Learning Library

← Back to Library

AI in Education: Future and Equity

Key Points

  • The panel highlights that AI’s role in education varies widely by socioeconomic background, with many students receiving little to no AI‑assisted learning.
  • Current AI applications focus on personalized curricula, teacher‑level content curation, and back‑office operational support within schools.
  • Experts stress the need to overhaul how AI itself is taught in K‑12 settings to prepare students for an increasingly AI‑driven world.
  • Parents in diverse regions (Nairobi and the Bay Area) report that AI tools are slowly entering classrooms, but adoption remains uneven and often intentional.

Full Transcript

# AI in Education: Future and Equity **Source:** [https://www.youtube.com/watch?v=JSnwZmbTMXM](https://www.youtube.com/watch?v=JSnwZmbTMXM) **Duration:** 00:36:25 ## Summary - The panel highlights that AI’s role in education varies widely by socioeconomic background, with many students receiving little to no AI‑assisted learning. - Current AI applications focus on personalized curricula, teacher‑level content curation, and back‑office operational support within schools. - Experts stress the need to overhaul how AI itself is taught in K‑12 settings to prepare students for an increasingly AI‑driven world. - Parents in diverse regions (Nairobi and the Bay Area) report that AI tools are slowly entering classrooms, but adoption remains uneven and often intentional. ## Sections - [00:00:00](https://www.youtube.com/watch?v=JSnwZmbTMXM&t=0s) **Socioeconomic Divide in AI Education** - The panel debates how AI assistants will influence 12‑year‑old learners by 2028, highlighting that adoption will vary dramatically with socioeconomic status and locale. ## Full Transcript
0:00it's 2028 3 years from now and you are 0:0212 years old how much of your learning 0:04is done with an AI assistant a lot a 0:07little or none at all feder baderas is 0:10the responsible AI leader for Consulting 0:12feder welcome to the show for the first 0:13time what do you think I think the 0:15answer is it depends in particular on 0:18one's socioeconomic background yeah we 0:20will definitely be talking about that 0:21Skyler Speakman senior research 0:23scientist welcome back to the show uh 0:24Skyler what do you think yes but a 0:27little and that is intentional for my 0:30three kids growing up in Nairobi kenu 0:32awesome and last but not least is Marina 0:34daneli senior research scientist uh and 0:37I believe we're talking about your kid 0:38here who will be 12 in 2028 what do you 0:40think I my son will be and I think it'll 0:42be a little also intentional even though 0:45and maybe especially because I live in 0:46the Bay Area all right awesome all that 0:48and more on today's mixture of 0:54experts I'm Tim Hong and welcome to 0:57mixture of experts each weeke brings you 0:59the analys hot takes and banter that you 1:01need to keep up with the ever hectic 1:03world of artificial intelligence today 1:05we're going to focus the entire episode 1:07as we get to the end of the year on AI 1:08and education what that means for AI to 1:11be used in education the risks and 1:13opportunities and where we think it'll 1:14go into the future so let's just drive 1:16into it and fedra I want to turn to you 1:18first to just kind of set the stage for 1:20our listeners um I think Ai and 1:22education is kind of widely hyped and 1:24it's sometimes difficult to know what it 1:26is that's going on and so I want to give 1:27our listeners just a lay of the land to 1:29start what are the big uses for AI and 1:31education right now and where do you 1:33expect it'll go in in the next few years 1:36well um I think it's really important to 1:38be thinking if you're thinking about 1:40using artificial intelligence in an 1:41education context first of all to be 1:43focused specifically on personalized 1:45learning and having customized 1:48curriculum but then also utilizing it to 1:51help teachers curate that curriculum to 1:54be able to augment their dayto day 1:57additionally there's all kinds of things 1:59that happened in the the background back 2:01office to help with operations but I 2:04think in in addition to to having a 2:06conversation about what is the 2:08usefulness of artificial intelligence in 2:10an education context I think it's also 2:13important to have a conversation about 2:16how do we need to be changing our 2:18approach to how we're even teaching the 2:20subject of AI in schools today and how 2:24that needs to be changing going forward 2:27yeah that's great I mean I guess your 2:28short answer is it's kind of happening 2:30everywhere front front of the you know 2:32the educational experience the school 2:34operations teaching people about AI um I 2:38guess Bri Sky I don't know if either of 2:39you want to jump in as both parents 2:40yourselves I'm kind of curious about 2:42what you're seeing on the ground with 2:44your kids I mean are you seeing you know 2:45teachers starting to use AI tools or 2:47people being encouraged to learn about 2:49AI kind of curious about how that's all 2:50kind of playing out in your experience 2:52so I am seeing it come out a bit more 2:55with our students and sorry with our 2:56kids and their teachers um I I think one 3:00of my props takes on this is at least 3:03for primary education this is an 3:05opportunity for us to Channel all of the 3:08people hours that ideally will be made 3:10available because of with the Advent of 3:12AI so I think it would be really great 3:15to really keep the Personal Touch in the 3:17primary education space because of all 3:21of the other enablement we've had in 3:22other sectors so I think really keeping 3:25this balance between the role of AI and 3:28that human connection from the in the 3:30classroom uh so important for what at 3:32least we are looking forward to for our 3:33kids yeah for sure Marina what are you 3:35seeing I'm kind of curious I mean you 3:37know Skylar in Nairobi you're in 3:39California very different places uh but 3:41are you seeing the kind of AI sort of 3:44wave appear with your kids education 3:47yeah absolutely uh I think that there's 3:49a lot to be said for some very 3:50interesting uh games and gamification 3:53that is going on in the educational 3:54space so one thing I'll call out without 3:57trying to be a sponsor is osmo uh really 4:00great games that would not have been 4:02available even a few years ago because 4:04of the capabilities of like the uh 4:07tablet camera to be able to see and 4:09directly interact so it's this really 4:10lovely mix of what's going on on the 4:12screen but also being able to do things 4:14that are physical for for spelling for 4:16math for coding uh those kind of things 4:18are really great same thing with 4:20programmable robots botly and things of 4:22that nature that's the kind of thing 4:24that that's showing up as well so I 4:26think that it's very interesting it 4:27gives a lot more options for how kids 4:30could be exposed to these Concepts and 4:32that seems to be a good thing since kids 4:34learn differently yeah for sure I think 4:36it's one of the things I'm most excited 4:37about is like you got all these options 4:39for learning the same topic now which 4:40feels which feels really interesting uh 4:43feder let me ask you this for you know 4:44we're looking ahead to the next year you 4:46know 2025 I think we're going to hear a 4:47lot more about Ai and education you know 4:50what is the big trends you have like the 4:51one thing where you're like wow this is 4:53really the thing that's going to kind of 4:54knock people's socks off in the next 12 4:56months um kind of curious about what our 4:57listeners should be paying attention to 4:59oh well I I was interested this week to 5:01seeing a an article come out of PBS and 5:06their use of artificial intelligence in 5:08particular uh enabling children to have 5:11conversations with some of their 5:13favorite characters in the PBS learning 5:17shows um they are not using generative 5:20AI but more traditional forms of AI um 5:23and it's specifically targeted towards 5:25younger kids who naturally talk to the 5:28TV 5:30uh so I I thought that that was really 5:32interesting I think we're we're going to 5:35see uh more very clever ways as Marina 5:39said about the intersection of of play 5:43and 5:44AI um and I I'm really looking forward 5:47to to seeing what how that shapes in the 5:50realm of education and in particular 5:53ways of being able to harness that to 5:55address more Equitable outcomes in 5:58education yeah for sure and I do want to 6:00talk a little bit about some of the 6:01concerns here I think fader do you want 6:02to go into that point just a little bit 6:03more I know you mentioned it in your at 6:05the top of the episode as well as kind 6:07of these concerns ultimately about you 6:08know the equity of these kinds of tools 6:10it's really important that we're 6:12teaching people how to be critical 6:16consumers of Technology at large it's 6:19really really important and in 6:20particular how to to teach people what 6:23is the real nature of AI and the real 6:27nature of data what one of my favorite 6:29definition of the word data is that it's 6:32an artifact of The Human Experience 6:34right we we humans we generate the data 6:35or we make the machines that generate 6:37the data but it's important to recognize 6:39we humans we have over 180 biases in 6:42counting so what's really interesting 6:45about AI is that it acts as a mirror 6:48that reflects our biases back towards us 6:51but we have to be brave enough and 6:52introspective enough you know to look 6:54into the mirror and decide does this 6:56reflection actually align with my values 6:59my organization's values if it does it's 7:02important to be transparent why did you 7:05pick this data why did you pick this 7:07approach and if it doesn't align that's 7:10when you know you need to to change your 7:13approach and it it goes back to the 7:15conversation I had up at the the top of 7:17the hour about not just having a 7:19conversation on how AI can be used to 7:22transform education but how we really 7:24need to be teaching this in schools 7:27because if you're lucky enough to be 7:28able to take a class in on the subject 7:30of AI or AI ethics or data ethics you're 7:34probably in a higher institution and you 7:36have self-categorized as a coder or a 7:38machine learning scientist or a data 7:39scientist and literally not everybody 7:42else on the planet um so I think we need 7:46to be thinking about how do we bring 7:48this kind of a curriculum that's 7:50holistic and multidisiplinary 7:53uh much earlier in people's academic 7:56careers in fact you know I see no reason 7:59why we should be teaching this in middle 8:00schools and in particular in social 8:02studies class versus computer science 8:04class which I think is where this 8:06subject ultimately belongs that's great 8:09and we want to I want to get some more 8:11kind of concerns out on the table I 8:12think I do want to talk a little bit 8:14about how we approach these sorts of 8:15issues and how do we address them I know 8:17both Marina you and Skyler said well 8:20hopefully a very little and that's kind 8:22of by choice in terms of you know kids 8:24using you know AI to learn and and 8:26specifically kind of AI assistance was 8:27kind of how I I had teed it up I guess 8:29marina maybe I'll I'll I'll choose you 8:31first and then we'll go to Skylar is why 8:33is that I mean what are what are your 8:34concerns there why would you want to 8:36limit sort of like access to these tools 8:38as a as a way of learning well I think 8:41because uh the nature of the tools right 8:43now if you actually we're going to 8:45generative Ai and not sort of the more 8:47traditional MLA aai is that it wants to 8:50adjust itself almost a little too much 8:52to the person and it's a good way to 8:54fall down rabbit holes that kids are 8:56maybe not yet very well equipped to 8:59handle there needs to be some structure 9:01around that so on the one hand it's good 9:03to have the adjusting personalization on 9:06the other hand it can be dangerous so I 9:09hope that there's going to be a decent 9:10amount of oversight for that kind of 9:13thing and a way also of doing critical 9:14thinking so I think that from a very 9:16early age what kids can be taught again 9:18in a gamification way is how do you 9:20trick it how do you break it how do you 9:22make it lie to you it's vs to tell you 9:24the truth it then you you start to 9:26really understand even as a kid how to 9:28set the expectations that it's not an 9:30oracle it's more potentially like Loki 9:33and like the trickster and and see that 9:35that's the maybe the kind of back and 9:37force mildly antagonistic relationship 9:39you might want to have with it it also 9:42it'll help with critical thinking yeah I 9:43love that I like part of the education 9:45here is like getting kids to break these 9:47Technologies it feels like uh very very 9:49rich and something we should talk more 9:50about um Skyla do you want to get I'm 9:52not sure if you share Marine's concerns 9:53or if you kind of think about your 9:55worries about this technology in a 9:56different direction first of all plus 9:57one on the gamification I think that 10:00really is such an important catch uh for 10:02these kids going into this technology 10:04space uh but I I do want to give this 10:07example that I saw earlier on Facebook 10:09this week I think of just a really great 10:11balance between generative AI technology 10:13and classroom leadership one of my 10:15friends from undergrad is a is a primary 10:18education teacher in the US and he had 10:20this really cool post on Facebook where 10:23he generated some prompts that he used 10:25in his class that wrote Act One of a 10:28play and and then his students were 10:30going to act it out and the students 10:31then had to write act two of the play 10:34and so having that type of dynamic 10:36leadership in front of the classroom 10:37with content coming from both generative 10:39Ai and from the the kids themselves 10:42playing back and forth on that I thought 10:43that was just a really cool example of 10:46balancing the roles that come from 10:48generative AI social interactions and 10:51Leadership from the classroom so yeah 10:53shout out to Don Donnie piery on that um 10:55and skyar I guess to kind of close this 10:57kind of section before we talk a little 10:58bit about you know ways of addressing 11:00these types of concerns I'm curious if 11:02there's any other items you might want 11:03to throw on the table I know you know 11:05fedra talked a little bit about the 11:06equity concerns here we just talked a 11:08little bit about kind of like dependence 11:09and personalization as things that we 11:11might worry about I'm curious if there's 11:13other things that you might want to kind 11:14of put on the table that come to mind 11:15you know as we we think about how to 11:17responsibly sort of deploy this time of 11:19tech yeah I think that those are some 11:20really great examples on the the 11:22responsible and the sorry the the equity 11:24angle in particular um my kids do go to 11:28a private school here in Nairobi Kenya 11:30and that is looks quite different from 11:33uh the global majority from around the 11:35world um you know uh north south east 11:38west and so I think making sure that 11:41that is recognized and top of mind for 11:44how these things are are deployed um 11:47across schools of all sorts of 11:49socioeconomic backgrounds uh is a is a 11:52key point that F just started off the 11:53conversation with I I do a lot of 11:55volunteer work with with the Girl Scouts 11:57and we have used games to introduce the 11:59girls to things like algorithmic bias 12:02but then it's I think it's important to 12:04have conversations with them like you 12:05know give me examples of where AI has 12:07delighted you now give me examples of 12:09where you were playing around with a an 12:11AI and the output made you feel really 12:13bad like you knew it was wrong or you it 12:16didn't make you feel good and listen to 12:20what they say it it is I think really 12:23telling when you invite a a a young 12:28person to be critical consumers of the 12:31tech and really be thinking about things 12:34like disperate impact or unfair outcomes 12:37it's it's very very telling and again it 12:41goes back to what I was saying at at the 12:43at the onset of this conversation which 12:46is about this is far more about social 12:48studies like whose worldview is actually 12:52being depicted in this AI model Beyond 12:54just you know can I trust the outputs 12:56whose worldview is in this model model 12:59is being reflected also teaching them to 13:03ask critical questions like who's 13:04accountable for this model how much 13:07better does it perform compared to a 13:09human 13:10being uh it's just I I think these are 13:13all important things we need to be 13:15teaching the Next 13:16[Music] 13:20Generation I think that's a great lead 13:22into the next segment which is you know 13:23thinking a little bit about how do we 13:26kind of address some of this right these 13:28kind of concerns in the tech technology 13:29and I know fedra you've been thinking a 13:30lot about these issues done a lot of 13:32work on it in the last few years uh in 13:34particular I know you know right before 13:36this episode I was reading a little bit 13:37more about your work with smarter 13:38balanced um and I'm kind of curious if 13:40you want to talk a little bit about your 13:42work there and kind of how it applies to 13:43some of the issues we've been talking 13:44about yes um this Ed tech company out of 13:47the state of California they were 13:49interested in addressing inequity in 13:53traditional educational assessments 13:56there's been a lot of research that 13:57shows that traditional educ assessments 13:59are inequitable for a wide variety of 14:01different ways uh reasons including you 14:04know English might not be your primary 14:06language or you might suffer from test 14:09anxiety or you might be neurod Divergent 14:13there's many countless reasons why 14:16traditional tests might not work right 14:19so they wanted to dive into uh 14:24discussing or experimenting on whether 14:26artificial intelligence could directly 14:28address some of this 14:30inequity and so uh one of the things 14:33they tasked us to do was to form a think 14:35tank that included students from all 14:38over the world and teachers in 14:40elementary middle school and high school 14:42as well as people who had leadership in 14:46neurode Divergent communities Etc ET we 14:48we pulled together this Think Tank and 14:51really dove into some very specific use 14:54cases for these AI models like if you 14:56were going to use an AI to to be able to 15:00ascertain the skill set of let's say a 15:03sixth grader's ability to comprehend a 15:06passage of text and have conversations 15:09deeper conversations about that passage 15:11of text right what would unintended 15:15effects of such a model be what are 15:17unintended effects and then given those 15:19potential categories of harm and the 15:22principles that this Think Tank came up 15:25with how would you detail what are the 15:27functional as well as the nonfunctional 15:29requirements needed to to be seen in 15:32such a model and the the principles that 15:35the think tank came up with I think were 15:37really interesting like IBM for example 15:41you know we detail fairness and 15:42explainability and robustness against 15:45adversaries and transparency and data 15:46privacy right this think tank when 15:49thinking about that these AI models are 15:51going to be used by children included 15:54principles like kindness and data 15:57sovereignty and 15:59agency and so a lot of the work was 16:02thinking through what does it mean for 16:04an AI model to reflect a principle a 16:06human value like kindness what does that 16:09look like in terms of feature and 16:11function it was absolutely fascinating 16:14work and that report is being made 16:16public yeah fedra I think that's great I 16:18think one of the things I'm really 16:19excited to see is all of these groups 16:20starting to articulate a lot more 16:22crisply like what are the values that 16:24they want out of these Technologies and 16:25I think that's such important work 16:27because it helps to kind of like really 16:29set up the the goals right like what do 16:30we need to do in order to make sure that 16:32like these systems are um doing what we 16:35want well part of it is we need to know 16:36what we want in in the first place um 16:39Skyler I know I wanted to give you a 16:40chance to give yourself a little bit of 16:42a kind of like travel report I know you 16:43were at the AI safety institutes 16:45conference um which as I understand is 16:48very much involved in kind of the 16:49process of trying to develop evaluations 16:51and standards for the space um did any 16:53of these topics come up I'm kind of 16:55curious about how that might plug into 16:56what we're talking about here uh yeah I 16:59think it came up in two ways one was 17:01directly with education as a use case 17:05and the second one was a bit more 17:07indirectly which is what are these kind 17:10of international AI safety institutes 17:12doing for capacity building and 17:15awareness so we've already hid on these 17:16kind of two topics already about how 17:18important it is for these young 17:20consumers to be critical about that 17:22technology that's the capacity building 17:24and awareness side and a bit more kind 17:26of on the policy side the this kind of 17:29uh technical gathering for these AI 17:32safety institutes were really trying to 17:34spell out how we do risk assessment 17:38everything from the you know the doomers 17:40you know end of the world type en uh 17:43scenarios and addressing the day-to-day 17:46harms that we already see in these 17:48deployed models uh so it was really a 17:51fascinating couple of days between 17:54technology um technology experts uh 17:56academics and policy makers is trying to 17:59come together and put language um so 18:04that in Paris in a few months from now 18:05in February uh you know these uh these 18:08countries can come together and sign 18:10these multilateral agreements about 18:12where they want to prioritize AI safety 18:16again from education from Health Care 18:18from Market competition really really 18:21cool space uh to uh to be a part of uh 18:24and that all just concluded uh last week 18:26in in San Francisco um I was there 18:29representing the uh the Kenya delegation 18:32uh quite quite an interesting event yeah 18:34that's really exciting and yeah I think 18:35part of it is you know especially in the 18:37US right education is kind of regulated 18:39such on a regional level it's excited to 18:41hear that kind of like at the 18:42international level we're trying to 18:43develop these kind of global standards 18:45you used two key statements there 18:47regulating and standards and the 18:50Secretary of Commerce presented at this 18:51conference and she was incredibly clear 18:54the AI safety ins students are not 18:56Regulators they are there to catalyze 18:59and provide standards so it was a really 19:02really cool conversation to have in 19:04there so um both both of those roles 19:06have both of those areas have a role to 19:08play but these AI safety safety 19:10institutes are much more about 19:12catalyzing and forming standards and not 19:15yet on the regulator side um so marina 19:17maybe I'll present to you maybe a little 19:19bit of a a hard question that I've been 19:21kind of mulling over um you know I think 19:24as we've talked about right I think 19:25there's like huge opportunity with 19:26technology there's certainly risk 19:29um but I think there's a lot of work 19:30being done to try to mitigate them um 19:33but I'm sure some of our listeners will 19:34be kind of listening to this episode and 19:36saying well there's maybe one thing 19:38which we haven't talked about which is 19:39can someone just like refuse in the 19:41future to use AI right I feel like do 19:43should we give students kind of the 19:45right to entirely opt out of AI entirely 19:48um it seems like you know a lot of the 19:50discussion we've been having here is 19:51well the technology will be here we'll 19:53just have to kind of mitigate its risks 19:55but curious about what you think about 19:56that is like you know should that be 19:58something we're trying to protect right 20:00as we build this new educational 20:01ecosystem or you know ultimately is it 20:03very challenging just given kind of how 20:05AI appears to be you know headed to be 20:07ubiquitous in the future well actually 20:09that's interesting because I would ask 20:11you what do you think would be the 20:12motivations for a student to decide to 20:14opt out because I can see a couple of 20:16things it can be parent driven it could 20:18be because a student sees that they want 20:20their you know their voice to remain 20:22theirs and not have any AI assisted 20:24anything I mean again can you learn 20:26things without AI yeah we've been doing 20:28it for a while so probably um what would 20:30be the motivation do you think for for 20:32opting out I would say there's probably 20:34a lot of fear of just like the 20:36technology itself right which is to say 20:38um I don't know much about it right like 20:40I learned the oldfashioned way I I'm 20:42sure I'm could imagine that being a very 20:44strong incentive like I learned with 20:45books right like I don't know why we 20:47need these new AI assistants um you know 20:50I I think that's probably one of the 20:51risks uh I'm sure there's also like a 20:53privacy risk I'm sure some parents say 20:54where is all the data about my kid going 20:56do I have any control over that so 20:59you're right I think that there's a 21:00couple reasons why someone might be 21:01concerned about it but I think you know 21:03like any of the new technology I think 21:05there's just like a lot of fear over 21:06what it is and what it might be doing to 21:08your kid right the data is a really fair 21:11risk although that's I think that's 21:13something that maybe parents understand 21:14better than their kids do especially 21:16today's kids they've grown up not not 21:17even thinking about the fact that 21:19everything they do is online um but the 21:22idea of what does it mean to learn with 21:24it I think this goes back to a lot of 21:26interesting things that feder had 21:27pointed out of are you going to be 21:29subjected to biases without even 21:31understanding that you are are you going 21:32to be re ending up in some sort of an 21:35echo chamber are you going to not have 21:37the breadth and depth of uh Concepts uh 21:42that you are trying to go through that 21:43like a human might find times that are 21:45the appropriate time to to push back to 21:47stop to pause to redirect and AI is not 21:50going to do that most of the time the AI 21:51assistants what they really want to do 21:53is keep hurdling along at speed in the 21:56direction that they've been pointed at 21:57least so far Maybe things will change so 22:01on the other hand part of Education 22:02needs to be how do you function in 22:04society and even if you opt out you do 22:07need to know how to handle it when it 22:08comes your way or when it comes the way 22:11of your friends or your family so even 22:13if you have that critical of an eye I 22:15think it's not great to say I'm not 22:17going to learn it's like I'm not going 22:18to learn to follow traffic signals well 22:21I guess you can opt out but it's 22:22probably not a very good way to be a 22:24part of society so you at least have to 22:26learn about it even if you don't want to 22:28fully participate yeah and I think this 22:30is the third topic I really did want to 22:31touch on is kind of we're now moving 22:33away from sort of the AI being the 22:35teacher here to kind of the the the 22:37difficult questions I think really 22:39interesting questions I think around AI 22:40literacy right which is well you know 22:43you might opt out but we actually think 22:44it's really important because you need 22:45to know how to work with these systems 22:47in the future um Skyler you're smiling I 22:49guess you might want to jump in well I 22:50think I was just reflecting a bit do you 22:52think that opt out conversation is 22:55happening at the family level at the 22:56classroom level at the school level I 22:59mean where where do you think maybe not 23:01the opting out but the decisions to 23:03really kind of you know engage with this 23:05technology how how do you see that uh 23:08how do you see that working out in a 23:09practical level what what level of 23:11decision-making do you think is going to 23:13drive that type of adoption yeah I think 23:15it's complex and I could see it I mean 23:16the short answer is I think I can see it 23:18emerging across any of these options 23:20right like a school district might say 23:22this is untested we're going to opt out 23:24I could imagine a parent saying I don't 23:25trust this technology we're going to opt 23:27out I could also imagine a kid just 23:29saying hey I don't learn great this way 23:30you know how I learn best I learn best 23:32with books right I want to opt out and 23:33so I can see it happening across all 23:35those levels I know fedra you're right 23:37in the middle of it I don't know if you 23:38want to jump in and kind of respond I 23:39would say the the reason why an 23:42individual or a group or a school or a 23:45state would want to apt out is because 23:47they don't trust it they don't trust and 23:49there are many reasons why someone might 23:52not trust an AI model and earnestly it 23:55takes a lot of work to earn somebody's 23:57trust it it takes a tremendous amount of 24:00work and it's not strictly a technical 24:03problem at all it is a social technical 24:07problem and with any soot technical 24:09problem it has to be approached in a 24:12very holistic way First beginning with 24:16accountability like do you actually have 24:20a group of individuals who are being 24:23held accountable for making sure that 24:25this model is behaving in the way that 24:27it's intended to be behave are they 24:29being 24:30transparent about this model and again 24:32the worldview that has been embedded 24:35within this model the data was it 24:37gathered with consent is it 24:39representative of all the different 24:41communities which have to be served in 24:43an educational system is it the correct 24:45data to use according to real domain 24:48experts who understand the context of 24:50this data and the relationships between 24:52this data and I I'll tell you I think 24:56it's very unfortunate that that so many 24:59organizations I think are ill prepared 25:02to be held accountable for these models 25:06and again it goes back to why the 25:08emphasis on AI literacy and really 25:12understanding what is the level of of 25:15effort that is needed to put into these 25:17AI Solutions in order to be able to earn 25:20people's trust and honestly the hardest 25:22part as I said the hardest part is not 25:25technical the hardest part is the social 25:27part and making sure that You' you've 25:30got the right organizational culture and 25:33the processes in place as well as the 25:35tools and AI engineering Frameworks to 25:37do this work in a responsible way yeah 25:39for sure and I want to unpack that a 25:41little bit more fedra I think you know 25:42what does this look like exactly A 25:44literacy and practice I mean is it okay 25:46districts okay parents okay kids like 25:48here's a here's a curricula right like 25:50you have to you have to go through the 25:52AI 101 class or is it something else 25:54that you're envisioning oh heck no no no 25:56this it first of all it has to be multi 25:58disciplinary it really now when I say 26:01multidisciplinary I mean like get it out 26:03of just strictly computer science class 26:06and you know have it be where you're 26:09you're bringing in uh School schools of 26:12philosophy schools of government it is a 26:16truly interdisciplinary and the the 26:18challenge I think at least within the 26:21United States I'm not going to speak for 26:23other countries but uh Public School 26:27Systems even higher institutions within 26:29the United States have been extremely 26:31siloed with respect to how they teach 26:35disciplines like artificial intelligence 26:39as I mentioned at the beginning like if 26:40you're lucky enough to take it right now 26:42you're in a School of Engineering most 26:44likely and you're not bringing in 26:47linguistics professors you're not 26:49bringing in philosophy professors to 26:51talk about worldviews and ethics or even 26:54disperate impact to give an example I've 26:58come across uh AI practitioners who are 27:01developing AI models to do something 27:03like offer uh predictions on what 27:06percentage interest rate people should 27:08be given with respect to a home loan 27:10that don't know what the word redlining 27:12is they've never heard it before and 27:15again this points to why we desperately 27:19need to have a multi-disciplinary 27:21interdisciplinary approach to how we 27:23teach this subject in other words AI is 27:26not the death of liberal arts education 27:28if anything it's more important than 27:29ever that's right she's right she's 27:32absolutely right and and even when you 27:34look at generative AI look at how much 27:38it's being used to do coding now what 27:41does that mean in terms of the the 27:42programming 27:43profession whereas now people are saying 27:46we need more English Majors to be able 27:48to craft the the right 27:50prompts right so it's she's 27:53right liberal arts education is now more 27:57important than ever that we understand 27:59what what is inequity what is human 28:02history what is disperate impact how do 28:04we approach ethics in a way that's 28:06holistic and representative of all the 28:08people that we need to serve I'm just 28:10now so much more optimistic about my 28:12undergraduate liberal arts 28:14degree yeah thanks it was all worth it 28:19yeah yeah for sure I mean and and I 28:22guess I don't know it strikes me fedra I 28:24don't know if you'd agree with the 28:24statement that the stakes are pretty 28:26high here in terms of getting this AI 28:28literacy bit to work properly um because 28:32it does seem like look you know 28:35irresponsible deployment of the 28:36technology could lead to some kind of 28:38incident that really reduces public 28:39trust that means there's going to be 28:41less use of that technology going 28:42forwards less opportunities to show that 28:44the technology can really create real 28:46benefit um it almost feels like the the 28:49kind of like getting the trust in 28:50education bit is going to be the thing 28:52that kind of like ensures that we can 28:54actually get to all the opportunities 28:55that we've been talking about here I 28:56don't know if you you'd agree with that 28:58all I I think in order to be able to get 29:00to the opportunities that we're 29:01describing that where you're creating 29:03models that earn people's trust you need 29:05to educate people on what the heck we're 29:07even talking 29:08about like I said what is the real 29:10nature of data because interestingly 29:13working with the clients that I do so 29:17often real domain experts who 29:19desperately need to be part of the 29:22conversations and have a seat at the 29:24table their perception in their mind is 29:27I'm not a machine learning expert I'm 29:30not a data scientist I I I don't I don't 29:34have a degree so why do I really belong 29:37here that's not really my swim Lane and 29:40that's what we've been communicating to 29:42people for decades is that they don't 29:45belong which in fact they desperately do 29:48we desperately need to have hear their 29:51voice at the table and even addition to 29:55those domain experts again where you're 29:57your trying to solution something in 29:59their domain like I mentioned you you've 30:02we've got to have far more diversity 30:04inclusivity in terms of who's developing 30:06these models and the systems of 30:08governance around these models and that 30:10I don't just mean gender race and 30:12ethnicity but earnestly people who have 30:14different lived World experiences coming 30:17to the table to have discussions about 30:21does this artificial intelligence is 30:23this solving the problem is it 30:26reflective of all these the needs of a 30:29wider variance of human beings what are 30:32the unintended effects of these models 30:34how do we Design This in a way to earn 30:37people's trust and as I mentioned these 30:39aren't strictly technical challenges 30:41yeah for sure um reeno I'm curious how 30:44you kind of respond to all this you're 30:45someone who spends a lot of time 30:46directly in the research um and you know 30:49I'm I'm sure like again when I talk 30:51about this with some of my friends in 30:52the machine learning space they're like 30:53this is overwhelming we're like just 30:55trying to get these models to work now 30:57you want to worry about all this other 30:59stuff um and I guess I'm kind of curious 31:01is like do you think like in effect I 31:03think what fedra is proposing is that 31:05you know people who do machine learning 31:06in the future will look really different 31:08right like from the people who are 31:09marginally doing it today um and in part 31:12it'll be that like they will have to be 31:14so strenuously 31:16interdisciplinary that like I think it 31:18might end up looking quite a bit 31:19different from kind of what we expect at 31:20like you know an icml or you go to a 31:22kind of technical conference today I 31:24don't know if you'd agree with that we 31:26used to think that only specific 31:28uh people needed the training to you 31:30know learn calculus and that wasn't 31:32because you're going to be doing 31:33calculus forever it was just because you 31:35needed to learn what it is and how it 31:37shows up and what does it mean to have a 31:39structure and a proof and things of that 31:41nature I'd make a plug uh to join fadra 31:44social studies class statistics early 31:46statistics statistics early statistics 31:48often because part of what you really 31:50need to do is understand how do these 31:52models even remotely work just an 31:54intuition not the deep math but that's 31:56what's going to help you combine that 31:58along with uh your work in linguistics 32:01your work in uh history your work in in 32:04language and and all the rest of it I do 32:06find my own um slightly more liberal 32:08arts background coming up a lot when it 32:10comes to trying to talk to people with 32:12examples that they can understand uh but 32:15also again intuition from my stats 32:18classes comes back time and time again 32:21the explanation of what do these 32:22generative models do they're playing 32:24Guess the next word simple things they 32:26might not be completely accurate but 32:27simple things don't try to boil the 32:29ocean if everybody has just a little 32:31more intuition and then you're going to 32:33be more effective again another example 32:35look at cars none of us understand how 32:38they work but we understand how to drive 32:39them we understand how to regulate them 32:41we understand in general how we we live 32:43with them and use them and and what the 32:44effects are it'll get to that point so 32:47I'm I'm not worried I just hope that 32:49we're not going to be rushing it it's 32:51going to take a little time for this to 32:53become pervasive and be become natural 32:55and sort of second nature to the to the 32:57point about how this is going to take 32:59time again look at the Traditional 33:01School Systems today and how siloed the 33:05approach is and how hard it is to get 33:07these different schools to actually work 33:09together on a collaborative curriculum 33:12like that I think is is what's going to 33:14be the hardest thing to move yeah just 33:16last week I was helping my my 33:1710-year-old make a probability wheel 33:20which is a spinner and it can fall in 33:22one of these things and then I told him 33:24that you know his dad me I do 33:27probability day in and day out at my job 33:30and I could just see his wheels spinning 33:32what what do you mean you know you spin 33:34this probability wheel but it goes to 33:36Marin's point about starting those 33:37conversations early and the the 33:39importance of that type of of of 33:42background and intuition um I'm seeing 33:45it play out already in some of these 33:47some of these young lives so um yeah 33:50again just a a great comment Mara and uh 33:52backing that up with a real world 33:54example from just a week ago yeah that's 33:56great I love your kid imagines like you 33:57just sitting in your office with a bunch 33:59of Wheels spinning them exactly he 34:01couldn't quite get it but I told him 34:03that this is really important and I use 34:05this on a daily 34:07[Music] 34:11basis all right for our last segment 34:13it's the end of November we're starting 34:14to think about the new year I want to go 34:16around and just ask each of you to kind 34:18of tell us your greatest hope for the 34:21new year if you could change one thing 34:23what would that be and uh Marina I think 34:24we'll start with you as much as possible 34:27uh get teachers up to speed and educated 34:30and comfortable and able to own what's 34:32going on they are after all the folks 34:35that drive how it's really used on the 34:37ground and any way that we can offer 34:40support to to teachers to meet them 34:42where they are and make this be 34:44something that's positive in their 34:45classrooms that's a great one uh Skyler 34:47you next doubling down on supporting the 34:49teachers but with 34:52their outside the classroom with their 34:55extra work you know that the the that 34:58sort of stuff I think there are some 35:00some areas that could be lifted off them 35:02uh to make them so much more impactful 35:05and involved from the front of the 35:07classroom so I think ai's got both the 35:09role to play helping teachers from the 35:11front of the classroom but also I guess 35:13what we' call back office stuff as well 35:15that could really really change the 35:18lives and aspirations of teachers that's 35:20a great one and last but not least fedra 35:22well as I mentioned I won Ai and social 35:24studies class and I want it taught much 35:26earlier like like I said middle school 35:29if not Elementary School you could twist 35:31my arm but then also uh I I would love 35:35to be able to see more schools making a 35:38concerted deliberate effort to make more 35:41room at the table pull the seats out and 35:44invite students who don't see themselves 35:47as being technologists and say hey 35:50having a conversation about Ai and what 35:53it means for you and does it reflect you 35:57is core to you having a seat at this 36:00table to be a a critical consumer of 36:03this Tech that's something I would 36:05desperately want to see within the 36:06coming years fedra Marina Skyler thanks 36:09for joining us uh and we'll have to have 36:10you back on in 2025 to talk more about 36:12this and thanks to all of you listeners 36:14for joining us if you enjoyed what you 36:15heard you can get us on Apple podcasts 36:18Spotify and podcast platforms everywhere 36:20and we'll see you next week on mixture 36:21of experts