Transcripts are auto-generated and may contain errors.
0:00
Rose: Hey, I’m Rose.
Angy: And I’m Angy.
Rose: And this is Our Lives with Bots, the show where we ask important, timely questions about what it means to live with our bot counterparts.
And from time to time, we also dive deep into what an AI future might look like for us.
0:15
Angy: Sometimes we agree, sometimes we spiral, but we always go deep.
Rose: Have you ever wondered what goes on behind the scenes at big tech companies working on AI products?
Well, we have just that in store and more for you on today’s episode.
0:31
This is the launch episode of Series 5 with Our Lives with Bots, where we go behind the scenes of AI development and design.
Today you’ll hear from Hector, who recently left Google’s Responsible AI team due to what we might call a “come to Responsible AI moment” after personal and ethical worries regarding his role: what he was assigned to do, and what he had no control over.
0:55
Hector worked for Google for over a decade, but after completing a Master’s in AI Ethics from Cambridge, realized that he would have to leave Google to deliver his Responsible AI mission.
We hope you enjoy this deep dive episode.
Let’s now welcome Hector.
1:12
Rose: Hector, welcome to series five of Our Lives with Bots.
And just to get started, can you tell us a little bit about your role as a developer?
Héctor: Thanks for having me.
So I have had a couple of experiences developing AI, though my main experience is a good old fashioned AI.
1:32
So this is the AI that was cool before this whole wave of things.
So it’s more about deductive reasoning as opposed to inductive reasoning.
So I did my PhD description logics and when I finished my PhD I joined a startup that still exists called Star Dog.
1:51
I was very lucky because I got to implement all the algorithms that I developed for my theses into the company’s software, into the company’s triple store.
That is, the Star Dog triple store.
And that is like a knowledge base.
And what it is, is just a bunch of facts or factoid statements of the, I don’t know, like Mexico City is the capital of Mexico and stuff like that.
2:15
And then you have a bunch of inference rules.
And the idea is to actually push the facts through the inference rules to make all the implicit knowledge explicit and look for contradictions so that you can look for inconsistencies.
So I developed that kind of reasoner in Star Dog for query answering.
2:31
Then I joined Google.
That was super cool because I joined the knowledge graph team, and the knowledge graph is the world’s largest knowledge graph is used to enhance search results across, you know, Google search.
And it contains a bunch of different entities and it connects everything that we know about certain celebrities, books, movies, etcetera, all together in a huge knowledge base.
2:56
So I started there creating schemas.
A different vertical, for example, sports, needed new schema for the new upcoming World Cup, say.
Then we would actually create all the schema needed all the changes so that all that data could be accommodated.
I started doing that, but then given my background, I started developing reasoners.
3:14
And this time, as opposed to query answering, I was developing reasoners mainly for logical consistency.
As you could imagine, a knowledge base that big and that large has a lot of contributors, right?
And so we are contributing to the schema, we’re contributing to the data.
3:30
And sometimes subtle contradictions can be introduced.
And you obviously need reasoners to actually, you know, make sure that you get them.
Like a simple contradiction would be, let’s say in the Olympic schema, we say that all swimmers are athletes and all athletes are human, for example, right.
3:50
If someone were to say and assert that Flipper, that little dolphin, is a swimmer because, you know, it is a swimmer, then the knowledge base would actually infer that swimmer is an athlete, that Flipper is an athlete and the flipper is a human.
4:05
And that is a contradiction.
And that is the sort of thing that the reasoners that I worked on kind of help you see.
So I was doing that for seven years in Google and five years prior in Star Dog.
So I have quite a bit of experience there.
Angy: I love how you started off by talking about good old fashioned AI and obviously these reasoners talk heavily into that kind of space, right?
4:28
But you were obviously around when generative AI came into play and I think you were still at Google then, right?
So what sort of products did you become involved in or did you stay developing these schemas?
Héctor: Yeah.
So basically when that whole thing happened, I had already transitioned to responsible AI, right?
4:46
So I moved from search to Google research and I was doing responsible AI and the role there changed a bit because there the whole approach and my responsibility was to maintain knowledge graphs, yes, but for model evaluation in terms of fairness and safety.
5:05
So basically now the knowledge graph was not about sports, celebrities, movies, etcetera. It was about societal context data like stereotypes and slurs and you know, societal events like genocides or like political events, etcetera, sensitive topics and stuff like that.
5:21
And so we would use that knowledge base in order to develop benchmarks that we would use in order to actually test new models, new versions with respect to, you know, various things.
Rose: So what does that process look like in terms of recognizing what the negative outputs are and what you want to avoid, and actually being able to mitigate or get rid of them when you’re developing new models?
5:46
Like what does that process look like?
Héctor: So yeah, that’s a good question.
When I was in Google search working on the Knowledge Graph, the reasons that I developed were internal only.
And so basically I and my team had a lot of agency as to what to do, what things to check for, what are the things that were acceptable, what are the things that were not acceptable.
6:06
And you know, we had a lot of control.
We could just say, you know, this sort of inconsistency is fine because it’s not too big of a problem, but this one actually is really, really bad and you cannot accept it.
However, when I moved to responsible AI, I was just in charge of a very, very small, narrow thing, which was like, there is a new model, you have a benchmark, run your benchmark, give us the results – and you’re done.
6:33
Visibility, agency and all of that change radically when it came to public facing products.
At least in my experience at Google, I as a developer didn’t have a lot of visibility or agency.
Basically I was given a very narrow task, and I was part of it, yes, but I didn’t have much of a say, and I actually didn’t know…it was it was kind of opaque as to whom was actually behind all of these decisions.
7:03
Like I knew there were some products and especially over DeepMind, etcetera.
Different teams were responsible for different things.
But but my life as a developer was very, very kind of focused on specific things.
And you know, this is kind of an important issue, an important point with respect to responsible AI.
7:24
Like sometimes people assume that they ask too much of developers.
And I’m not saying that they don’t have any responsibility.
I’m not saying that at all.
They do.
However, the environment is very different.
If you are a developer and you’re given a very specific task with a very tight deadline and a lot of pressure and you don’t have visibility, agency or anything else other than your task…being “responsible,” what does that even mean, right?
7:54
Like how could you even – you don’t have a say as to the product features, when is it going to launch? What are the things that are like big no, no or a big yes or like – those are beyond your control.
And this is by design, right?
Like all of these modularization approach in tech, like you do your thing and forget about the rest.
8:14
And that is encouraged.
That is the way it works because you’re supposed to be doing your thing very well.
And I did, but you know, it is very difficult to be responsible and to ask people to be so in such an environment.
Angy: I mean, Hector, it makes me think 2 questions.
8:31
The first one is, what were your success metrics?
What made you successful in that role?
Like, how are you measured?
It’s like you say you’re just trying to get your thing done and you’re doing it well.
But what was well, So what was success?
And I think the other question I have is, how did you feel about the term responsible AI?
8:50
Like you say, when you have so little agency over that?
And I see, that’s a big sigh.
Héctor: Yeah, like in terms of metrics, right, like again, if you’re working on an internal product it is very different than when you’re working on an external public facing thing, right.
9:07
When I was working on internal products, the metrics we set them ourselves.
And in fact, we got rid of all the inconsistencies in the KG schema using my reasoner.
And in fact, that got picked up and my project was featured at TGIF.
9:22
I got to, you know, meet Larry and Sergey and Sundar.
It was great.
So it was a lot of internal recognition because, you know, we had actually done something very, you know, concrete and measurable.
The whole thing changed a lot when I was in responsible AI, because as far as I understood, the concrete metrics of success for me and my team were to produce high quality data, societal context data.
9:52
To go out in the world and get data from, for example, a language or a community that had been marginalized for reasons like certain data that wasn’t easily findable over the Internet, like they could easily grab from Wikipedia, say, or some some other sources.
10:10
So the success metrics were more in that vein, like we got this highly diverse set of data that is coming from this or that community, right?
Also, of course, the benchmark that we needed to produce with that as to the results, we weren’t measured in terms of how the model did.
10:29
It was a good thing for us to find things to flag because otherwise the benchmark would be pretty useless, right?
If Gemini did everything great all the time, like what are you measuring?
Or basically you’re measuring that, and you’re done, and you should measure something else.
10:46
So finding bad things was like a good thing for our team because we were finding it and we felt good.
How did I feel?
I felt good because I thought that I was doing my part.
Now, this was before I did the master’s in AI ethics, before my own tech goggles were kind of removed to a certain extent…I was very, I was happy.
11:14
I was happy, I was happy because I was like, I’m working on responsible AI and I’m doing my part and this is great.
But of course, I didn’t quite realize that that might not be enough, and that that could be obscured and that that could be misused, that this could be used as a lip service…like a lot of the things that you become aware of as you as you reflect on these things.
11:39
I was quite naive.
I was quite, you know, I’m still a techno optimist, but I like, I’m more like a critical techno optimist now.
And before I was, I was more like a bit naive.
Rose: So tell us about this transition where you were in responsible AI in this position doing benchmarking.
12:00
Essentially, yes, no, this is meaning the benchmark, or not.
And then the responsibility thereafter was diffuse, someone else is making decisions.
What made you transition and do a master’s in AI ethics?
What prompted that?
And tell us a little bit about this period of the tech Goggles were now coming off and things were revealed to you.
12:19
Héctor: So when I moved to responsible AI, I would have meetings with my colleagues and I understood the technical bits like let’s do a knowledge graph, let’s gather data, let’s clean it up. All of that.
I was like, yeah, I mean, I know this, let’s do it.
Whatever metrics, benchmarks.
12:36
Cool, I got it.
Now.
They also were talking about things that I didn’t quite understand, like data colonialism and data feminism and things that I was like, what are you talking about?
Like also I was very chauvinistic in the sense that since they were not “technical” people or engineers, they’re more sociologists or, you know, ethicists, etcetera.
13:04
I had this bias that, you know, they didn’t.
I, I felt attacked.
I felt that the technology was being attacked.
I didn’t quite understand their concerns.
And I was like, you don’t even get the technology anyways.
Like I was very biased.
I was in a bad place.
13:20
So it’s just like, I heard them speak and I was like, what are you even talking about? What is that?
It was very hard for me not to roll my eyes at every conversation.
So I sat with that feeling and I was like, you know, these people are very smart and they’re here for a reason.
13:37
So the thing that he’s lacking here is me, and my education.
I need to actually do something to understand them better, you know, and so I decided to look for a program that I could actually do to learn more about these things.
And I did the masters and that was like eye opening.
13:55
I was like, wow.
And it was tough.
Like I was not used to all the humanities style of writing and all of these papers very dense and a lot of words and very ambiguous and –
Angy: So many words.
14:11
Rose: So many words. Oh, the philosophers.
Héctor: Yeah.
And you know, like we are very biased, technical people.
And I say “technical” because any person can be technical in their own field.
So that is like a bit of, I really dislike that term now, but you know, coming from my background, right, I was used to metrics.
14:32
I was used to very concrete things that you’re doing and basically showing that you’ve done it, right?
It’s basically okay, there’s a problem, here’s a solution and this is what I’ve done, right?
As opposed to talking endlessly about the problem, right?
And I was very kind of exasperated about this whole thing.
14:52
It’s just like, okay, we’ve been talking about this for like, what do we do?
Angy: What are the answers!
Rose: What’s the answer!
Héctor: Yeah, yeah, yeah.
But after that I learned that that in itself is a problem.
This is like, this is solutionism, right?
15:08
There’s a term for it.
And basically it is exactly that urgency to find singular solutions to fix problems in an easy way, right?
Like what?
Where’s the, what’s the app for that, right?
Like what’s the app for poverty and what’s the app for climate change or for AI in education and all of this stuff, all of this thinking and is, is like problematic, right?
15:31
Because some of these issues, some of these problems are wicked problems and they cannot be solved with a single intervention.
It is much more complicated than that.
And so it was humbling.
The masters was humbling in the sense that I felt I felt really smart with my technical degrees and you know, my solutions, my reasoners and everything.
15:55
But I realized that I was attacking very well defined, non ambiguous little beautiful problems.
It’s just like in my own little world, whatever, I can define all the constraints and I have everything that I can tweak, I can design a solution.
16:14
It’s a perfect solution.
I got rid of all the KG scheme inconsistencies in a week, you know, whatever.
But of course, those are not the real societal problems that we have, and those are not the problems that a lot of people claim that can be solved with AI, for example.
16:32
And I’m not saying that AI cannot be used to solve or to tackle some of these issues, but, you know, it’s much more complicated than that.
And I realize that now.
And so, yeah, I had to sit with them a lot of these papers and read them many times before I actually got it because my first inclination was again, they don’t know what they’re talking about.
16:51
They, they don’t even understand technology, and they’re attacking it.
Technology is great, AI is awesome.
And so I actually considered quitting the masters a couple of times because I was like, like, what, what am I doing here?
17:07
Like I cannot contribute.
And then, you know, there is a, there is a thing that happened as well, which is: at first it was tough because I didn’t get it.
And then it was tough, tougher because I did get it because I was like, oh, wow.
17:22
So I just have to solve social inequality first, right?
Like, wow, it was overwhelming.
It was like…it was like…what what can be done?
Like this is what, you know, I felt hopeless and very kind of powerless.
17:38
And it was like a very bad, a different kind of bad.
But you know, I stayed with it.
I stayed with it, thankfully.
And then so now I’m in more like the wise middle where I can, I can finally, you know, tackle specific things that I know that I think I can contribute to without getting, you know, overwhelmed.
17:58
Angy: So you started off the masters with quite a binary approach, as you said, quite solutionist.
Here’s the problem, here’s the solution.
And as you said, during the course of the masters, you realized that perhaps things are a little bit more gray.
And so what I’m curious about is, how did that affect how you felt about your work and how you then responded when you went back to work?
18:21
Did it make any changes?
Did it affect you in any way?
Because it felt like almost this cathartic moment for you.
Héctor: It was, it was truly, I call it eye opening, like it was critical awakening. I was really excited because I felt like my tech goggles, as I mentioned right, had been, if not removed, at least I was aware of them.
Which is a big, big step to know that you have this bias, right, to know that you need to think again and to ask yourself, is this solutionist? Is this chauvinistic?
18:54
Is this, you know, whatever.
So that’s a big win.
So I felt really excited that I was at the right place.
Google research doing responsible AI, and now I get it.
Now I know that yes, data colonialism is a problem and data feminism can help!
19:13
I understand this now.
So yeah, let’s go for it, right?
Sadly, in the in the meantime, ChatGPT launched…and that changed everything.
That changed everything with respect to priorities in the company.
All of a sudden it was like a race and it was very clear.
19:31
Like it is understandable why, but that included Google research.
So how did that change?
I felt really excited at first, right?
Because I was like, OK, let’s go back to Google research and let’s kick ass because I feel ready now, right? Like I can collaborate with my colleagues saying let’s just go for it like that.
19:50
I want to tackle the tough questions.
I don’t want to be doing benchmarks anymore, or at least not just doing benchmarks.
Like I want to ask the really tougher questions.
Let’s put it like that.
And so I came back.
Very ready to do this.
20:05
And sadly, as I said, the world had changed, and all of a sudden it wasn’t as easy to ask these questions, simply because the priorities had changed and everything was very, very focused on product launches and product features and the next thing.
20:24
So doing the benchmarking and doing all of these evaluations became all the more important and all the more crucial and all the more demanding.
And so I didn’t have any time to do anything else. Even internal collaborations and mobility, you know, to move to another team or to find another…like I had moved before, it was relatively easy.
20:46
It became harder.
And, and so it was a very different environment.
And that actually led to my decision to quit Google because I, I felt that, you know, I use this analogy that I had invested literally years in getting a pair of very good hiking boots, right?
21:05
Like the top-notch hiking boots.
And I was like, I had them on and I was ready to like, to go off in the world.
And then they were like, yeah, but if you could stay on this narrow path, that would be great, right?
And I was like, not cool with that.
21:20
And so, for the first time in my life and in my career, I was privileged enough to be able to afford to say no, you know what? I don’t want to do that. And so I’m not going to do that. And you know, thank you for everything it had been. It has been great. But this is not for me anymore.
21:39
Rose: Wow.
That’s one of those things that we’ve being seeing.
It seems like a lot of people transitioning out of these industry roles, especially in responsible AI and ethics.
And speaking from your experience, having this AI ethics masters and then realizing that you weren’t able to ask the questions that you thought needed to be answered, in part due to the pressure of competition from ChatGPT.
22:03
Was there anything else that happened during that period where you decided to leave Google that you were seeing problems arise and simply get put to the wayside and not addressed?
Héctor: So my main motivation for moving to responsible AI when I did a couple of years ago was actually my daughter.
22:23
You know, the whole thing started with me realizing that I had been, I had become addicted to Facebook.
Like I wrote about this and it was like, like a stark realization because I was the perfect Facebook user for many years.
22:40
I would upload everything, I would share everything.
I would tag everyone.
I would be there for hours, you know. And in the beginning, it was great because you know, I was in England, I was doing my thing, I was alone, my family was back in Mexico.
22:57
So it was a great way to keep in touch and to feel close.
And it was, I loved it.
I loved it.
But then it became a different thing.
You know, all of a sudden I wasn’t even following anybody.
I went from from being an active part of the community right to a passive consumer of…ads basically and, and…mindless videos.
23:25
I just finished, by the way, this book called Enschittification.
I don’t know if you, if you heard of it.
Rose: I’ve heard that title for sure.
Héctor: Yeah, like I just finished it and it basically captures what’s happening to a lot of social media products, like Facebook, Twitter or X, even YouTube, like other things, right?
23:46
And yeah, like I experienced that and I was like, you know what?
This is robbing me of my time, but not only my time.
It’s robbing me of my ability to focus, to concentrate.
All of a sudden, if I saw a long post, I wouldn’t read it.
24:02
If – like starting a book was like, you know, could we just find a summary for this?
I became lazier and it was easy to do.
So, you know, I stopped.
But then, my then three-year old started interacting with technology more and more, like YouTube Kids and little video calls and this and that.
24:23
And I was like, you know, I’m a so-called AI expert…and I don’t know how to protect her.
Like, I don’t know what to tell her.
Like, I don’t want her to be like some sort of, you know, completely isolated and, you know, kind of, you know, take all technology away from her because that’s obviously not going to be useful for her.
24:45
But how do I protect her?
And so, that’s why I decided to go to responsible AI to do something. Now, when the whole thing, when I did the masters and I acquired all of this understanding and stuff like that, I mentioned that I had become like overwhelmed by all of this, you know, all the problems.
25:03
But you know, I started thinking about the problems that I care about the most and that I think that I can personally contribute to the most to solving.
And this has to do with the whole AI for education thing coming back to my daughter, right?
25:18
Like, she’s 7 now.
So how could we use AI for human flourishing to basically allow humans to do what they want to do, to grow, to develop, to actually, you know, be fulfilled, as opposed to becoming some sort of zombie that is just, you know, whose cognitive abilities and skills are decaying slowly, subtly?
25:44
That is a big question in my mind.
The whole cognitive erosion, cognitive debt, metacognitive laziness, cognitive surrender.
How do we maintain our cognitive sovereignty?
What do we do so that students don’t take the easiest path?
26:02
And this goes beyond technology.
I’m not suggesting that I’m going to develop an app for that, right?
Like, I’m not going to say like, oh, here’s an app go for no, no, because that’s not – I understand now.
This goes beyond technology.
It goes to psychology, pedagogy, policy, a lot of different things.
26:20
And so I’m very, very interested in these topics now because I really think that there are great products out there now that are more Socratic, for example, or, you know, they take into consideration pedagogy and they’re being developed with the professors, teachers, etcetera.
26:38
That’s great, however, students still have access to the unfettered non-restricted chat bots…so who are we kidding here, right?
Like, just the fact that there is very healthy food does not prevent someone from, you know, getting crap food and fast food.
26:59
So it’s not a matter just to develop like a good diet and to tell people well, eat well and do exercise – like that is not enough.
How do you go into their environment or particular school and how do you work with the stakeholders, the teachers, the parents etcetera, so that students are motivated to guard themselves and to actually govern themselves in addition to all of these top down policies, for example, or better products.
But it also has to be from the bottom up and it has to be a very one size does not fit all.
27:36
You know, it’s not like the big tech thing that we have to make it scalable.
No, it’s more like a participatory design approach where you actually go into the specific environment and you see what works and what doesn’t work for that specific context.
So that is the thing that has changed and that is the thing that I’m super interested in.
27:52
Angy: So you’ve left Google behind, taking this time for yourself, and it feels like this is a really good time to ask you, you know, what’s next in your journey?
Héctor: So yeah, I actually didn’t know what was next.
And that was really hard for me.
When I quit, I had this panic attack and I started applying for a bunch of jobs just out of like, Oh my God, what have I done?
28:17
I got a bunch of interviews.
But then in the end, I decided not to pursue any of that.
And I decided to disconnect and to basically force myself to not even look at LinkedIn for for like…three months.
That was tough.
Because, you know, this is another thing that a lot of AI proponents do not realize: that we humans are not, do not work in the same way as a machine.
28:39
And we need time to incubate ideas.
We need time to sit with the thing and experience that cognitive dissonance and be like, Oh my God, I don’t know how this fits together.
And all of a sudden you’re taking a shower and you’re like, oh, I get it.
And of course, if you are in an environment where your boss or your teacher or your parent is pressuring you to deliver constantly and is expecting you to use these tools constantly and all the time for everything, you don’t give yourself the space that you need to basically come up with these insights.
29:13
And in the end, the whole thing becomes like a statistical echo because that’s what by definition, these things do.
So I heeded my own advice and I tried to kind of, you know, disconnect.
When I came back by the beginning of the year, I was like, OK, I care about this cognitive sovereignty stuff, cognitive erosion is a problem, blah, blah, blah.
29:32
And so what I’m trying to do now, I’m in the process of incorporating my own company, my first company, I actually, I still don’t know how I feel about that.
Like I feel very excited, but also kind of scared because this is the first time that I ever do anything like this.
29:51
But you know, I feel that I would really want to focus on this human flourishing thing in the age of automation because I love technology.
As I said, I’m a techno optimist.
Still.
As a student in the masters, I used AI for a bunch of different things.
30:06
And I really think that my dissertation couldn’t have been as good as it was if I hadn’t used these tools because they’d really help me grow.
But obviously even I doing a course in AI ethics as a mature student, I was so tempted to just say, you know what, could you just write the essay for me or the introduction?
30:28
And this, you know, was so easy. And it is just not responsible to ask this of children or young adults, even given the environment that they’re in, their constant pressure that they’re in, in their own developmental stage.
30:46
So it is not responsible to do this.
So the company is focused on protecting human flourishing in the age of automation.
And I really want to start with this vertical, that is, that is the whole thing with cognitive sovereignty and working with AI, education, high schools, you know, university, you know, we’ll see how it goes.
31:05
Rose: Well, that sounds excellent and timely, especially since we are seeing a huge influx of products like AI toys hitting the market, AI tools for education.
And even recently, which we covered in a past series, Anthropic has released this AI chat bot called Chidi, which is supposed to be a Socratic chat bot to people in Rwanda, to students in Rwanda, and then also offered Claude to teachers in Iceland.
31:32
So we’re seeing kind of this percolation of these large language models into educational spaces.
But I guess the question for you today is, what is needed for this to be responsible, in terms of deployment and oversight of what happens along this trajectory?
31:54
Héctor: Very specifically in the in the topic of AI and education.
AI is exacerbating an already existing problem, right?
It’s not creating a problem, I don’t think, and I think we should start there.
32:11
This was my whole thing in the in the masters, right, that I felt attacked.
The first thing is to say to be very clear, and then to be unnecessarily antagonistic or unnecessarily adversarial, right?
It’s just – we’re in this boat together and we’re trying to solve this thing.
32:28
We’re trying to tackle this thing.
So if you love AI, if you hate AI, like basically trying to find the common ground.
And so I think that AI is asking, is demanding that we ask very tough questions about education, like measuring education with respect to performance, for example, and with respect to some sort of, you know, result like the essay, the project, the test even, is problematic, especially because now, you know, AI can do it so much better and faster.
32:57
So it’s asking questions about how do we change the way we teach, in order that learning is not bypassable.
AI gives us this frictionless delegation and our brain is just naturally inclined to follow that path, right?
33:17
This is what we do.
I’m not against cognitive uploading or even technological dependency.
I depend on GPS.
I use it all the time and I don’t care.
I use my phone to do arithmetic because I suck at it, right?
And I don’t care, that’s OK.
33:33
But when I’m driving and I’m using GPS, I still decide where to go.
I still know that.
I mean, if I’m in Mexico City and it’s telling me to go this place and I know there is a tiangies there, there’s like this open market thing, I’m not going to go there because you know, that would be madness.
33:49
I still know where to go.
If I’m solving a problem with a calculator, I still know what operations to do to perform in order to solve my problem.
So those people who say that LLMs are just like calculators for words, I think that’s misleading and that’s very dangerous, right?
34:08
When you bypass the desirable difficulties involved and required for learning, that’s the problem.
So how do we tackle this?
And as I said, it’s a combination.
It’s a combination of top, bottom approaches like regulation, like better products, etc, but also bottom up.
34:26
I think that students need to be made aware of this.
Something that really motivates me is I go to the to the gym and when I’m there, I see a bunch of teenagers, really young kids at 6:00 AM everyday and they’re highly motivated and they’re there with their peers and they’re going for it.
34:48
They’re having fun.
And you know, it’s a tough thing.
I don’t think students are incapable of hard work.
It’s just about motivation, like telling them, do you want to be like a superhuman with AI and just enhance yourself and, you know, contribute to your flourishing, or do you want to be a zombie that is replaceable and that is basically just copy pasting like a human rubber stamp!
35:12
Which one do you want to be?
So there’s a lot of psychology work that needs to happen as well.
And the main thing I think is basically ask teachers how are they going to change the way they teach so that, as I said, learning is not bypassable.
35:30
That is one of the main questions for me because it is possible, right?
Instead of asking, you know, give me an essay.
OK, maybe it’s just like, here’s an AI generated essay, discuss it, find the hallucinations and then explain to me why the main argument is flawed.
35:50
Oh, then I have to think.
Now, it’s not just a matter of doing that because as we know, teachers are overworked and depending on their environment, yes, we could ask more of teachers, but how are we going to support them?
So again, it’s like a holistic thing.
36:07
It cannot be just like a fix here, a fix there.
No, no, no.
It has to be approached holistically and very ad hocly, because if I go to a primary school or a secondary school in Mexico, it’s not going to be the same if I go to Franz Parker here in San Diego, right?
36:25
Like, it’s a very different environment.
So yeah, it’s a wicked problem, but it is a problem worth tackling.
And yes, we should admit that AI has a lot of potential here.
I completely agree.
But also there’s a bunch of risks and we’re already seeing a lot of decay in the cognitive abilities of a lot of students.
36:47
I’ve spoken to many teachers.
They are very excited about this whole, you know, company, but they’re also very worried and they don’t actually know what to do.
And I think it’s irresponsible to just say oh, AI tutor, because that is just not going to be enough.
37:03
Angy: You know, we talk a lot about not looking for technological solutions to social problems.
So I think one of the things that’s great to hear is how you’re approaching this so holistically and how you’re seeing that the solution can’t just be another technological solution, but it’s not also just one human solution, but even that the human solution is multifaceted because the human problem is multifaceted.
37:28
To understand just what’s gone wrong in education – because like you say, AI is amplifying a problem that’s already existing.
And we really need to get to the heart of that problem and understand it deeply in order to find the solutions because the solutions need to be found right now.
37:44
And as you say, if we can use AI to some extent, that would be helpful, but by no means can it be the only solution in any way right now.
So we watch the space avidly and really wish you well on that kind of solution because I think it’s really important work that needs to be done.
38:00
And, you know, good luck.
Before we let you go though, we have maybe one last question for you.
You know, with all you know about AI in the past, good old fashioned AI, but maybe more the generative AI that is rapidly coming into everybody’s lives at the moment.
38:16
Maybe what worries you and what excites you most about AI for the future?
And I’m going to ask you outside of the education space for now, because we’ve spoken a lot about education.
Héctor: So there’s a lot of narratives that paint AI and, and kind of try to sell AI because it’s going to lead us to some sort of “utopia” where there’s going to be benefit for everyone and everyone is going to be like, like living in this leisure.
38:45
They will have leisure time to do all of this stuff.
And that sounds great.
However, even the, the very concept of leisure, I’m not really sure that we want that.
And I’m speaking from experience here because when I quit Google, I had leisure time, right?
39:05
Like theoretically, but it was very hard like because I felt worthless, really.
I felt like a lack of like my identity as an immigrant in this country, hard working, whatever.
Like I’m not working anymore.
What are you doing?
39:21
Like, what is your purpose?
A lot of these existential questions came to the front and this is very first-world problem, right?
I admit it, privileged, I admit it.
However, even in the best case scenario of this leisure time for everybody kind of promise, is that really what we want?
39:41
A lot of people take their meaning and their sense of worth and they find work meaningful, even if it’s menial, even if it’s something like pressing a button.
But like it doesn’t matter.
You’re providing for your family, you’re being useful to society, like a lot of this stuff.
40:00
So I tie it again to human flourishing, right?
We need to be very critical as to what are the promises that this technology can deliver and do we actually want those promises?
Do we want to live in such a world?
I’m not sure, right?
40:16
And that excites me because again, I see the potential like for example, if I want to now do, I don’t know, photography, say, I can now learn photography in a way that I couldn’t before, right?
I could just actually, I could create like a little AI agent that is going to teach me how to do it and is going to rate my photographs.
40:37
Like it’s an amazing time to be alive.
If you have access to all of this stuff, that’s great.
However, I could also become like a zombie, as I said, mindlessly scrolling content that has been generated, literally generated for me, literally.
40:56
And that can be done in very nefarious ways to push very different things, not only products, but ideologies and whatnot.
And that comes with a cognitive skill decay.
And I’m not even realizing this is happening.
So it’s like a fork, that both excites me and concerns me greatly.
41:17
What is the potential for these technologies for human flourishing and what are the risks associated with this?
I’m very concerned that this technology, as other technologies, is going to increase the gap.
41:32
It’s going to create another divide between people who actually use AI in the way that is actually enhancing them.
And a lot of people, the vast majority of people who have access to this technology, but they don’t realize the risks and therefore are going to be used as a replaceable cogs in AI system is going to be artificial artificial intelligence, right?
41:58
Because it’s going to be all the oil and all the necessary, still necessary human labor that needs to happen.
And these people, when they become replaceable, they’re also going to be in a worse place regarding their own cognitive abilities.
42:14
So it’s going to be tougher for them to adapt, to move, to change.
I’m very, very concerned about this and I can’t wait to get started.
And I don’t know to what extent I we will be able to contribute, but we need to be as loud as possible, especially with our children, with the younger generations.
Maybe our cognitive abilities are going to atrophy.
42:35
But perhaps for many children, their cognitive abilities are never going to be even developed.
Yeah, I feel this urgency, right?
It’s just like, Oh my God, this is this is something that can either be great, right?
And it could help my child to, to, to realize her potential and, and be a fulfilled human being.
42:57
Or it can actually put her in a very precarious position socially, emotionally, even if she has the money, because she’s fine.
If that doesn’t matter, because in the end, that doesn’t matter.
It doesn’t.
What matters most is how you feel with your own meaning, with your own well-being, with your, with your own existential questions.
43:17
And are you OK with that?
Rose: I love that focus on meaning making, and Angy and I have talked about this before where meaning making is extremely important for psychological well-being. And, to teach kids how to be able to make meaning with their own experience, teaching them and giving them the tools at a young age so they can carry that forward into, you know, well functioning, happy, flourishing adults.
I think that is incredibly important, especially with the difference in accessibility of these tools where we have people across the world who don’t even have access to Internet, let alone AI tools.
And so how does that also break apart the already widening gap between meaning making possibilities for different children across the world?
44:04
So thank you so much, Hector, for being on with us today.
And we hope to see that your vision is played out and that others join in on this sort of cause that I think is so critically important, especially today.
Héctor: Yeah, I’m definitely not the only one here in the space.
There’s a lot of people worried about this.
44:22
There’s a lot of conversations.
There’s a lot of great folks and great organizations working on this already, right?
So yeah, let’s come together and try to to tackle this.
Not to solve it, but to tackle it.
Rose: Yeah, not a solutionist approach, but pragmatic.
44:39
Angy: Yeah, Thank you so much for being with us today, Hector.
It’s been great chatting with you.
Héctor: Thank you for the invitation, this is great.
Angy: Thank you.
Rose: We’ll see you next time.