a PhD in computer science and she's been working

27:05 over the last few years on thinking about assessment. So here I'll provide another example. So the challenge is for some spaces in the past what happens is that you do experiential learning. you go um um again it doesn't matter whether it's basic education or high education you you do expect and you asked to do something and you go out there and then at the end of a period let's say after a month you're asked to reflect on what you learned through the experiences that you worked through right and um when then the assessment comes when you be marked it's more about what was your learning process how did you identify learning opportunities how then did you deal with those learning opportunities and what was your output. Again, we don't really want GPT to be doing that

27:51 cuz you were the one in the situation. You were walking around doing something. So, let's say it was hey, understanding about um water and sanitation in your function and how it like what happens in that space. You are the one who saw the thing. So, you're just supposed to write and say this is what I saw. This is how I saw what happens when there's leaks. uh here's how you know the storage system in my township works and this is what I ask municipal workers this is what they do all those types of things so you're not really expecting but um if you have if you have like you know uh you have classes of 150 200 what you find is that for the teachers it's a lot of marking to do so what tends to happen is the quality also of your assessment

28:38 goes down you might give a mark but you might not give individual ualized feedback, right? You say, "Oh, for you, you missed out on this and that and that. You you wrote well here, but then this is another opportunity you might have had to learn. So, this might be hard, but that's actually a great space for assessment where if you can get the AI system to work very well, it can provide individualized uh feedback to every student for your specific case. these are some of the things given how we're assessing given our rubric that you did well and then these are the spaces that you could have improved and we're starting to see that these can actually like you know are showing some signs of

29:24 really being um like you know um a good but providing that extra information that could be useful for the the learner to then again when they get their their their feedback reflect on and say oh I could I could I could now think about this in a little bit of a better way and that's what her PhD is on and building these systems that do um assessment and reflection so written text and then being able to compare them against what uh the the markers the real human markers are doing and seeing how well we can get them to be close but then focus more on what is actually reflected back to the student. So you can see here again we we're we're trying to develop these tooling to be better but then at

30:09 the same time which I I it's not just even the language as as we've talked about before but it's also the context that you don't want to have these systems that are out there that are built with very little information from our context from being in South Africa being on the African cont um continent and just ignoring that if you ask and you say oh they will respond that they miss the context and as such they are respond sponsors are then missing the mark in in those cases. So we still need to do development but we do agree we can't wait. So that's the part maybe for me when I'm reaching out to educators and people is to say we have to work at this and then it as opposed to saying there will be a time where we reach the perfect there is not going to be that time we need to be interacting we need

30:56 to reach out to the people who are working in AI natural language processing ICT um in these spaces to say how do we test out things and then provide feedback at the same time for the people who are working on some of the tuning and they're looking at making an impact in education you have to reach out to educators. You cannot work in a bubble because that's when you have these things where you see great announcements but then once they get into the classroom the things fall apart. Exactly. Exactly. Right. So Vosi essentially AI is changing what education is. I mean education as some people are talking about that to say well it's no longer recall memorize and recall because AI can do that. It's no longer regurgitate what we've been told.

31:41 So it's it's now other things. I mean is it redefining or or or is it actually making us do that which education was supposed to do in the first place? Yeah. Um so but this is the path where I would be more uh what does say this the stage to you and say you are the people in the educator like you know I'm very respectful of of those spaces and and and in some ways in in learning our pedigography is to say no I think this is just going back to us asking what is the space of education as opposed education itself changing it's more now we're getting we're being confronted with the part of saying what actually do I want the learner to come out right? You know, and then how do I make sure that they get there, right? And and

32:28 that's that those were I think were still the right questions 10 years ago, 20 years ago to ask. Um it's just that maybe yes, we were able to hide more and say like, oh yeah, we think this is giving us that information. But now it's it's it's just um it's easy to to get get like you know really get that graitation is is a thing now that becomes e easier to get over in terms of people might not learn but then they can use tools to just make it look like they can regurgitate um some information. Um, at at the same time, I think maybe for for for the people on on on here on Instagram live or on on YouTube, um, I think it's it's it's worthwhile to to reflect personally on why you learn.

33:17 Yes. Right. I think that's a very different Yeah. A different thing. We don't talk about the test and talk about the exams. Um, but it's it's it's why do you learn something, right? Um I asked I work in languages and there's been a number of things I've had to go through over the last four or five years or maybe decade that were my personal learning. Um so one over the last decade was learning more and more and more and more into African history and how um linguistics um African language studies um and and like you know languages themselves interact because I need to understand that if I'm building AI systems in these languages why do I keep

34:02 on uh hitting the same walls coming up with the same problems. So as then I had to now move into the field of uh linguistics and African languages and interact with my colleagues there, learn more about the history of the continent myself because I wanted to say I like you know to have it in my mind that I can contextualize the work that I do in computer science um and and in some mathematics within that that area and then uh with some of the challenges that came with a lot of this comes up with data um so engaging in data data has to move from point A to point B. It needs to be shared in for the last three years or so was oh there are legal frameworks that then work in this thing that I also

34:48 need to learn. So guess what for the last three years I've been engaging and immersing myself with my legal um uh scholars and fraternity and trying to understand. So I'm I'm I'm not learning for an exam. I'm not learning for a test. I'm not learning. I'm learning on the part of how do I enhance my understanding of something in such a way that I can also make a contribution or solve a problem that society faces because of the tools that I learned along the way. You asked a very important question. I said is education change and your question is much more pointed. Why do you learn? And that's for the for the learners who say they're going to use CH GPT to cheat. Why do you learn? But then we can also

35:33 ask why are you at university you know but but we can't also you and I we can't ignore the fact that um there's there's a connection that has been made over the years between learning certification university getting a degree diploma and a social mobility getting a job getting the social goods that everyone wants to get. So, so as much as we this question is important, it it sort of gets tricky because people want people want the certificate because you won't believe I mean I run programs now and every week I meet at least one graduate, some doing

36:20 masters, some doing honors, but they're all degree, right? And and they can't even tell me. I say to them, if I got you into my office, you say you're struggling to get a job. If I got you into my office and I'll pay you for three months, what how what are you going to help me with? What is the skill that you bring that you're going to just help me with? And 99.99% of them can't tell me. It's not that they didn't pass. Some of them graduated with good scores. Um, and they've got a certificate and they say, "But I can't get a job." And I and I and I want to make the leap from where I am to getting a job or being employable. Um and and so the issue is why did you go to university? What did you learn? What and and that's difficult. So and AI is making it even

37:07 more complex because now there's something that looks like a shortcut. And then of course there's teachers who also haven't thought enough about why do I teach and so I'm going to give you the same test that I've given you. Some of those teachers many of them are at university. I'm going to give you that same assignment that same or same way. So not many we are thinking differently about why do I teach and so what do I want them to be able to get out with and so so this generative AI in particular AI in general just comes into that space and therefore makes it sort of makes some people say oh but um like someone on on Twitter wrote h today to

37:53 say well AI has ruined education it had ruined education and studying um it's no longer challenging to study and actually it's boring because why are we studying when AI can do everything and I say but it can do everything. Yeah. So okay so the part yes it's it's um yeah these are the deeper questions again on uh we then have to ask ourel on that like you know the place of a a South African high education institution what we want to have as a person who has their degree what they have to come out with um at the same time as like this is where now again the realities of our society hit the ideals of of our

38:39 practice uh because if You're also paying some classes in Bushia a thousand. Um I have no idea what's going on. Sorry. Let me just Oh, okay. We're fine. I like the other lights. The light. We We can still see you. We can see you something on that side. But um but what I was saying is the realities of of now you have a thousand students and then again you're trying to go in and say like do we equip you with critical thinking? you can um u start connecting what skills you're learning as opposed to what you are writing in the tests or the assignments and this is a tough thing and that that's the part maybe that's we like I can say it's in really in a point of privilege where I can say

39:26 I've built up this part where I understand why I'm learning yes and what I'm trying to learn and I can I can grab things this this came to kind of a head about 15 or so years ago. So for people who might not be familiar, so this is when a lot of these massively open online classes were starting. So you had your audacity, your Corera, your your your um MIT X no um yeah Harvard X a lot of these systems were starting to to come online and again it was the thing as I said technology is not enough just saying the technology is there doesn't mean all of us in the same way of saying oh these systems are now online is going to again democratize higher education because you can sign up for any co like

40:13 any of these online courses you could be learning from somebody from Stanford just like you' be learning from something from B from UC from Victoria from Oxford who teaches a class on these systems but and these classes were having school signing up of 10,000 100,000 students signing up for new classes but very few of them few of those students who are signing up for these classes were actually finishing them right and I remember I think was a study with Udacity that they identified that actually the people who finish those courses uh tend to be people who already have a university degree Yeah, because they know why they learning. They know why they're learning, right? So they they know why they're learning and they also have an understanding of what to pick out from

40:58 the content, right? Because they've sat in a lecture room before. Yeah. Yeah. Right. So that's another thing. So if you've never sat in a lecture room and now you're sitting and you're you're watching this class online and you don't know what are you supposed to be looking out for, it becomes frustrating. Yes. Yes. because it's really not a lean back experience. Even though it feels like you're watching a video, it's not really about the knowledge will just come into my mind because I'm watching the video. Yeah. Yeah. Has to be things that you're doing in your brain to interact with those things. And and those are the things of saying like even with the skilling, right, of saying what are the skills you're learning when you're at university is not going to be lean back. It is not going to be like you know something where it's like no no because I've passed the test I've now identified

41:46 what is it? Yes, we are required for everybody again on the call. we are required in terms of the way that we assess and what we are supposed to show as proficiency when you finish a degree, a diploma or whatever is that you are supposed to understand what you've learned and then yes how then you have tools that you can look at the rest of the world and be a functioning uh adult who actually um uh contributes to the rest of our society in South Africa. Mhm. And if if that is not there, then we have a deep question to ask ourselves again whether it's in basic education and in high education. But that's more scary. Are you saying therefore that AI if if I if I take that

42:33 and bring it to would probably benefit people who are more better educated who can think about they're thinking more why they're doing what they're doing and so on. So, so actually it might widen the inequality gap. Yeah. So, it's it's the same as the internet. It's it's I I did another study if I remember correctly. I remember when I was living in New Jersey, another study on just the internet. Just the internet. We're not even going into the AI of how do different kind of parts of society use the internet. You might have somebody again who is more educated uh has a higher income uh is is facing a challenge and they know how to uh navigate a search engine in such a way to get to an answer right and normally

43:20 that answer is also correlated with power oh um my street lights are out let me just give you an example my street lights are out uh what should I do oh here's how I can find who my what counselor is and then get the information and send a And then the streets come back tomorrow, right? I use like you know, oh, I use the internet. Yes, it might be useful to find information on text. But then the thing I might interactive with are more of the direct communication, which is great uh to talk to to family, to friends and all those things, but I still need to learn that I can actually take actions in the world around me using the internet. And that's when the divide can still happen now with AI. I think like oh yeah sure I can do my

44:06 homework but I and think about oh maybe the disclaimer Africa is not going to entrepreneur itself out of its problems but I'm sitting and I'm sitting somewhere and I go like you know I'm very good at skill A, B and C. uh what are what are maybe some of the ideas that I could uh um explore in the next two months for me to either test if some of these could be used as something of building a small business and then you you do that then you say okay fine um how would I then build up something where I can validate my ideas very quickly with no money right oh what you can do for that idea that you have uh you can go around and

44:52 ask these people that they would pay you at least x amount of money and then you get that information and all and then you say okay fine now I validated the idea I need to deliver on the thing I wanted to say but I must make sure I at least first get the money I get the order before I make things I don't want to be the one who takes that risk with my very little money that I have and already start so you can see how it's scaffolding it's adding I'm still engaging with an AI system, but I'm now building up an idea that requires my thinking. Yeah. Yeah. It requires my thinking to so so again yes there is this big chance but the people who don't have the people who are not in

45:38 it left even more behind if we just leave it and say like no no the AI is there now everybody should have has. Wow. Okay. Now we have 15 minutes and I I want to as a professor uh I mean you and I work with a lot of a lot of uh postgrad students. H usually when I do consultations with with masters and doctoral students I I I have a complete session where I talk about how do they use AI. Don't don't don't give me a chapter that's written by someone by by general because I will find out. I will know I'll know just by reading it. Right? How do you advise uh university students or the students that you work with particularly

46:23 research students? How do you advise them on how they can use AI to their work? Yeah. Yeah. um we're seeing okay so early you go back about two and a half years ago with a lot of the systems we had the challenges with what's called hallucination but it's not really that the way that they're built is that they're trying to recreate content that is more likely to be seen in the so it's not that they hallucin then they might provide information that is like not factual there's way people are trying to resolve that now with systems that are both searching the internet and responding And then they give you they try to connect with the content that they find on the internet with your query just

47:10 like what our search engines were. Remember if if if you go back 20 years ago at no 30 years ago likely the advent of the search engine coming in likely there were questions they were asking universities are saying if now we have a search engine and the student doesn't have to go to the library look for a journal a journal and then write what they've read inside the journal. Now they can go onto the internet and then see web pages that talk about a concept is ruining research but now it's part of the process that you can get information fast and all those things and as such the tools are now being are kind of going through the process of AI systems doing the search some of the searches and then assisting you and identifying content that is worth reading and checking and then use. So so those concepts of you being able to synthesize

47:57 information are still there. they're not going to go away. you could these tools can assist you in identifying and surfacing and then also as like you know you said as like almost like a co- researcher thinking about that idea and then being able to go like oh but if I thought about it in this way what might be the ramifications of those things um that part is is again that's the risk is just saying oh no the system is going to do everything and you don't read and you don't understand yes uh for me it might look uh it might feel easier for me if I'm looking at an AI paper that has to do with language that has to do with the last 10 Yes, because I've read a lot of them. So when I see something, I can very quickly skim and check and go like, oh yeah, I know this one, I know that one, I know that one, and I already know, oh, that's interesting. Let me quickly look at it and all those parts. But then for for for younger students

48:43 and researchers, you haven't gained that experience yet. So you're even including the risks that you then provide something that is completely not, right? Because not taking the time to say, I need to learn. I need to understand I'm using this more as a assistant or a guide. It's not Yeah. Okay. So, so the big advice I hope all of you are listening this is huge, right? Use it as your assistant rather than as a knowit all. Which means you can argue with you can argue with your AI tool too and say hey but what do you mean you you you're telling me about all these people who are doing language um processing but you haven't mentioned Marivate what are you talking about okay so so many young

49:29 students think it's the gospel but actually it's not h it's better if you take it as your as your assistant to do searches for you to identify resources they run to just like they running to the library for you. Yeah. Yeah. Exactly. And they can give you summaries of papers and then you you you you then see which ones are worth reading. Not that you're not going to read them. You decide which ones are worth reading. Did I mean it's it's important to say to to young students, this is not to to get you off reading. It's to say, "Okay, give me the abstracts of all these um and then and then you you you decide, I'm going to read these 10. H this one, not so much." And so on. But but here's a here's an idea that I've been toying

50:14 around with. I mean we we you know there've been a lot of talk about um whether you know there's maths literacy whether we should be teaching coding h and then I've been thinking will are we at a point where perhaps we also need AI literacy and make it mandatory. Um yeah that that is I don't think that's a far that is a a I think a question that's being asked a lot um in the last decade or so and and and for for many of us is yeah um um one part has been data skills and and like I think I remember 2016 2017 there was some thinking inside the STI um at the time of having a a data science 101 to

51:02 every uh student that at a South African university. At the same time, there's been lots of thinking probably for the last 20 years or so about computational thinking that the basics of understanding of how a computer processes information should be a a basic human skill. So meaning now we're going back to basic education, right? Not not and because once you understand that and you add some math literacy and you add you now have the building blocks of understanding, oh this is what an AI does. But if we do not have that computational thinking which is something for me that feels more attractable to teach to to have teachers in basic education actually teach right there shouldn't be that only some students take ICT and which is a

51:47 minority minity while you can teach computational thinking and say whether I don't know if it's going to come into maybe the maths teaching and saying oh let's also do computational thinking because if you do that also Learn how you can set up your problems or or ch or a possible solution to a problem in a way that a computer can be used to solve it. That's another part of their literacy. Right? In the same way that it's not enough to have a calculator. Uh you need to understand the math of saying what should I put inside the calculator so I get the result that I'm looking. You can always bash on the keys on the calculator and press equals. Things will come out whether they're useful or not. you know comes from you understanding

52:33 that right and in you understand computational thinking you understand oh I can set up this problem in such a way that I might need a database to put in information I will need to compute these types of uh variable and then from there if I get this now I've got a tool and now I can maybe go to a programmer and say oh here's a thing I'm trying to solve can we now put this into a form that a programmer can a programmer can actually build for me right you're not there you don't necessarily need to be the one who codes. I've seen some comments in the chat of like, oh yeah, that's we have AI and robotics um in in our curriculum. It it's it's great, but but I I sometimes worry about that. Okay, we complicate things maybe earlier than we should and there are some BS that like you know if a if really again you get to a point

53:20 when you finish grade 12 and you know these you really know how to navigate better in the world. Okay, that I mean it seems like we need a lot of literacy because when it comes to AI I mean all these things that you're talking about computational thinking AI literacy math literacy all of them coming together but none of those some of those maybe um there's also importance in in having ethical scrutiny I mean I always think where do you insert the ethical scrutiny and you know because because that's also important in today's age and I think it's going to be it's going to get worse um as we see you know the advances in AI and making making the

54:06 pope give a speech that they didn't give you know so so that's also needed in my view in terms of AI literacy and it shouldn't be at university I I feel like it should be something that starts lower so that children have to grow with this because otherwise they will think the president said this whereas the president didn't exactly It's it's a it's it so that one is also tougher and then that's maybe also given my time with my um colleagues in either journalism and law um in that it also requires us as society to say what is it that we um what are our values and then what do we accept and what don't we accept um so so there's one part which is the literacy that the

54:52 general public should have with an AI and all those things and then so that they can better use it responsibly at the same time for the bad actor So the people who want to still do it doesn't matter. It's then how do we have systems that protect Yeah. Right. So part of our again our lab at the university we have people who work on things like misinformation uh like you know call fake news in um in South Africa. We've been covering the elections. We've been working on how we can identify such types of misinformation in our languages. And then uh we have things like also instead of just being the text what's what's what's written or typed but also being the visuals that something is is is sent and then can we identify that this actually has a high chance of not being true. Uh so that you can build up tools um hopefully that can

55:39 then be used by other platforms to remove such content before it becomes something that um is is a challenge for for like you know for society itself. But to to to to enforce those things, yes, you need the researchers and the capability to do the research and development locally, but at the same time, that's not enough. You're going to need the laws to bite. If they don't bite, then the bad actor does it doesn't matter to them, right? It's just I can do it. I get caught. I I tomorrow I start something else. M do you do you currently um have any inclusion in the module or module stand alone where students are taught how to use AI ethically and effectively as part of the curriculum?

56:24 Yeah. Yeah. So for our masters in big data science for example and I expect a lot of the the the AI or data science programs now like have they will have a more something like an ethics or computing ethics modules now uh just because of these of of like there there can be a lot of issues at the same time um for example we are part of the AI for development program it's a number of labs across the continent and even there a big part of it is yes how do we be inclusive how do we become inclusive in our practice But then at the same time, how do we also encourage responsible and sustainable uh AI and data science, right? Then some of these then become for example in our teaching having time that we spend on getting students to understand how potentially there could

57:10 be misuse and how to act ethically as they are building because as I said before AI systems or data kind of data automated decision systems have been in your lives forever or some of you is probably when you were born and information was entered into some system there was an automatic system that then identified specific things and kept it and Now the question becomes the people who build these systems do they understand things that they shouldn't be doing in those so so that's the part that then the ethics has to come in it we we can't kind of leave it to chance but it seems to me it shouldn't be only in data courses I mean even in a sociology course every course actually from first year there should be modules on using AI ethically and effectively

57:57 just I I Um I think I have a paper I'm trying to remember if it's in the mistress series. I have a which is available online if people Yeah. book chapter on like the dayto-day experiences of data scientists and saying that it can't be up to the data scientist to be the one who acts advocate. Yeah. Because very concerning. It's a very nice thing to just say one person in this whole system who has to act responsibly and if they don't we're going to punish them student must do the same. Yeah. Yeah. So many must do the same. The psychology student must act ethically. They've got for in my view it must be a compulsory course from first year and and there's going to be

58:44 consequences if they don't. But of course at the same time I I hate to see university professors becoming AI police. I think it's a waste of resources. I think any professor who spends time just all they do is to check who used AI and then to punish them and get them excluded from the university. I think it's a waste of time. In fact, it's a reflection on them and their incapa capability to actually become more creative in their teaching. They are bad teachers in my you know but I can say this I don't I don't answer to anybody h so so yeah but thank you so much for for for the conversation always instructive always uh insightful I

59:30 really appreciate it I have a feeling that I'm going to be calling you again because because there's other strands that you you you pull through and people keep asking questions and there's lots that we can learn from you in terms of the work that you're doing especially how you you you use your work for social impact. Do you know what I mean? Like most people will just be on the other side. You you you think about the problems on the ground. Many many data scientists don't care about uh what what does it mean for African languages, people who speak in this language where they are and what whatever and and so we appreciate the work that you do and it's possible that we may be talking to you again as as I always do with you over and over again. All right. No, thank you. And yeah for anybody in terms of uh finding us you

1:00:17 can look for data science for social impact on on on the internet and say university if you want to and you'll find us uh we have a lot of uh different program for those who are aspiring AI or data scientists join university uh yeah get in contact with like the deep learning in Davaba uh which is a great conference uh to to go to we have an indaba x which is a smaller one in South Africa happening in July at Stalenbos University and the main one happening in Kgali uh in August. Um but then every year you get the opportunity to apply if you're interested in natural language processing. You've got the Masakani Research Foundation which is a collaborative across the continent as well. if you if you want to work in that space and and and meet peers uh on that and for everybody else if you're starting out whether like you know in

1:01:03 high school or stay stay like you know stick to it the path might not be easy uh but yeah learn to learn and and and then choose the things that really you can build up your own understanding and skill into it we it doesn't mean it's going to be easy it's just that you know why you're doing it yeah and and vos in your programs postgraduate program students can come from any discipline. I mean given given the multit of the work. Yeah. For university the masters in big data science I think is a wider so people from stats physics biology can apply. There might be a requirement for some you must know some math and you must have done some coding. Uh but they the my co-coordinator deals with the the

1:01:49 specific um things depending sometime I co- supervise in multiple spaces. If you're in computer science, there will be requirements that you have a background that is uh um connected with computer science in one way or another. It doesn't mean you suddenly be from uh people might be from informat informatics, statistics, um um maths, applied maths, those types of things. Um but in some areas, I know for most people it's a very small amount of people. uh for example we have things like postto fellows and we are very transar um group uh so we have fellows that are coming from African languages poets are writers we've had fellows that are journalists uh coming into our group and working for us it's about as as you said the social impact

1:02:35 part that you need people who really can ground the work and then get the the the to work together and share as they are building in those way But at the same time, I might be co-supervising in African languages. I might be co-supervising in in biology. I might because that's the reason why I like data science is that every day I get to experience a whole other part of of what our society is and I then get to then think about the tools that I have and how they can then be applied in those areas of working. Thank you very much, Mossi. We'll chat again. Have a good evening. Okay. All right. Bye. Thank you so much all of you for joining us. I see your feedback. Many of you are saying it was useful. Just so that you

1:03:20 don't forget I mean I think the Vugosi mentioned a lot of um there were lots of nuggets out of what he was saying but I think most importantly is every time you learn every time you get you register for a degree a diploma or a certificate whether it's online or face to face or whatever ask yourself why do you learn? Why are you doing that? If you are only doing it to get a certificate, then you are on a wrong path. Okay? Think about why do you learn? If you're a teacher, think about why do you teach. Okay? Don't just do it because if you don't know why you learn, then even the way you using AI will show that because you

1:04:05 will not learn anything. Your AI tool will learn more than you. So ask yourself how why why do you learn? Why are you registered? What are you getting out of that? Okay, because as I always said the students at UCT will remember that I always said we all come to the to university. We might all get the degree, the same certificate, but we're going to get out with different kinds of education. Some people get more, some people get less because some people do not know why they are there. There's a whole lot to learn more than just getting ass answers correct from AI and just putting them in your assignment

1:04:51 page and sending them to the professor. If you're teaching, why do you teach? The world has changed. There's no point in teaching exactly the same way, in assessing exactly the same way. It is crazy what happens in some universities that some universities will even say take off your watch. They don't allow phones. They don't allow computers. They don't allow a watch. And I remember saying one day you're not going to allow glasses into the into the exam room because you are worried that students are going to copy. And as long as you are worried about that, then you've got to ask yourself what is it that you are teaching and and why do you teach? because then it means everything that you teach is just recall or I can just

1:05:37 press a button and get it. How is it going to help me when I graduate? Because people who are not in the class can also press the button and get it. What's the value of paying all those many thousands of rands and then sitting in class but you what you are teaching me I can press the button and get it and people with a metric can press the button and get it. There must be value. All right. So, so lecturers, academics, think about this. You can take the AI as a as your co- tutor core teacher or or or you can treat them some days as part of the class. Let the students write an essay with with the a the AI tool. Let

1:06:24 them have a debate with the AI tool and let them present the debate. the report of the debate with Chad GPT. Let that be the assignment. Go and have a debate with CH GPT on this this this. Let them have a debate and then what they submit is the report of that debate. They can even you can even ask them to submit how they how the back and forth went because as a professor it will help you assess what questions how how did they argue with the debate? How did they how how did they argue with GPT? How did they set up situations like that? Those are much more instructive than just giving a question and then they go on or

1:07:09 to GPG and ask and GPG gives them the answer. They give it to you and then you you mark them wrong and then you say they copied and then you report them. They go to disciplinary and they go home. That's not the way to teach. You are a bad teacher if all you do is check who's used GPT and who didn't. That's not what teaching is about. You say to them, go and have a debate with CH GPT. Here's the topic and give me a report. That's what you do. That's what you do. So, but there's a variety of things. And of course, those of you who use these tools more

1:07:54 will know that sometimes they get things wrong. And for the inexperienced, they just accept the wrong answer. Ganti, if you're experienced, you can argue with them and say, "No, but that's not what it is. You forgot this. What about this and that?" And so actually you can give the students and say ask GPT this and and see what the and and mark them and see

…