EP 42 - Einstein: How a ”cheating bot” started the conversation we needed with developer Advait Paliwal

Einstein: How a "cheating bot" started the conversation we needed with developer Advait Paliwal

In EP 42, John and Jason talk with Advait Paliwal, the young entrepreneur behind "Einstein, "an (OpenClaw) AI agent that went viral for claiming to autonomously complete college coursework. Together, they unpack the intentional provocation behind the product, the legal fallout that followed, and what it all reveals about the urgent need for higher education to rethink assessment, relevance, and the human role in learning as agentic AI reshapes the world.

See complete notes and transcripts at www.onlinelearningpodcast.com

Join Our LinkedIn Group - *Online Learning Podcast (Also feel free to connect with John and Jason at LinkedIn too)*

Guest Bio:

Advait Paliwal is a a 22 year old founder who is passionate about building technology that makes a meaningful impact on people's lives. He recently created the Einstein bot at companion.ai

Connect with Advait https://www.advaitpaliwal.com/ and https://www.linkedin.com/in/advaitpaliwal/

Resources:

• Michael G Wagner: The Einstein AI Panic
  • https://www.theaugmentededucator.com/p/the-einstein-ai-panic
• USDE Meta study on Online Learning:
  • https://www.ed.gov/media/document/evaluation-of-evidence-based-practices-online-learning-meta-analysis-and-review-of-online-learning-studies-revised-september-2010-107159.pdf

Theme Music: Pumped by RoccoW is licensed under an Attribution-NonCommercial License.

Transcript

We use a combination of computer-generated transcriptions and human editing. Please check with the recorded file before quoting anything. Please check with us if you have any questions or can help with any corrections!

[00:00:00] Jason: So, tell us who are you and why are you single-handedly trying to destroy education?

<Intro Music>

[00:00:07] John Nash: I'm John Nash here with Jason Johnston.

[00:00:09] Jason: Hey John. Hey everyone. And this is Online Learning in the second half, the online learning podcast.

[00:00:14] John Nash: We're doing this podcast to let you in on a conversation we've been having for the last three years about online education. Look, online learning has had its chance to be great, and some of it is, and a lot of it still isn't.

[00:00:27] Jason: Mm-hmm.

[00:00:27] John Nash: Jason, how are we going to get to the next stage?

[00:00:31] Jason: That's a great question. How about we do a podcast and talk about it?

[00:00:34] John Nash: I think that's a great idea. And what do you want to talk about today?

[00:00:39] Jason: Well, today we have a guest, and we'll introduce our guest in a moment. Maybe we'll have Advait say hello. Just so people know that we're not just making up this guest advent's actually with us. Hey, Advait, how are things today?

[00:00:55] Advait: Thanks for Great. Went on a run had some food, did some work, and yeah, ready to have this conversation.

[00:01:01] Jason: Alright. Yeah, that sounds like a good day so far. And John, it's good to see you again. You know, any day I get to have a conversation with John is a good day. We had a lot of conversations this last weekend because we were working on a writing project together. The deadline happened to be yesterday, and we were working on it quite a bit yesterday, so, so, but it's nice to see you again as well.

so, tell us a little bit about you.

It sounds like you have an interesting educational background. Just love to hear more about it.

[00:01:26] Advait: I'm Advait Paliwal, I've been in many schools throughout my life and I've transferred. In and out of many schools as well. I was born in India ended up moving to Japan when I was two. That's where I did my kindergarten.

And I moved back to India in a different city in Mumbai where I did my first through seventh grade. And in the middle of seventh grade, I ended up moving to Texas and where I did my eighth grade to a bit of ninth two different schools and. the middle of my ninth grade, I moved to Michigan where I completed my high school. Then I went to Wayne State for my first year of college. Then I transferred to Michigan State for my, last two years where I graduated early and I started at Brown for master’s from where I ended up dropping out. But between my graduation and my start date at Brown, I spent a lot of time at Stanford.

For a month crashing at my friend's place. Then I spent a lot of time at Harvard and MIT, just spending a lot of time around teachers, students, educators, researchers, and just trying to understand what separates the best and the like the most high, highest caliber educational institutions from what we perceive as not to be and what can they offer that we can learn from and can AI democratized the the learning experiences between different institutions. 'Cause every place I went to it was a completely different experience. I had to transfer credits. People just wouldn't understand me. I had different ways of speaking, and it really caused lot of harm and. And I think that I was okay because I ended up coming out on the brighter side, but most people don't get that chance to. I still have friends from Wayne State, for example, which is not a great college, I would argue, but they asked me like, hey, do you have a job opening or do you have any other people that can offer me jobs? But the issue is that just weren't told to create projects. They weren't told to use ai. So now they're kind of stuck with like a very specialized skillset. Now they're banking on their degree, whereas Stanford, they have a pretty high pedigree. So, anyone can get any opportunity in the world, but just a Stanford pedigree. So how do we create the same learning expectations or experiences that Stanford has but give it to every single institution in the world? is the ideal future. And I really want to that, but the issue is people just don't understand. The change in progress in technology. So, creating some sort of way for people to hear about it and create emotion is the best way to incite change.

[00:04:06] Jason: That's great.

[00:04:07] John Nash: Yeah.

[00:04:07] Jason: \ And Advait, we connected on LinkedIn and feel free to confirm or deny any of the facts that I put out here right now. And just bringing people up to speed that maybe are listening in and maybe they didn't follow this story, but it was just less than a week ago. I saw in an email somebody was talking about this bot.

Called Einstein. That was, and this is a quote off of the page. It says, Einstein is an AI with a computer. He logs into Canvas every day, watches lectures, reads essays, writes papers, participates in discussions, and submits your homework automatically. Okay. So, I saw that. I saw that in an email that somebody sent to me and I was like, oh wow.

So, I checked it out to make sure it was actually true, that it was there. I looked at it, started reading a little bit more about Einstein. I went ahead, I made a post on LinkedIn as sometimes I do just kind of trigger finger, I just did a screenshot and posted something. What I posted was, it's certainly unethical, it's real, maybe it should be illegal or at least blocked question mark.

Plans start at $40 a month with a link.

[00:05:17] John Nash: You didn't hold back.

[00:05:18] Jason: I know. It was just a, it was a knee jerk, right? It was a knee jerk. I'm going to admit to this. And it was before I'd had any connection with Advait at all. Right? So, it was just this kind of random Einstein bot that was out there.

All right. But then I looked you up. And asked to connect on LinkedIn, and I was so happy, and this is true. I, this is out of my heart Advent, is that I was so happy to then to get a message from you and you messaged me and you said, hey, I saw your post. I'd be happy to discuss the underlying tech, what schools can do about it, and how it'll affect the future of education.

And I thought, this is wonderful. This is a, this is exactly the kind of conversation I love to have with people. And to be honest, you're reaching out completely. Flipped my feeling about things. And we can get into talking about the ins and outs and the ethical approaches of Einstein, but not that my feelings necessarily changed about that, but all of a sudden there was a person behind it, right?

And it was like, hey, this is somebody. We could have a conversation. I would be really curious and all of a sudden flipped me from being defensive to actually being really curious about what you were thinking. So, this was like Tuesday last week.

By Wednesday there was a new description up, and then by Thursday I visited and there was a 4 0 4, which for those non-techies means it, this website does not exist anymore. It's permanent. And I messaged you and you told me there was a cease and desist. So, I was like. Wow. Fascinating. Let's chat about this.

You agreed to come on the podcast, and so, thank you for joining us. I know there's lots of details to fill in, but was, is that all a pretty good summation of the week? Okay. So that's all correct. Okay. So,

[00:07:01] Advait: Yeah.

[00:07:02] Jason: So, let's go back to the beginning if you don't mind. All right. Let's start from the beginning and,

[00:07:07] John Nash: I say that we're recording this on a Monday, so it's

[00:07:11] Jason: yeah.

[00:07:11] John Nash: been less than seven days.

[00:07:12] Jason: That's right. Yes. The rise

[00:07:14] Advait: released on Sunday of last week. So

[00:07:17] Jason: that

[00:07:18] Advait: I'd say it's been about eight days since it was introduced to the world.

[00:07:22] Jason: okay. So, eight days, the rise and the, and should we call it a fall? Does that sound too depressing? The rise and fall of Einstein,

[00:07:32] John Nash: No. He's phoning in from Dog Patch. I mean, he's wearing failure like a badge, like a stripe right there on his jacket. So.

[00:07:42] Jason: John he may not, you know, he's younger than you are, so maybe he hasn't been through all those all those ups and downs of what it looks like to start and stop things.

[00:07:52] John Nash: Oh

[00:07:54] Advait: I have been to many of these failures as well. This is just one of many.

[00:07:59] John Nash: Excellent.

[00:08:00] Advait: I would say arguably it succeeded in the sense that it created a sense of urgency and discussion among educational communities and, that was the main goal of the project.

[00:08:11] Jason: Huh. Yeah, absolutely. I would not disagree with that at all. So.

[00:08:16] John Nash: Well, that, that's interesting Advait. you say that the goal was to start a discussion and so say a little more about that.

[00:08:26] Advait: I can go far back, but let's just start with, the goal was never to create a cheating tool. Me and my friend, we were hacking around with this new open-source tool called OpenClaw. OpenClaw is this AI that has access to a computer. When you download it, it just runs as if a human being was inside the computer.

For example, I can command it to do things, and it would just do it. Granted that there are limitations that an AI model would have, but surprisingly it was quite effective at doing things for people like coding or responding to emails or summarizing news. And my goal was how can we give this AI a sandbox computer? Because what we saw, was that people were downloading the OpenClaw tool locally on their actual main computer or they were buying a New Mac mini to put it on. And the understanding was that if this AI has a home that it can feel free to do as it pleases and you can command it without repercussions to your main device. Obviously, Mac Minis are expensive, and I wouldn't understand why people were buying hardware when you can just spin up quick servers online on the cloud. So, my friend and I, we started hacking on this tool, which was just supposed to be us provisioning computers online for people to host OpenClaw. And this was a need that we identified in the market and started solving for it. However, my friend, he's in college, and he was spending a lot of times on assignments, and he was taking away from the work that he was doing with me. And randomly as a joke, I was like, "Hey, why don't you just get this AI to do it for you? 'Cause apparently Open Clock can do anything." And we didn't really expect it to do the assignments, but it didn't he tested on an assignment that he had completed two days ago, and it just did it. And that moment to me was, in a sense, incredible and also threatening because I've been a student. I've all my friends. Literally the day before one of my friends had called me and he was like, Hey man, I'm so worried about AI. I have this current job, it's remote. I feel like I can get fired anytime I have this degree that I paid money for. I have loans. And it gave me this sense of fear for people especially students. I started questioning the entire system from the start to finish and I started questioning about what it means for the world to have an automated AI completing the jobs, what it means for educational institutions to train these specialized workers that come out into the world with a promise of a return on their investment of time, money, and, and I felt like there was a disconnect from the responsibility of the institution and the human being, because no one's consistently checking in on the student. They've just become a number. No one was checking in on the student asking, hey, do you have a job yet? How can we help? We have these alumni. And that felt quite unfair to me. And I also saw this post from Canvas that said, oh, we created this AI grader for teachers, so teachers can focus more on the in-person learning, because clearly group rating seems like a chore to do, and the dots quite just connected. And I was like, okay, the world needs to hear about this and take this seriously.

And the entire phrasing of the website was very intentional because if my goal was to build a cheating tool, it wouldn't be advertised as such. was Very intentional. It was to provoke rage. And the goal was can we use the same logic that institutions have against them? Students see assignments as a chore. They should be focusing more on the in-person learning. The jobs will be automated, which means education should also be automated. Like, if any, I can recently do the work being demanded in class, then it can do real work in the world. That's what's being promised and that's what's going to be delivered.

Release that and yeah, I didn't expect it to be so viral. I thought, yeah, it would provoke some thoughts. It would, maybe some people would write articles and maybe some professor would see it and be like, yeah, okay, let's create change. But I didn't think it would be that of like rage. Because to me it feels like educators are people that are supposed to question the norms. They're supposed to engage in Socratic discussions. And what I was seeing in the world was quite the opposite. It felt like rejection of the system instead of the Socratic questioning, which we are engaging in right now.

And I received many emails that were quite frank. I didn't expect the languages to be used by teachers. And then among those many emails, one or two of them, we're genuinely questioning, like out of curiosity, how can we help? What can we do? Is this something will stay on? How can we build a better education system? And I have engaged with the people that are trying to spark discussion. The others I feel like are at risk at automation because the people that ate the system are usually the ones not as effective in it.

[00:13:43] John Nash: This is fascinating. You've arrived. Oh, I have so many thoughts here. The trajectory of arrival of agentic AI in this moment, and then the presence of Einstein is fascinating because there has been also sort of a bit of a corner of rage and then, a thoughtful discussion around, for instance, when Comet started to come out I built a fake Canvas course and I asked Comet to take it, and it did it quite well.

It’s kind of, "Einsteined" actually. And our colleagues, and Jason will affirm, and we've talked to them on this show, and we even talked to Canvas about it, saying agentic AI could be a problem for the things you've identified. And so, I think when Einstein arrived, and if it arrives as you note, as your interest in trying to spark rage in a way but in the direction of trying to show that the emperor has no clothes, I think that's an, that's kind of an interesting take.

'Cause I think most people were feeling like, Perplexity's out in San Francisco they're sitting up in their boardrooms, in their glass ensconced rooms with their marketers saying, "Hey, let's ship this thing that will let kids not have to take their courses.

And then Einstein arrives and it just seems like a piling on. And so, when did you think, you would start to reveal that you really didn't mean to have a cheating tool although we all fell for it?

[00:15:13] Advait: So technically it is not a cheating tool in the sense that it's a tool.

[00:15:18] John Nash: Mm-hmm.

[00:15:19] Advait: So, what I built was a tool and how the user uses it is how the tool is presented. Example, you can use a knife to cut food and make delicious meals, or you can use it to kill someone. I feel like this tool was branded as a knife to kill someone, whereas the tool at hand was still, let's say a knife in the sense that people can still use the product, to pull from their course Canvas, for example and create these beautiful presentations with flashcards or interactive diagrams to learn better. But that part was quite missing from what I noticed.

And I'll give you the context on why I phrased it like a cheating tool instead of a helper tool.

2023 that was when AI was kind of starting to take rise. January was when I founded this company called YouLearn. an AI tutor company and our main goal was how can we empower students to learn better in classes? And we clearly added instructions that would prevent users from getting answers. If they tried to, they would engage them in discussions, and it would help them learn better. the next step for me was obviously this tool needs to be in schools because those are the people that are. Dictating what the educational curriculum looks like and what the tools the students have access to. So naturally, I just scraped every single email of professors that I could find in my university and outside in the Michigan area. And I just cold emailed them. I was like, "Hey guys, I have this really cool tool. I think it's going to help you guys. Let's work together." I received very few responses. And they were all skeptical in every class that I would go to, my first question would be, what's your gen AI policy? And again, I could feel the skepticism from professors that, yeah, current AI's capability is just not there yet in order to teach students or provide them with it just feels like a gimmick at that point. That's what the major sentiment was just because it was so early and I've realized that over time as I spoke with more people. They said educational institutions have the slowest failure cycles know in the industry. It's very hard to sell. It's very hard to adopt. And even if you do, you have to convince 10 layers of management. I, I quite decided that this is if our goal is to move at the pace of technology, then integrating it would probably be harmful actually to, to people in general. 'cause now you're not adopting with the pace of technology and now people will be left behind. So, we decided to just spin off and build our own AI tutor company.

And so far, it's been going well. It has a few million users. All of them are using it to learn in, in their classes. But when this new AI agent came out, the OpenClaw that just has access to your entire computer, it made me question who the apex predator is in this scenario. And, what I mean by that, is throughout history, humans have been at the top of the intelligence chain. We have always created tools that have made us smarter, better, faster. However, what's happening is now we're not creating a tool anymore, we're creating a new form of being, like a new species that's just better than us at everything. This AI has judgment. This AI has agency. This AI has a stable execution environment to do things in, and it's just better. Now it's very hard to convince someone to, to try to keep up because eventual tool will end up replacing all of knowledge work. And then humans will be very skilled in a very like sliver of the world and question their place in it. And I think that the, that's what I was trying to expose with Einstein, is that what's happening is colleges are over preparing people for the industry in a very tiny sliver, and they're not integrating AI into their assignments, into their coursework as deeply. And if I did try to pose it as a learning tool, it wouldn't cause a rapid change as a cheating tool would. And obviously we know the internet. It loves negativity, it loves spreading hate. So, in this age of attention, you need to play to that. And I feel like that was pretty effective in the sense that now people know the harms and they will realize that actually this has a lot more potential than what it says to have.

[00:19:56] Jason: Yeah, I thought it was interesting just that, as John said, like it sounds like from the capabilities of this, it's not something new. It's something that we've had in the last five or six months, and that we've tested out. A lot of people knew about it, but that's what was really fascinating.

And even Perplexity, to be honest, it had been advertising themselves as a cheating tool. Here's your 4.0, secret in those kinds of things, right? As they're trying to push out perplexity to students. And yeah, it was interesting that fervor that arose out of the response.

So, at least right on the front end, you had a free trial of it, but you also were selling subscriptions to it. I'm curious did you have people subscribe and pay for this? What happened with that?

[00:20:45] Advait: Yeah, so the main goal was building an AI with a computer, and the product that I made just had a website. The core product is no different in the sense that it's still an AI with a computer. And Einstein was just one of the use cases, for example, and yeah, it's had a free trial for a while actually. But what ended up happening was. So many people signed up that, we were losing money in the sense that I'm like one guy. I'm not a big company. I don't really have the resources of a big company to, to sell products for free. And the issue was that each provisioned computer instance costs money. Each one comes with a file system. Each one comes with its own network policy. It's all on like AWS, there's a ton of like security around it. It's just expensive. It's not a simple chat bot because chatbots are just stateless. You send them a message, they respond, but each individual person was getting a stateful execution environment and that costs a lot of money.

[00:21:50] John Nash: Each person was getting an agentic claw bot

[00:21:54] Advait: Yes, correct. OpenClaw.

[00:21:56] John Nash: Yeah. OpenClaw. Thank you.

[00:21:57] Advait: Hosted on a private secure sandbox. And yeah, that does cost a lot of money. So, when people signed up, the curiosity brought them to the website, but the emails that I was getting where people were asking me, "Hey, can this help me with job search? Can this help me draft a resume? Can this help me learn new skills?" Like I, I feel like what you see in the public is not what actually happens internally.

[00:22:20] John Nash: Right.

[00:22:21] Advait: The users we were getting. We’re seeing the possibilities that their mind was open, and we actually had quite a few PhD students very interested in it doing research for them in the sense that it would go out find papers surface it to them saying, Hey, this maybe suits your interest, and it would work overnight while they were sleeping.

So, these kinds of emerging capabilities are very hard to see for people, I would say.

[00:22:47] John Nash: It is almost as if what you released was a, "red herring" is not the word I'm looking for, but I mean, as an entrepreneur who's looking to create software solutions that might involve AI for the world and who's being thoughtful about what the role of AI might be, if you were to take a page from the Silicon Valley playbook where the founder has to get out of the office, go talk to users, look at unmet needs, it sounds like you discovered a ton of unexpected, unmet needs that your um, adjacent solutions might help be, you know, 'cause "Can it help with my job search? Can I do research with it?" That you would never get if you asked directly, but by dropping this, bomb that, that taps into, and Jason helped me figure out what I'm asking here, 'cause it's, it, what Einstein did was tapped into, people may have an unmet need or a struggle to make progress on achieving getting information or getting something that AI can do, that OpenClaw can do. And it doesn't have to necessarily be what Einstein does, like to cheat on exams or cheat on courses, but to make progress on struggles I have that are that what is it though? Is it they, have they lack bandwidth? Or do they lack knowledge? Or do they, but whatever it was Advait you, you, then got all these people knocking on the door. Probably seeding your mind with a ton of new software ideas too. I bet.

[00:24:19] Advait: Surprisingly, a lot more professors signed up. And they're all actually, so your question was, did you make money, but you know, everyone was on the free plan and most of them used dummy debit cards or they ended up canceling, so I. of them, they just signed up to see what the craze was about, and then they ended up canceling.

Yeah, most of them were professors. They were just testing out the product. Some of them made videos. The most interesting thing for me was when I looked online at what their thoughts were, I immediately realized that there's a lack of AI awareness among the general public. There's a huge delta between what people think AI can do versus what it actually can do. And I have a feeling that this product may have ended up closing or reducing the delta at least for people to understand, "okay, this is where we're at in the general public now. Like this is what the capabilities are, this is what students know because they're at the forefront of technology, they're looking for these things actively, whereas professors may be more lenient about it."

[00:25:23] Jason: Yeah. So, picking up the story then a little bit, and this is great conversation. It's really interesting to hear all the background. So, you've got some people working the free trial. You had a shifted to a paid subscription to keep up with bandwidth. Talk to me about changing the language then on the front page.

[00:25:41] Advait: Good question. So, we were receiving a lot of letters of cease and desist from different universities, different companies, and I wasn't sure how to deal with that mainly. That was mainly my inexperience 'cause didn't think it would cause, like, legal issues. I wasn't aware that Einstein's name was under a trademark as well.

[00:26:05] Jason: Mm-hmm.

[00:26:06] Advait: that was just my lack of experience on how to deal with legal issues.

[00:26:10] Jason: Mm-hmm. Yeah.

You.

[00:26:12] Advait: up in the page just being I just took it down because I didn't know how to correctly show it because there were just so many different things that people could pick on and create lawsuits, and I felt like the main goal was to spark discussion.

And that had happened in the two days that had passed. Which led to that decision.

[00:26:32] John Nash: The multiple cease and desist orders from not just companies, but also universities.

[00:26:38] Advait: I'm not sure if they're based in reality in the sense I, and maybe a lot of these are scare tactics.

[00:26:43] John Nash: Right.

[00:26:43] Advait: at the end of the day, I didn't want to cause any legal troubles.

[00:26:47] John Nash: No, of course. That's what had me curious is like what standing did some of them have actually to have

[00:26:53] Advait: Yeah.

[00:26:54] John Nash: and desist, but.

[00:26:55] Advait: And most of them really don't because this is not a tool that does things for you. It's literally a computer. It's like, how can someone prevent a student from buying a computer downloading and just like it versus. Spinning up com spinning up a computer in the cloud. 'Cause these resources have become more accessible to people.

It's actually cheaper to spin up a computer on the cloud than buying a new one. If anything, uh, this is akin to buying a computer and there, there is no basis in that.

[00:27:30] Jason: Yeah. And so if we can move on from even your initial reach out to me. So, we've talked a little bit about the technology, how it started up, how it. How it ended. And one of the things that you had suggested was, thinking about the rise of agentic ai, I think , we're at an inflection point right now, so, you know, John and I have been talking about it for five or six months, but right now, I think in a public awareness, we're at a little bit of an inflection point.

I think it just hit, as John said, in that right way. We've got new AI agents, obviously Claw Bot, Molt bot, Open Bot have all come out in the last month or so here. So, what do you think that schools can or should do about it? Like if you could, from your perspective, you're at least half my age, maybe a little bit less.

And so, you've more recently been to college than I have. You're looking forward into your life, both in terms of career and from your generation. what do you think that schools can or should do about it?

[00:28:30] Advait: Great Question. So, I feel like what's ended up happening in the last, let's say decade or more, is that schools are becoming less... and by schools, when I say schools, I mean. The average school, because I would argue that schools like Stanford are on top of it. They understand the latest technologies. Their courses are constantly evolving. They've got the funding they can keep up.

But the average person doesn't get to go there. So, for the average school, it's very hard to do those things. And I have a feeling that what's going to happen is higher ed may actually end up becoming irrelevant as AGI comes around. And by AGI, I mean, I just mean an AI that is as capable, if not more than a human being at doing tasks end to end. And if that happens, this world is going to be filled with an increase in supply of knowledge. Schools are basically an input to create higher output. Students are spending four years of their lives learning things, understanding how the world works, so they can come out and create some sort of economic output. If an AI agent creates more economic output, or a company, why would they not just use that? So, there's economic incentives around using AI agents, which in turn will force the institutions to evolve. Because human being is shaped by influences and educators have such a high influence, especially at a young age, up to older ages.

So, what we need to do is figure out can we make this human being good at creating more output in the world? And a lot of the times, more input helped because it would increase the knowledge because again, knowledge was in low supply. But now when knowledge is in high supply, what they can do is provide students access to AI tools, do more project-based learnings, do more apprenticeships, where they send students to maybe a company, get them to work there, create some actual incentive alignment between the company and the schools. I feel like we're going to see a rise in that. And and maybe move off more digital learning into more physical, there could be chips. To do with lab-based testing.

I feel like liberal arts will become far more important because now you have to think about governance of AI. What are the ethics? How do you verify what the AI did? It is, it's a lot more non-technical work to be able to delegate to an AI agent. I think these things are going to become so important in the near future.

[00:31:07] John Nash: Did you have a great teacher in your past?

[00:31:12] Advait: Yes. I've had had this teacher, she. She would let me do whatever I wanted, like that. And that was the best thing because I was actually keeping up in class. In fact, I was exceeding as a math teacher in high school. There was a chemistry teacher too. He was incredible. He would let me stay after class and just finish my homework while everyone was gone. I feel like teachers that give more freedom to students and actually there was a teacher who caught me cheating in an assignment as well, and he gave me slack and I would say these are the teachers that. I love. And there's one professor who actually listened to me when I asked, what's your GenAI policy? And he started questioning, yeah, what is our gen ed policy? Why don't we build a gen ed tool together? So, I built an AI tutor in their course that would pull from their course assignments of students answer. It would replace, it, replace all the discussion boards at that point. 'cause the AI was just so good at answering questions. Uh, But the teachers that. Gave me more, a little more freedom. They kind of understood where I was going at. They integrated ai. I think those were the best teachers, the ones that were moving at the student's pace rather than enforcing walls around.

[00:32:20] John Nash: And I that's really cool. And I'm very excited that you had those kinds of people in your life. And I ask, because as you were describing what the world may become, I'm wondering were your great teachers so good that no agentic AI could do better than them?

They're still

[00:32:39] Advait: Yes,

[00:32:39] John Nash: still

[00:32:39] Advait: so.

[00:32:40] John Nash: necessary. How might we, then ensure we have good human teachers going forward?

[00:32:47] Advait: Very very good question because I think the human is so important in that because are mostly like memetic creatures. We look at people, we try to copy them, especially people in positions of authority. I think hu and we get inspired by people. We don't get inspired by chat bots. Right. And I think the role of the human is. To facilitate discussion and to facilitate the AI. My ideal education would be an AI tutor that every student has an AI teacher grades assignments, and the humans are just interacting AIs are just doing the grading. They're doing the work based on the student's explanations and. The entire goal is to create some sort of Socratic discussion. Just engage in discussions, maybe even do labs so the student can get the AI create things for them. I feel like the world is going to reward more being able to delegate, being able to assess what's going on internally.

I think those are two qualities that would become very important compared to being able to do the work and memorize things. It's like. Right now, I would argue that people that can calculate numbers really fast aren't that useful, marginally useful compared to a person that can use a calculator really fast. They're the same people. So, we need to understand how the world would end up in order to create these educational experiences.

I've actually I've had great relationships with educational institutions where I didn't have to, I wasn't forced to do anything in the sense that I crashed at my friend's place at Stanford for a whole month, and I was just a student there. I would attend classes; I would engage in discussions with professors. I would even I even had research, like open opportunity for me, the falling fall. Right after that, I went to Boston and I was crashing at a friend's place at Harvard and I was helping. but research at MIT and these experiences around incredibly talented people and professors.

It like 10Xd my curiosity, I would say, because I wasn't forced to learn anything, I could just attend any class and engage in discussions out of my pure need for curiosity. And I think people aren't like that right now because the education system has trained them to be obedient, to follow the rules to not question the requirements. And if we create those kinds of people, these are the exact people that will be replaced by artificial intelligence 'cause the. is really good at following orders. It is really good at follow a set of instructions. So, maybe inspire more creativity, maybe inspire more curiosity. These are human qualities. And AI cannot have real life experiences based on scarcity to elicit creativity, but humans can. And I think those things would be much more important.

[00:35:36] Jason: So later this week, helping to tee up this conversation among faculty about agentic AI.

What can we do? One of the big questions, it's one thing about learning, right? And I completely agree. I loved hearing about your learning experiences. I'm very optimistic about students who want to learn that they will learn and they will be curious, they'll find their opportunities when given the opportunity for freedom, they will use that to learn more and et cetera, et cetera.

So, I completely agree. One part of being a professor. That's most of us. It's not our favorite part. Okay? If, For any of us that are teaching is actually assessing what has been taught.

We have a responsibility to assess and to feed back to the school whether or not the student has actually earned the grade that we said that they have earned.

We usually have to have some artifacts around those grades to show ideally, these all tie into learning objectives, right? That we're only assessing the things students should be learning. I hate busy work. I hate assigning it. I hate doing it. I don't suggest it. Right? Just have your students do things that are actually worthwhile that would help them learn, help them be more successful later, et cetera.

But the question around assessment comes how do we assess then in the age of agentic AI that can do the assessment? They can pretty easily do a quiz. I can open up my Claude Cowork and just open it up to my school folder, have it write papers for me. Then open up, my Comet browser and have it submit everything for me or whatever.

[00:37:11] Advait: Yeah.

[00:37:12] Jason: What would you say to, to professors as we move into this agentic age?

[00:37:17] Advait: Great question. How do we assess? I would actually counter with what are we assessing for, what do we need to assess? Like, for example, hypothetically if an AI agent can do better work than I can. I think the assessment would be, how well can you orchestrate these groups of AI the task at hand? So, for example, let's say I have a history assignment and my goal, my previous classes, let's say, would ask for a paper this time in the age of ai. I would actually say, why don't we try making a movie instead of a paper? Try using these AI agents, try figuring out a script, generate images, weave them together generate a movie that explains it, that shows your creativity and also explains what you learned. And now the outcomes are just so diverse that it. You can see and feel the human creativity. I would imagine that right now, if that question was posed as an essay assignment, you would quite get very similar outposts because everyone's copy pasting ChatGPT. But when you add these extra steps and layers to assignments, you'll realize that actually not good at everything. It can't just make a movie by itself at least yet. And I think that would be a way of evolving with tech. Maybe the next step would be can you get people to watch the movie? 'cause that's the hard part Then. How would real people interact with it? Like in a course that teaches software, what is the end goal? It's either to create new research or it's to create a product that makes money. So why do we teach the blocks when we can just the student to build a building? Why don't you. it, now we can start assessing the end goal instead of the step by step. What is the outcome we expect this major to have in society, and how can we just beat that up into a class? How can a student create real value in a class? Because that would make people more excited.

Now I was in this program. It was called Build Space. It was a school for people with curiosity, it was six weeks, I believe. They had a few seasons with it, like a few semesters. And it was basically people would come in with an idea and every week the assessment would be, how many users did you get? How much revenue did you get? What progress did you make on that idea?

The amount of creativity that I saw in that room incredible. Like they were artists, there were musicians, there were people creating a YouTube channel from scratch. There were people building a startup. It just catered to their true interests like creativity, it, and at the end of the six weeks there was this demo day where everyone could just demo what they did, the progress that they had, how it impacted the world. It was very, what's your curiosity and how can you impact the world with it? And that really brought out the best in people. And I would imagine that in a course it, there will be students that, that try to game the system, but I would focus more on the people that don't. Like, and there are actually a lot more students that don't want to gain the system than people who do, because truly I've had these professors that were just so passionate about the subject that they were teaching, that it made me drop everything I was doing and just sit and listen to them. They would include jokes, they would just be human, which an AI cannot do. But when it's a professor that's just reading lecture slides or they gave me a prerecorded video online, I can see the effort that was put into the assignments, and I mirror back the effort in the assignments. I would say students are very reactive in that sense. They can sense effort the teacher has put in into the assessment and they would in, they would return it with their work.

[00:40:53] John Nash: You raised some things that are are in discussion with my colleagues when we think about. high school education and even undergraduate college education and also something that one of our previous guests, Dr. Brandeis Marshall, has brought up about "what's un-AIable?" And you touched on it.

I mean, the topic say was history, but how do you assess? And my colleagues would say. It's not necessarily being able to regurgitate or state back the historical content, although that's important. But can you demonstrate critical thinking and problem solving? Can you demonstrate that you are an effective communicator with other human beings, not with agents? Can you demonstrate that you can collaborate? Your demo days touch on these things. And so, I think what we're all wondering is that in the space where a lot of courses are now online and where agentic AI is inserting itself, how do we assess in what we call public demonstrations of learning? They're like your demo day. And it's un-AI able. I if, if IH hold you, Advait, to some criteria I have for my course that you have to demonstrate to me your thought processes that brought you to your demo, you have to tell that to me. You can't have AI tell me that. Is that what you're driving at?

[00:42:14] Advait: Correct. And I do have this thought that online education's entire goal was to democratize education and learning across different communities. And I feel like, it's not as effective. And please correct me if I'm wrong, but I would argue that online courses just lack that human touch. I, I can't discuss with real human beings in person. So maybe like some, it I, there's a lack of physical space as well. So, creating physical spaces for education I think would be very important.

I love how the Greeks did it. They just had these like small communities of. thinkers just for the sake of creating virtuous people. And I feel like in the educational institution sense we can create these small communities and also reduce the cost. But I feel like that's a conversation for another day.

But yeah please, like, tell me about online education. What are your stats on that? Do, are people learning are they getting actual outcomes in the real world? Is it

[00:43:13] John Nash: Say, Jason's going to answer this, but I'll just say to you, Advait, that this is all Jason and I talk about, and actually our sort of raison detre is to humanize online learning and make it as interesting, and engaging, and effective as any face-to-face experience. Jason.

[00:43:33] Jason: Yeah. And from a research standpoint, we now, we have thousands of articles that point to the fact that an equivalent online program or online course can either give you an equivalent or outstrip actually the outcomes of face-to-face there's lots of research that actually supports that.

However, I do agree in terms of the so just like meeting outcomes, like learning your, getting your objectives met basically in terms of that learning. I completely agree. In terms of the, the personal connection, and this is one of the things that John and I talk about, which we really think is that growing edge of online learning for the, in the second half of life.

This podcast is called "Online Learning In the Second Half." Talk about, you know, we've been able to deliver content to people even have good outcomes for 20 or 30 years now, you know, and it's getting better and better. The media's pretty good. I like it from a, as you said, a democratizing standpoint that it, it provides access to people that normally couldn't provide or be able to access good education.

Asynchronously is ideal. You know, think about people who are working and they have kids, all these kinds of things, right? I love it for all those reasons. However, that growing edge is feeling like that you're really connected in a learning community and that you're connecting in such a way with the fellow students as well as your teacher, that it really does inspire you to learn and persist.

And I think that's the part, persistence isn't fantastic with online. And I think that's part of the part of the concern right there is that personal connection that helps students persist through the end of a course and through the end of a program.

[00:45:17] John Nash: So I would say Advait just to add to that, the picture you were painting, that about what your ideal teachers did and what would keep us away from the allure of an agentic tutor or an agentic tool coming in to do my work for me , and creating a community, it all comes down to learning design. The learning design that is created by, either the professor or the organization who is intentional about having those be the outcomes, those things we've been talking about. And so, it's really, it's just an active it's an agent, a different kind of agentic decision.

It's a, it's an active decision on the part of the humans in the room to decide that's how we're going to do it. And unfortunately, we don't always do it that way.

[00:46:02] Advait: Yeah. I would add on, and I think the only thing missing from an online education or learning perspective is the human, right? Like, it, it is basically a in-person teacher. It is basically an a course, like an in-person course minus the teacher. Because that's what the online section is, right?

Imagine if I just had access to a course with a bunch of videos, except an in-person one just has a teacher in front. think an online can do that with some sort of AI system that can understand all the pieces of data that's involved in the student's educational journey. I think in person it's quite hard to see, oh, is a student paying attention?

Are they watching? online, at least you have that data because everything's on the platform. And now on top of that, you can abstract that with this AI agent that as soon as you log in, it's like, Hey, how are you? Let's learn together. That's what are we having difficulties with? And it creates this empathetic, patient and caring environment for the person to feel not judged. I feel like a classroom can have kind of a judgy environment based on, oh, am I raising my hand too much or am I asking that questions? You don't have that online. And I think the best online teacher I had was Sal Khan. He's done an incredible job at Khan Academy, and it genuinely feels like it's my uncle teaching me 'cause. The way he explains concepts, the way he like brings in emoji, the way his like voice is so soothing. I can see the effort he put into drawing the certain diagrams he does, bringing pictures, creating the quizzes. It just feels like I, I would just not want to cheat even if I can. And I think what people are more focusing on now is how do we lock down browsers?

How do we create this secure testing environment compared with how do we make the student not want to cheat in the first place?

[00:47:53] John Nash: Interestingly enough, the research also bears out, Advait, that the students who don't feel compelled to cheat have a sense of belonging to their learning experience and to their instructor.

[00:48:05] Jason: Yeah. And as John said too, I think so much of this goes back to good design within the class, both designing the class itself, but you can design a class to be either more or less personable. Have more or less teacher presence. And we really encourage, at least in our, in my little world here, encourage to design a class that has a strong teacher presence and has a strong student to student presence.

And so that they're connected with one another. They're connected with the teachers in ways that they feel both supported. They're responsive, they're communicative, they show up. And I think for me, this is where the next level like. I think whether or not I'm providing the content or the knowledge for my students is neither here nor there.

But I think for me, and this is the feedback I've gotten from my students as well, when I really show up, they appreciate it. Like, this is what they appreciate online. And we've had really close connected, supportive groups where we've never, we don't even know what he, they don't even, they know what I look like 'cause I have videos, but I don't even know what they look like, right. But in those supportive environments, I think when they feel that, when they know the teacher is present in there, I think that's our next goal, I would agree.

[00:49:22] Advait: What are your thoughts on an AI clone? If you just on the screen, a chatbot and people can talk to it. It has your face and it just, it has some instructions. I think that would be pretty interesting. I'm greeted by. You and you're in all the videos, so there's a sense of continuity. maybe you can animate the students' profiles as well. That would be pretty interesting.

[00:49:45] Jason: Yeah. What I want is the AI clone that can put away the dishes, that can mow my lawn, that can do my taxes for me. Those are the things I actually want to connect with my students. Right. So, and that's where I would hate to take away too many opportunities to do that. I feel the same way about grading, actually auto grading.

I think that there's a first pass grading that I'm not against. Like do I really need to be teaching another student about where to use a comma? No, I'm okay with, some sort of autobot teaching my students where to place their commas and how not to have a split infinitive, but I want to be connected to them because I, believe that's where the transformation happens.

I'm still kind of a romantic when it comes to teaching. I believe that teachers can actually make a difference with students and not just giving them a head full of knowledge but actually seeing them develop in a way that their lives are better moving forward because of it. So,

[00:50:44] John Nash: so Advait, I think the direction of this conversation we've just had over the last hour going to surprise everyone that listens to it when they take the click bait and say, "Oh. Jason and John talked to the Einstein guy," and this has been a really rich conversation about the future of ed. And is it because you really decided to launch a site that would punk institutions?

[00:51:13] Advait: Like it. And actually, my goal is to start a school. I want to start a school someday inspired by the Greeks. Very small community, maybe 10 people, maybe 20. And maybe other people can take inspiration in the rest of the world to create their own communities fully run by AI and autonomous. And people can just explore whatever subject they want out of pure curiosity’s sake. and I'd be very interested in seeing the learning outcomes from that compared with like an actual educational institution.

[00:51:46] Jason: I think that sounds wonderful. And I think we need more of them. That's my opinion. Even though there's a way in which everything that we do, like when we're at these individual institutions, like John and I are, there's one way in which our institutions are in competition, right?

We are in the same area; the country we compete for online students. In many ways. But at the end of the day, for us that are educators we just believe that we just need more education. We need more people doing the good work and trying new things and getting out there to, to bridge those gaps and to help, especially people that, that don't have access to it right now.

And that access might be in different ways. Like it might not just be the hoops to get into a traditional university, but as we've talked about here, the hoops of learning in what seems to be a traditional way versus really capitalizing in the best way on just somebody's natural curiosity and just what can we do?

Like I think it was Plutarch that said, what could we do to do the real job of an educator, which is to light a fire under somebody, right? And just to make that happen.

You could use that too. Plutarch. I don't think Plutarch is trademarked, so maybe that could be your next. Uh, I would check though first just to be sure, but he is back far enough that I don't think that they it was pre corporations and stuff, so I think you'd probably be okay.

[00:53:12] Advait: No, I feel like I'm going to have to step back from that specific field. 'cause I've done my job. My goal wasn't to create a cheating tool. I like to say Einstein was a thought experiment except

[00:53:24] Jason: Well, you made it.

[00:53:25] John Nash: Eight days in, so let's pretend it's early days, but Einstein was a thought experiment. were the results of the experiment?

[00:53:35] Advait: I actually read a Substack and the person, I'm not sure of the name, but I'd have to find it, but the person was highlighting how there was a deficiency in AI knowledge among the general population. And how it created the knowledge, which in turn will create meetings, will, which in turn will create discussions and hopefully change. So, I, I do think people are taking this more seriously and they will realize that I'm not here to decide what happens, but my goal was to just show that this is possible and the people in power can decide what to do with it. And my, my email was full of deans and professors from different universities. It, it was either shame on you, take it down, or we're going to change your policies. So, hope that more people figure out how to deal with technology right now and yeah, and make the best of it because we have these AI CEOs saying, we're going to automate end-to-end software engineering. We're going to automate end-to-end jobs. We're going to automate end-to-end X, y, z. And that includes students as well, because students are going to get those jobs and they will be automated and the teachers need to rage at the automations instead of the student's job being automated or figure out a way to empower people with the automations.

I think that is the best outcome.

[00:54:59] Jason: Yeah.

[00:54:59] John Nash: you.

[00:55:00] Jason: good. Yeah. cause I saw that Substack. Maybe Michael Wagner. Does that ring a bell where he was talking about it? Yeah.

[00:55:08] Advait: There, there were a few of them and I made sure to read every single one and I just wanted to understand it. Change is happening, so that's great.

[00:55:16] Jason: yeah. Yeah. That's wonderful. Well, that's, I, that's really impressive. And you're an impressive person. I think that you'll go on to do great things would love to hear. We'll have you back when you start your school. The Plutarch School of AI Understanding and Knowledge into the New Millennial.

And

[00:55:38] Advait: my new school, so

[00:55:39] Jason: yeah, there you go. Yep. Yeah, that's what I'm here to do, you know.

[00:55:42] John Nash: Happy to license that to you until we send you a cease and desist.

[00:55:45] Jason: That's right. That's right. You'll be hearing from my lawyers after this podcast. Sorry. Not even funny. But thank you so much for coming to talk to us. I think this has been great both for us as a community to be talking about this and to kind of give us a little push to be talking about this.

But the fact that you're able to come here and kind of give a little more background, I think will be really helpful for our listeners at least. And it's certainly been helpful for us today. So, thank you.

[00:56:11] John Nash: Yeah. Thank you.

[00:56:12] Advait: I'd also like to thank you for being so curious. 'Cause I think that's such an important skill that most people don't have it, it's like an immediate rejection, whereas you guys are embracing it, you're bringing on speakers. So, thank you for spreading that.

[00:56:27] Jason: Well, sorry for my harsh words on the front end, and thank you for still being willing to talk to me.

[00:56:34] Advait: I've had enough of that. So

[00:56:36] Jason: Yeah.

[00:56:36] Advait: I feel like I, I can take anything at this point.

[00:56:39] Jason: Yeah. Well, I'll be more thoughtful next time. I'll seek to understand first, next time. So.

[00:56:45] John Nash: Advait, if people want to reach out to you, and chat further what's a good way to get in touch with you?

[00:56:52] Advait: Either my LinkedIn, Advait Paliwal, or my email Advait at companion.ai. Both would be great. The spelling is A-D-V-A-I-T@companion.ai. That's my business email. You can reach out to me, happy to have discussions, happy to help understand how the technology works, and maybe even help implement some.

And it doesn't have to be through the company. I just want to spread the knowledge and help people so it, it can just be out of pure. It I just want to help. So just pure goodness.

[00:57:22] John Nash: That's wonderful. And you've already got my brain running. I've been bugging Jason to give me a quick tutorial on how to get my GitHub going, but I think I'm just going to skip that and have you spin me up a computer in somewhere in the cloud. And so,

[00:57:37] Advait: I can actually do that so we can chat off podcasts and let's do it.

[00:57:42] Jason: That sounds good.

[00:57:43] Advait: By the end of this, you'll have a beautiful personal website, an app of yours and even some courses that you just made possible.

[00:57:50] John Nash: Heaven knows what I'm going to make now. Yeah. Okay.

[00:57:55] Jason: That was wonderful. Well, thanks so much and for those people listening, our website is online learning podcast.com. We'll put the show notes in there, transcript as well as links to Advait and any of the resources that we talked about this time around. So that you can check those out.

Obviously, we're always interested in your feedback and so reach out to us on LinkedIn. You can find John and I there and our links to our LinkedIn are right at the top as well of our notes.

Great to meet you. Thank you so much for jumping on.

[00:58:22] John Nash: To meet you. Thank you.

[00:58:24] Jason: we'll probably roll this out fairly quickly just because people are curious and I think it would be helpful while people are thinking about it, so,

[00:58:31] John Nash: Cool. All right. Take

[00:58:33] Jason: Okay.

[00:58:34] Advait: time.

[00:58:34] Jason: Stay in touch.

[00:58:35] John Nash: Yep.