AI Jobs for CS Grads in 2026: Why Startups Beat Big Tech

(soft music) Welcome back to another edition of AGI Schools out Saturday. Today I am joined by another boiler maker, Bo Hyun-Jang, his junior and computer science getting ready to go. I just finished in your junior year, right? Can I get a move in your year at Purdue University? - Yeah. - This is a Purdue household, both my kids are Purdue graduates, so welcome to the show. Tell everybody a little bit about yourself and your background, and then we'll just kind of dive into the state of AI and jobs through the perspective of somebody like yourself. - All right, thank you for introducing me. I'm Bo Hyun-Jang, I'm from South Korea. I'm an international student, and I'm studying computer science, but I'm also working on completing entrepreneurship certificate, so I'm doing CS and both business, next to the CS. - Have you been in any of Professor Woods classes through the Burton Martins School of Entrepreneurship? - Not yet. - Okay, she does a great consulting class, Susan Woods. She's been on the show a couple of times. I think the mixture of entrepreneurship and computer science is like a match made in heaven right now. I really want to go into what the job landscape looks like, what it feels like from your perspective. Like what are you hearing from students who are graduating Purdue? Are they worried that they're not going to be able to get a job when they get out? Or is that just the news media saying that? - I think there's two sides to it. People who are aiming for the big tech company, like Amazon, who are probably struggling more since AI came out and people who are willing to kind of look for different opportunities such as start up or small and medium enterprises. I think they're doing pretty good job. - I just have to tell my audience. I did not pay him to say that, right? Because this is something I've been saying over and over. It's the first Purdue student that said it rather than me singing to Purdue student. You are not wrong, my friend. Like if Metta, Amazon, all these big companies are not hiring right now because they want to show off how they're AI can handle entry level positions that may change in the future. But there are so many smaller medium size businesses who have no idea how to navigate this AI thing that they keep hearing they should be doing that people with data and computer science skills are like gold to those people. - Yeah, 100% agree. And one thing to add, there is mostly in startups and medium size companies. You'd have more agency and autonomy rather than working in big tech company. And as the AI rises, I think one of the important skill or properties or something that people needs to have more of is taste and judgment. And I think those can only be grown through experience and iteration and feedback from the expertise in the domain. So I think the combination of startups and medium enterprises and CS student who just wants to get more experience and get more hands on experience, I think it's a great place to start at least out of the graduation. - I couldn't agree more. And you gotta keep your finger on the pulse. Like I did a little experiment because I listened to a lot of podcasts, I listened to a lot of tech podcasts and individual contributors like a trending term for people at a lot of these companies, right? They're an IC. So I decided to add that as a job I was open to on LinkedIn. I'm not looking for a new job at all. I love the company I'm at, but I like to mess around with my LinkedIn profile once in a while just to see how much traffic I get from it. And when I put individual contributor in there, I got floods of post opportunities. And a lot of them are for people like yourself who actually have coding skills, like I wouldn't really fit them, but that is definitely one of the trending terms for someone who wants to come into a company and contribute with their AI skills and their computer skills. Has that been on your radar at all? - Yeah, 100%. If I'm. So if I'm clarifying your question or. So you're hearing more of the coding guys, the tech guys looking for people in domain expertise is that what you're saying or? - Know what I'm saying is a lot of the small to medium size businesses or the startups that are looking, like you have to know exactly what language they're using when they're talking about who they're looking to hire. And the individual contributor is the language that seems to be what a lot of them are hiring. So whereas previously it might have been, we're looking for someone who's a marketing person. We're looking for someone who's a design person. We're looking for someone who's a software engineer. Now they're starting to bend towards this individual contributor, which is essentially like saying, we're looking for someone with a skill set who's willing to work across different functions, leveraging AI and become a part of our team that's a lot more dynamic than just a job title. - Yeah, I think one of the terms that keep hearing is, for example, when the washing machine was first introduced to the household, they thought like the people in household was working less in the chores, but instead since washing machine increased the productivity in the house chores, it actually raises the floor. And then you're actually sometimes working more than ever. So I think that's what is happening is, when people have access to all these AI tools, and you don't really have to be one thing, you can be in different areas and you can have marketing, you can have sales, you can have development, website design, all the stuff. And I think it's more important to kind of focus on what skills are you willing to spend a lot of time and then you're actually enjoying it, and you can actually leverage AI properly since you already have some, at least the minimum amount of taste and judgment that you can leverage it. - I couldn't agree with you more. I think that is definitely the best place for anybody to start. A lot of, it's an easier sell in a company, right? You can actually go in and say, okay, what are you hate doing? Let's see if we can automate that away with AI, but if you're actually trying to get people to understand how powerful these tools are, saying, what are you love doing? What are you somewhat good at? Like marketing and writing newsletters and that kind of thing is kind of where I was working when AI came out, and it was playing with it in that arena was really what made me fall in love with it, right? If I would have tried to do something that was completely not within my skill set, probably wouldn't have known the right words to use or the right questions to ask or the right guidance to give the model to give a good response from it. - 100%, absolutely. And I think the minimum what people expect, how they use AI is keep increasing because information is basically free and you're overflowing with all the social media, like telling you different tips about AI. And I think it's not really hard for one person to implement new skills in terms of AI. And I think it's important to know how to work with AI and a longer project timeline. For example, the term that is trending right now is called Harness Engineering, which is basically you're creating a system or environment where you can keep giving feedback loop to the autonomous agents like AI to work on, reliably without going to a different direction and losing its direction. So you're giving us clear goal and clear set of environment that is secure to play around with it. And you're just giving it autonomy to work towards certain goal. So I think that's one of the things I want of the term that is kind of trending right now. - Yes, 100% and this is the thing that I think a lot of small to medium sized businesses don't understand when it comes to kids with either, or anybody with a computer science and a data analytics background is building those harnesses and figuring out exactly what company data you need to give access to these things to and how to maintain that data hygiene. Like these are all still very technical things that without them just launching chat GPT in your company is not going to be as effective as you hope and it's probably gonna make some mistakes that could end up costing a lot of money in the long term. - Yeah, 100%. 'Cause it's all black box, right? You don't really know what is happening behind the cloud code or behind chat GPT. You can only see the output of it and you're given that your specific input. So you can only play around and have some kind of loop of input at output since you cannot really look at the source code behind it. - Yeah, I think some of the exciting, well, I don't know if I would call eVals exciting, but eVals are something that is a skill set. I think it's gonna be very necessary and a lot of companies going forward. Like how do you evaluate the efficacy of this thing? How do you evaluate? And the other thing that people don't understand and don't really look to open source for that, if you have a business and you're trying to run a lot of processes, like my audience has heard me say, like you don't want quad opus 4.7, running everything or it's going to cost you more than a human being. There's plenty of times where an Amazon Nova model that costs like.05 cents per million output tokens is just fine for something that doesn't need a whole lot of thinking power. And orchestrating what agent loops and what processes are tied to what model and how much the inference cost is for each of those models. Like that is a whole that that's a whole ball of wax that could save a company a whole lot of money. There are going to be those who just lean into well, we let co-pilot make all those decisions or we let cloud make all those decisions. But when you're talking about smaller businesses and getting efficient use of these tools, orchestrating all that behind the scenes and then keeping up with it and figuring out how to evaluate it is definitely something that you need someone who knows what they're doing to do. Yeah, absolutely. And I think it's one of the part, especially in computer science is I think architecture is really important of designing the actual architecture behind it is because AI is some sort of brain and you need the backbone and you need the bone and you need all the muscle and you need all this skeleton in order to have that AI brain flourish and that specific environment. So without this specific architecture, you're actually falling into a pitfall of have you noticed the galos law which is basically a complex system comes from complex system will only work when it works from simple system. So if you don't really have the simple system made and properly validated, it wouldn't work to add it more and make it more complex. It would be harder and harder to debug and know what the root cause is because you used to have this technology and human, but now you have AI thing. So you have so many different things that can be like kind of like the stakeholder or the thing that is participating and building the system. So I think you got to be the foundation is more important than ever. That's awesome. I love the brain body analogy as well. I will tell you, I get jealous. I don't get jealous that often. It's awesome. I was always horrible at math, right? And so I was never really a computer. So it's awesome. I can build a website using clawed browser extension and it can write the code for me and everything. And I'm not complaining at all, but there are times I get jealous and when open claw first released, right? And I wasn't going to touch out with a 10-foot pole because I knew there would be all kinds of security things. But then in video, I went ahead and created Nemo claw inside of their Nvidia Cloud environment, right? Where you can use a lot of different models to do the exact same kind of thing. But I watched the demonstration building in there and I was like, damn it, man. Like you like this is just not something you could do if you don't understand computer science, at least a little bit. This is way over my head. I wish I had those skills. But I do think that it shows there are so much value to this in note for no coders, but the people who have computer science, like who really understand how those architectures fit together and what that type of thinking is are still necessary if you really want to build something that's going to be long term and that's scalable. Yeah, absolutely. I like I mentioned before the productivity, the floor is rising. So people who had no code can now build websites, but CS people already knew how to build website before AI. But now CS people are getting highlighted in terms of their skills and judgment skill in terms of building the infrastructure or the architecture. So they're going to the next level because all this AI thing at the end of the day, if you're implementing to a business, it should be sustainable and it should be reliable. And if you're just a non-technical guy who is using cloud code, you will always question yourself if this is the best solution or best architecture for my specific problem or specific solution. So what is your favorite model? What do you like to work with? When AI first came out, I thought it was chat-gb-t, but when I switched the clouds, now I'm not switching back for a while now because I work at VS Code, which is basically like ID where people code and there's like a terminal, I work inside terminal. And there's all this skills that people already built on published on GitHub that I can use. And frankly, like it's one of the most fun time to build stuff because the things that used to be not efficient 10 years ago can be done with so much little time and energy. That is why you can do all this stuff. For example, if there's public data or public, if you have a favorite person, let's say, I like the person called Neval, Revicant, and I like that guy. So what I do is, if I scrape off his website and then I make AI just put that in like the markdown file and then I make it has Neval voice. So I can basically chat whenever I have a problem, I can do slash Neval and I made a skill. So whenever I have a problem, I come to Neval, like the my version of Neval and I chat with it. And it's completely legal because like I'm not publishing to somebody. I'm not monetizing it. It's just public data. I'm just scraping it and using it for my own personal use. And that can happen. And you can do this for anyone with public data. You can just take their data and they didn't to cloud and cloud will scan all that thing and make like a voice MD, which is like basically copying their voice and tone and like how they think. And you can you can basically have like your own consultants that you can always chat on. And I make it question me. So whenever I say like let's say I struggle with working out and it questions me about the problem. And then depending on the answer I give, it asks me deeper question to understand deeper understanding of the problem. And then finally it gives me solutions. So I think those are one of the use case like outside of the technical part that I'm using. It's an awesome use case. And it actually can be done for a company or a person personally for marketing as well. So I've done the same thing with a bunch of stuff that I have written created a mark down tone voice and style file for that so that when I want AI to generate LinkedIn posts or Instagram posts or whatever, I can upload that mark down file or I can slash that skill and say write this in my tone voice and style. A lot of companies don't understand. You don't have to generate AI slop if you just say write a social media post into chat GPT and give them the URL of your website, then you make it AI slop. But if you've got a mark down file like that that actually encompasses your brand's tone voice and style, it will stay pretty. You don't have to edit it a whole lot and it will sound like you like you want it to sound like. Absolutely. It's also like I mentioned, you're creating that constraint or environment so that it stays within like let's say I don't like M dashes. So it's like, oh, don't ever put M dashes and like, you know, all of those like whatever that looks like AI slop, if you just have the eye to look at the detect the pattern that AI is using, just give that as a rule that do not ever use it. Do not do something. Do not do something. Do not do something. And it actually listens and whatever mistake is happening, you're actually, it's not actually bad thing for you that is making a mistake. It's the opportunity for you to fix it and then make it as a rule. So you're making it a better version for your personal use. Yeah, right now it's using periods and sitacamas like the brain period, the body comma, the, the, I'm like, it was M dashes for a while and I was, I so adamantly put that in all my system instructions that anytime any of the AI tools I use generates a prompt, even if it's a video prompt, it ends with, make sure you don't put any M dashes. It knows that I do not want M dashes and any of my stuff. And I'm sorry for the technical writers who like to write with M dashes, but that was the tell tale. And now it, it shifts. So you have to pay attention to these kind of things and update them. This is one of the things that I've struggled with. And I love the thought of this at you. Like, so if we're saving everything to mark down files, right? And we're saving all these skills, like I have all my cloud skills, my cloud skills library and I have them in projects. At some point in time, does that end up being like the old share point where there's just like a bunch of crap stored in half of them aren't even relevant. Like how do you keep that cleaned up so that you don't end up with a data store full of irrelevant skills that have been replaced 60 times? First of all, like I'm kind of guilty too. I have like, I guess like 15 plus skills. But the thing about the skills is it only pulls up when you call it. So for example, if you have prompt and you say, like they released a new feature called slash goal, that's what I like to use is I start with slash goal and it says. I say do not stop until you achieve these tasks below and it has all this task and I always feel like if you need Like use impeccable, which is one of like the design skill I use impeccable or Let's say oh my cloud. It's just another skill and then If you just like put the skill name inside of the prompt it only calls That skill that is inside the prompt. So it won't ever use the skill outside of the prompt So that is the thing that is keeping it kind of divergent and not really intercept with one another, but I think what matters more is like the context problem is I have something called primer MD, which is basically a markdown file and whenever the context window is over 65% I always say update the primer MD and what that what it does basically is it summarizes the Session you're working on like every what you complete it what you're working on what you have to do left and all the thing into a markdown file So you can do slash clear which is clearing all the context window to zero and then you just say continue from primer MD Then it reads the primer MD and continue with the conversations So I think that is one of more slot thing that you have to be messing around because if the context window is this big It only really focuses on the start and the end and it usually ignores a lot of the middle so whenever it is 65% Full you either want to slash compact which is basically compacting the context window or I like to usually do a slash clear I love that. That's yeah, and then clawed has started doing that some but I have found that it's if I if I If I Prompt it to Consolidate I will say distil a lot of times I'll say distil this down to To a summary doc so that I can transfer to another conversation as what I used to do before you had the ability to do it in the same Conversation chain have you played with claw design yet? I did I did all while ago, but I Mean I think it's pretty good to but I'm not really a big design guy in terms of Using the clawed claw design. I don't really use it that much honestly. I usually just use the skills within the terminal to like directly Mess up with the code because the thing is If I if I edit any design I actually have to implement that specific design Using clawed so I just like directly edit the code that is responsible for design so I don't have to work twice That makes sense. No, it does make sense. I think clawed code for people who are who we call I think clawed code work and clawed design are kind of that Intermediate step to help people like me who don't really know how to write how to write they or read the code to begin with But I will say that I love clawed and I've got a clawed account Chad Gbt is way better at aesthetics and Images so I find using the tools together even like when I designed a website. I would just say listen clawed I want you to Articulate the style and I'm gonna take that to Chad Gbt And it's gonna write the code and then you're gonna you're gonna tell me how to commit that to get up to publish this to publish this because Frankly, you're designed sucks and and it is so even with clawed design It it's just not as if you were doing something technical like if you had a technical website I could see that it would be very useful, but it is the aesthetic of it is not anything outside of Software engineer technical time. Yeah, I'll Understand. Yeah, I think it's because the root like the start of the Chad Gbt was Focus on the consumer base because like they wanted to be the next Google who runs ad but instead of the Google search engine They run it through their own like chat window and they were more consumer focused and but Entrafix which is still like that for the clawed they focused more on the company size So they worked on like they focused more on the code and those Technical side so I think that's the fun difference and Chad Gbt now is like oh, oh no I'm now missing out and then they're doing it the codex and I see many people switching also back to codex because Claude They have a thing they're using too much talking and Codex will last longer, but I haven't made that switch yet But I might try to use Claude and codex together and see what works Yeah, I think I think they both have their strengths. I will say everything you just said is a hundred percent sure of the business models and the way they approach things One of the things aren't me One of the things I think people don't realize oftentimes when you go after business instead of going after Consumer is that you really reduce the sick of fancy in the model because you can't tell Enterprise level business people who are trying to use your model that every idea they have is brilliant Because if they go if they go try that because the model told him it was brilliant and they lose a whole bunch of money They will stop using the model So I think that Claude has always been much more realistic feedback to people much more candid feedback Whereas because it was so consumer facing Initially Chad Gbt has a tendency to kind of tell you what it thinks you want to hear just to keep you engaged with the model Yeah, 100% I tried to put the I still remember like few most back when I used it a lot and always Like whenever I give them business idea. Oh, this is like best business ever and I'm like but when I talk to Claude it's like Sometimes I feel like oh, you're talking too harsh to me and it's like because Claude is like it doesn't mess around like sometimes It's just like how's your straight to a point one of the thing I want to push back is like xyz Yeah, I'm like, oh, okay, but I like it better because I don't I don't need somebody who just Cruced me then I'll never grow because no one will ever object to me But if I have Claude then it's kind of having my own feedback loop even with Claude And that is I think one of the things that people Might want to look at is because You got to ask yourself what are you without any AI tools? So like what can you do without any AI tool? So if you keep using this AI tool because it's fast and quick you get the output right away You can build anything and it would in the few days Those are the traps that people fall for but at the end It's actually shortcut to take the longer route of actually learning and growing your internally So that You're able to be in a position where you can orchestrate this orchestrate all this AI tools with better judgment and decision-making skills and decision-making Like frameworks or any like problem solving skills. I think you will only grow if you can Not be just replaced. You're thinking just all with all this AI So I think that is one of the thing people have to be careful because I think in elementary school I saw one of the pockets like people are getting dumber the students are getting the kids are getting dumber Because of the social media take-top raw blocks all this stuff And I think they're just Mostly replacing their thinking or not even thinking at all. So I want to share your thoughts on that Two thoughts a question and a thought so first off I That's really pragmatic advice and I couldn't agree more and and Claude Claude annoys me because it tells me to go to bed a lot of times when I've been working a lot And I'm like okay enough of that but I want to tell you a story of my like standout Level being leveled by Claude one night. I started thinking really hard about what it would look like if you trained a model Just on the best of humanity like if you could scrape the internet clean up the data and you just add like the Buddha Mother Teresa Gandhi like like if you if that was your foundation model and then you fine tuned it based on Pragmatic like what real world is like and so I ran this idea by Claude and it said Can I be frank and I was like yes, and it said You don't have a you do not have a PhD in AI And you do not have access to a billion dollars So I don't know why you're losing sleep thinking about something that you're never going to be able to bring to fruition Go to sleep and get up tomorrow and start working on things you're good at I was like I tend there too That hurt that now to your other point of dumbing people down like this is a question I want to ask you and I wasn't really going to get into your to your heritage But I there's a theory that I have that I like to run by you which is I think that Asian cultures in general Kids tend to grow up more using technology as a way to create things and build things and Americans Are youth grow up being entertained by or you as consumers? And I think that's the real shift that America is going to eventually have to make is how do we shift our Youth in general from a consumer mentality to a creator mentality as it relates to their relationship with technology I think the great question is I want to talk about habit because let's say you were eating McDonald's And then you order a burger and fries and a Sprite let's say you cannot tell them to just stop going McDonald's at once You got to be like oh you got to stop eating the fries Just eat the burger and the drink but just stop eating the fries and then you replace the Drink will like earn a die coke and then you start replacing with water I think those are the change you have to make is Technology I don't think banning technology within school or household would be ideal solution because People are living with it and even yeah, no technology within your household you will like face it anywhere else in the world and I think is Technology is just a tool But I think the user or the thinker of how they use the tool to do something is what matters is Simon Synex Find your why it's like the thing I just called golden circle is like the why and What and how and I think the right hemisphere right hemisphere is like controlling your why and like left hemisphere is Your what and how I might be confused but I think it's one of those and I think it's important to Understand what you want to do. It's like what's your purpose in life? It's like what you want to do in life and then technology is just a tool to get you through this milestone and Head towards the direction that you want to go. So I think technology Should actually be I agree with you in terms of like technology is be used to for the builders Because I think one thing AI cannot replace is Creativity Yeah, so because AI because creativity if it is defined It is no longer unbounded but since creativity is unbounded You can never define it and therefore you can never you can never have AI Be creative because AI will always be finite So if that makes sense so AI I think can never be creative It can act like it's creative by like mixing one word and mix it on other words and Making a new word and say it's like let's say you're coming up with a company name in my look creative because you're just mixing out Different words from the internet and creating a new word, but I don't think it's actually Really fundamentally creative. So I think that is the one of the thing that I want to mention is yeah I can now replace creativity So I think school should focus more on how to create the environment for kids to be creative even using technology Well, let's take that even a notch further. Okay, so this is one of the reason one of the times I started kind of harping on this whole idea of American kids need to be more creators than consumers if you look at One year ago today. I think it was June of last year China had Their first half marathon with robots. That was what was trending on social media. It was awesome I don't know if you saw there was like there were there were some bipeds that were impressive They were like little janky old star wars type fallen over robots right at the same time the AI that was trending the AI You so was trending in America was creating Italian brain rot Oh, yeah, you see the difference Like we're making up stupid videos of a teapot saying weird words in America to to doom scroll ourselves the death wall and China They're using this technology to create robots that will someday power factories and that like how do we how do we How do we become more exploratory as a culture here in America? So what this technology can do outside of create something to entertain me 15 seconds at a time I haven't fully read the book but I know that How innovation works is a book called how innovation works by Matt Riley is a thing he talks about Innovation in terms of innovation He term he talks about innovation in terms of the builder and also the government So what is actually government doing to Create this environment for innovation to happen like what incentives are they giving what programs are they making For this people to work on this project that is more creative and not really Consumer focused, but also like solving like real world impacting problems So I think that is one of the direction that you can have all is what are the programs and environment is government doing to Boster this innovation and obviously I'm not here. So I don't want to get too political of the things that I don't know But I don't think you even need to get political about this is because like no matter the parties here just Create an environment for people to move the direction that is more right or more Clear direction that you want to head on so there's no right or wrong answer But I think there's a direction that you can head to the more right answer So I think that is the one of the question that people have to answer is How do we make a place for like solar solar founders or like small companies to Stribe whether than having all this monopoly and big companies or just Consuming and absorbing all these companies and didn't getting bigger and bigger and whenever somebody's creative They just hired that person to and they're Becoming part of the bigger project. I think that is one take That's a really wise take I actually read breakneck which is about the difference in Innovative approaches between American and Chinese Societies and one of the things that they pointed out there which is really resonates with what you said is if you look at Underneath the the ruler of China if you look at the parliament or are there their government like it's full of engineers That is who makes them it makes up their government there different kinds of mechanical engineers software But they're engineers right if you look up if you look at Congress and the Senate and the House in the United States 60% more or more of them are lawyers So all that they really want to do is sit around an argue about and and let and legislate things to death right so what you're describing is actually What Silicon Valley was born out of it's big companies now, but initially when Silicon Valley started It was kind of a deregulated. How do we empower? Start up I'm Steve Lawsniak and Steve Jobs started in a garage when they started out They didn't they didn't start at meta as a $6 million a year software engineer from our Zuckerberg So how do we foster more of that? I think is that's a really really brilliant way to describe How to move in that direction? Yeah, and one of the point also mentioned in between met rightly which was the author of how we Innovation works and Naval ravicant and they did a podcast together and they mentioned like what is the Place on earth that is that is ideal for hub of Innovation and they said it's not Silicon Valley anymore. It cannot be India. It cannot be China But it cannot be just open space online because there still needs to be this cool thing like robber's drones all those things also need innovation so that was one of the question like where in the Planet Earth will be the hub for innovation the next hub. So I think that's one of the question That's a great question if it wasn't for the EU AI act I would probably point towards like Norway Sweden Finland that that they seem to have like culturally the the type of the type of the type of people and thoughts that I have but I the EU AI act I think is so over regulatory that they're really gonna hurt themselves in your Yeah, I heard about that too, and it's like whenever you build something I think you have to account for the EU user so whenever you're if EU user is one of the customers if there is even one customer that is from EU You got to like change the policy you got to change the terms of service. You got to change the privacy policy all those things and I think there's a right there should be a balance. You cannot go extreme on any size And you have to spectrum and I think it all boils down to the why problem like what is the purpose because AI doesn't have desire So when you tell something to do like unless it's illegal or unless it's like program to not do it it'll do it It won't ever say I'm not gonna do it unless there's something illegal so AI just doing things but who is orchestrating the AI the human right at the end of the day and Who is orchestrating human to move forward is like the why like What do you want to achieve like what problem do you want to solve so I think that is one of the question you Have to keep asking yourself is like why are you doing this like if you're building something fast cool But like why are you doing this at the first place? So when we look at it I couldn't agree with you more. I just wonder now like who are your heroes in the in the AI science world AI science work I don't know hero invest the second field Like do you mean heroes in general or a hero that I always learn my AI stuff like Elias chance giver Demis is so this you know Sam Altman obviously probably isn't, but if he is, that's fine. Bob Elon Musk, a lot of people, you know, set Elon's politics aside and just think he's, he's brilliant because of his vision for what leaving the earth and and populating space and doing data centers. But for me, it's Demis' office. I love what he does with Google Deep Mind and the fact that he's really focused on like curing cancer and doing a lot of things that are more pro humanity and pro pro life sciences outside of just like let's create a chatbot that we can monetize in businesses. Yeah. To be 100% honest, like I don't really have a specific person that comes out to the mind, but the mission, if any of the companies are doing something that it is focused on more than the 20, not the 80, it's like, there's like famous this 2080 rule, you got it, like the 20% of your work accounts for the 80% or what I don't really recall like that code, but I think the 20% is the value that what human can do. And if we use AI to help human flourishing, then we use that human flourishing to have more humans and have more AI orchestrated by those flourished humans. And then I think you have to cycle off like you have to focus on like what value you're providing to people first, whether than only focusing on the AI, because at the end of the day, I think AI is just a tool, really powerful tool, but at the end of the day, I think it should all lead to human development and human flourishing. That is, I mean, you're very wise for your age. And people are going to listen to this conversation they're going to say is this the typical computer science student at Purdue University, like I mean, I advocate, I think Purdue is the MIT of the Midwest. I really do. And that's not just because I'm a boiler maker, Fed. I think if people understood like what level of STEM science is going on, whether it's the drones and Sabrina Brunswick program, program to the Department of Defense contracts we have, and like sound based defense initiatives to building large language models, like all of that is absolutely phenomenal. And I will say my complaint previously up until this year was that the humanity side of Purdue was still kind of banning any kind of AI use in the classroom, while the STEM side was like leading the charge. And now that you're starting to see more cross campus like AI literacy as part of philosophy classes, it's part of humanities classes that aren't STEM related. What is the conversation amongst just your fellow Purdue juniors and seniors? Like what do you guys talk about when you talk about AI? Is it something you talk about a lot? Not a lot. Is there a lot of the environmentalists like, you know, the data centers are going to destroy electricity and water? What do those conversations sound like in social gatherings? I think it, I think there's two sides like because at the end of the day you pay tuition to come here and there's all these resources provided by university. And one of the resources provided is the actual class that you're taking. So I think it is, I think the conversation is happening on like AI cheating with AI. And at the end of the day, I think the course should be heading towards an exam because I think it's measuring in one of the previous classes, but I think there should be more explain your thought process. So at the end of the day, your thought process is what is commanding the AI. So you should like you mentioned like show your work in math problems. And I think those will be really important among yourself too. And maybe it'll be helpful if it was in the exam environment. But I honestly personally want less exam and more, I want more project based for creative thinking I think communication class is really important too because people are on their phone texting, but they don't really know how to verbally communicate with the like body language, like proper tone and manner. I think those communication skills will be more important. And I think people need to know how to be some sort of leader because no one's going to keep telling you what to do like in the future. There's going to be always you're going to be orchestrating either person or either AI tools or you're going to give command to something. And if you're not clear about your message and if you're not clear about the mission or the purpose, then it's going to the wrong thing and you cannot blame them because you're the one who is not clearly communicating to them. So I think communication class is one big thing. And the thing about the AI cheating is like people measure like, oh, why are you cheating if you paid all this $10,000 to actually like come to Purdue at the first place. So I think only you can be the one who's controlling the amount of the spectrum or extreme you want to go of using AI in the class. And I think there's always two sides of the pancake. There's like good sides where you're using AI as a feedback loop. This is what I did is I actually gave all the coursework and then I made this interactive HTML website. So I can like interact for hours and hours, like studying the exam without going through all the slide and notes. I don't like reading all the thing and I don't like like memorizing. I learn more through like interactive. And one thing I do is like I solve practice exam and whatever I get wrong, I review them, I stay like, why did I get this wrong? And then I ask AI to create a similar problem and that specific type of problem. And I solve it again and I get the feedback loop, which is more healthier version than just asking chest-chipity to like give you the quiz answer and you're just like, you're just async the quiz. But at the end of the day, you have to study them somehow. So you're creating this that of knowledge that you have to somehow later pay back. And I think that is one of the conversations that is trending right now among the students. >> Well, and I think that you brought two really, really, really good points, which is like, I didn't graduate but I did go to that other school down in Bloomington for like a couple of semesters when I was younger. Cheating is not new to AI. Like the people who are going to find ways to shortcut systems like AI just kind of amplifies abilities. The kids who really want to study are using it the way you just described and finding new or creative ways to have like an almost personal study coach or tutor that they don't have to spend $150 an hour for, right? And the kids that like to cut corners are they good luck figuring out how to detect that they use AI because they'll get that will put some energy into figuring out how to beat the AI detectors and like why play whack-a-mo like that. Why not lean into the technology? What I think that you said that it was more interesting is moving away from exam base to more project base. And I think that there's deeper layers to that just than the experience and the ability to see if somebody's learned something that they can't regurgitate using AI which is A, the communication has to be there when you're in a project, right? Because it includes more people than just yourself. B, you're starting to build a portfolio that you can then use to demonstrate what your skills are outside of class with employers. The soft skills that we talk about, my concern with the soft skills conversation which continues to come up and I've started to talk about this is like how do you measure those? We don't have rubrics for what is critical thinking? What is being a good teammate? What is being a good leader? What are those rubrics and are those industry dependent? Is it different to be a good critical thinker in healthcare than it is in manufacturing? Then it is in marketing. So I think that project base work can start to uncover what that really looks like in different settings with different, I love the scrum. I don't know how familiar you are like agile and scrum project management. But scrum essentially is like a bunch of different people who all have certain skills and whatever, wherever we're focused on in the project that you're the subject matter expert in, you're in charge of that one time. So if you're a marketing dude or we're working on the marketing stuff, you're in charge right now. If we shift to finance and I'm the finance guy, now I'm in charge of the project. There's one overall kind of project orchestrator, but more or less it's a democracy where we're leaning into our skills to lead in that. I think a lot more of that and then being able to demonstrate how that impacted your education when you go to the workforce would be very useful at a post high school level education. Yeah, absolutely. I couldn't agree more in terms of project and communication. I think sometimes you got to with this, all this confusing and overflowing information, all this internet you have now you have like multitude of information after AI came out because there's AI generated blogs and all over the place. How are you going to actually navigate through the like sea of information is I think one of the things that is really important is having the ability to think and one of the things you mentioned is like there's no rubik for critical thinking and there's no rubik for those. And I think that's why people should read more than ever because you're actually thinking through somebody else's lens without actually having to experience the same experience. And you said if the critical thinking is different in healthcare or different industry, I think there is no exact way to replicate somebody's thinking to solve the exact problem because every problem is unique therefore every problem needs unique solution. But I think there's some models because they say like all models are all but some are useful and if there's some frameworks that people can use to navigate their thinking and if they are keep replacing their thinking with AI, it's going to come back somehow because people who read are going to have compounding knowledge that is keep growing and then people who have know how to communicate will have compounding network or human relationships they're growing. So if you're using AI as the captain of the ship instead of tailwind, then you're going to have to pay back somehow and it's not going to look pretty. So one of the thing I mentioned is AI just amplifiers and I call it tailwind. It's just helping you go to the direction that you're heading, whether that is wrong direction or right direction and to navigate through that you have to be the captain of your own ship instead of replacing AI because sometimes like you've got to know how to be independent on solving problems and that is when critical thinking and problem solving skills comes in and you cannot replace that. What is such an amazing advice? Do you have an internship going this summer? I work with Hive which is called for innovation venture entrepreneurship which is based in engineering and produce and that's what I'm going to do and I have a project working during the summer but I don't have internship yet for the summer. That's all I'm working on right now and I'm really focusing more. I'm actually in West La Fiet for the summer so I'm going to focus on meeting people locally and then meeting new cool people and try to grow my network as well and grow my communication skills, all those self skills that I'm trying to grow. I think that's what I'm going to invest during the summer. That's amazing man. I think you have a very very bright future in front of you so I wouldn't even have to ask you questions because you're answering them before I even asked them and I think you definitely have a bright future ahead of you. If you had a recommendation for the student who doesn't have computer science skills and who's looking at this AI thing like what would you, what recommendation would you give them right now? I say first try on your own before asking any help because I don't like people who come up and ask for help without even trying on their own and then after you try it on your own just go to somebody who is more technical by them dinner, by them food and then actually just go to like core working studio or go coffee together with your laptop and actually ask them different questions. If I was having this problem like how would you approach this problem and instead of telling you what to do you're actually going to ask different questions and you're going to learn through that process because what you want to solve might be different than what they want to solve and there's different use case for different AI. So you're just going to ask them which direction I need to head on then and how do I research on my own so you're going to basically learn to be independent researcher and technical faith and I think the best way you can do is have someone close to you or meet someone with more technical and just ask them because as much as technical people might not look too superficial like they like to brag their knowledge but you kind of just ask them and then they're more willing to be like oh how did you do that and you're just just kind of frame it well and then they're they're going to be willing to help you. That is great. I actually when I joined this company so I'm like the only non-technical person at Citizen AI. Most of our people are out in San Francisco, they all have software engineering or data analytics background like they're building the technology and it's interesting because the first three years that I was diving into kind of no coding and figuring out AI. Most of the other people I talked to were people like me who were who were not very technical backgrounds and were trying to navigate this up and when I started talking to people who have those technical backgrounds I did find totally different questions I needed to be asking. I found things I wasn't thinking about that they were and it's a great exercise because you don't really communicate the same way a lot of times different people with different. I have a very marketing and very literature. I love philosophy. I love psychology. I like I think in words a lot right and math again math, graph, spreadsheets, all that stuff is like a foreign language to me that puts me to sleep so like learning to communicate because the world a small team in the world that is to come is going to need both of those kinds of thinkers to be able to communicate with each other and figure out what problem they're trying to solve for a business use case or for the world and who their end user is and how they combine their skills and then transfer that into like you said getting the tailwind of AI so that you can amplify what you figured out together as a human team. Yeah absolutely. I think I love that you mentioned about asking the right question. I think that's all it takes is I'm not a biologist so somebody with the degree of in biology will ask better question about the topic and biology than I do. Even asking for asking in cloud or chat to pt. I can ask like what type of what type of butterflies is and they might ask like is this butterfly this type of butterfly or this type of butterfly like they might ask like deeper level or like next level question and I think that is when domain expertise comes out and getting back to the question you ask like how do non-technical people be more technical and I think that another question should ask is what skill do you have that technical people cannot replace you so instead of like try to it's important to like address your weakness but I think it's also important to think about what your strengths are that is non-technical and kind of amplify that it's kind of like puzzle piece like there's no perfectly matching puzzle piece between like small teams there's not gonna be a perfect team mate but I think the good approach for non-technical people is to at least like learn at least shadow level of technical space so that when you're communicating to them they can better understand it and if they see that you're using all this technical terms that you didn't know before they're actually thinking oh you actually put effort to actually research and you actually put effort to like kind of kind of sacrifice your time and energy to actually learn to fit the puzzle so they're more willing to help you and on the other hand technical people like me should learn to have the self-skills like communication skills and all the skills so that when I'm talking to marketing my product marketing the technical product that was built for the product to market for a business then I will have better understanding of the marketing space so I can communicate better so I think you should go both ways not just non-technical people learning to be technical I think it should be technical people learning to talk to the non-technical people. 200% agree I think you know if you looked five years ago before Chad GPT launched before we got into this kind of artificial intelligence error if you were listening to McKinsey or BCG or any of these big consulting firms they were starting to say that the HR director was the CEO of the future for a long time it was the finance director the chief finance officer was kind of the person that most companies looked for to move into CEO and right before artificial intelligence it was leaning more towards the HR director because they understood human factor engineering and how to motivate the workforce now that AI has launched the chief technical officers are really the guys that are going to be probably running companies but they will have to develop some soft skills in order to do that I love what you said about people who aren't technical leaning into the skills that they do have that can't be replaced by technology. And I want to run something by you that I've kind of been thinking about on the flip side of that. So I'm not saying that people who are who are who are introverted, extremely introverted, extremely nerdy smart like science, science, sciencey people shouldn't learn more soft skills. But one of the skills I think that they may not understand they bring to the table is if you go outside of computer science or outside of software engineering into the like main workforce. Most of the processes pardon me are not mapped. It's all just in people's heads. It's real messy. Like if it does not, it cannot be articulated properly to an ML or an LLM model in a way that it understands and can replicate it. And so the people who think like that, whether they have social skills or not, have some serious value going into a company and saying, tell me what you do and let me translate like your natural language of blah, blah, blah into, okay, this is the process. I hear you defining. This is what we need to put in a document so that our artificial intelligence systems actually understand how replicate what you're trying to do. Does that make sense? Yeah. So, if I want to clarify, you want to, the technical people needs to be able to transfer the mission or message that non-technical people want to like build or have had tours, but they want to translate that natural language, let's say English, to the way that AI understands it better. So, you're not losing the meaning along the way of the translation. Yes, and not only the mission in, but the actual processes, like most people who do something, let's take a nonprofit, right? A lot of people in nonprofits wear lots of hats. So if you were to say ask executive director or the fundraising person of a nonprofit, like list what you do on a daily basis, it would be very difficult for them to write down, step by step, like how to do their job. So there are, there's a lot of companies, especially outside of enterprise that don't have documents that say, these are the actual procedures that have to happen in order for you to do this job, right? But all that needs to be documented in order for machine learning or large language models to be able to replicate those processes. You can't just say, well, you know, sometimes I come in and I write a few emails depending on how I feel, maybe to some people who I had a conversation like that's way too vague, but that's how a lot of people who don't have data or computer science backgrounds think and talk and articulate. So you've got to translate all that into what does it actually look like to operate this company instead by step procedures that we can feed to a machine so it understands what we're trying to do. Yeah, I think that makes more sense. And I think one of the things is computer science students used to have to use this big technical words to prove a point to the professors who are more expert than them. But now when they're going to the workforce and you say all this big technical stuff, it's like, bro, what are you talking about? Like, we don't use that big technical, technical words and jargons. You got to like, dumb it down. But you don't want to dumb it down too much. You got to like keep it concise. The word I like to use is like dense. You got to keep your words dense. I like, you got to dumb it down. You have to make it simple enough so they can understand, but dancing up. So all the details are there and as simplicity. So I think that is one of the things that computer science students or any technical people needs to do. They got to dumb their process down and then they got to ask better question to the people. So they might not know like when they're like asking the question, but if you ask like better or more detailed question, then you're going to start to like kind of try to understand their process by asking better questions so that you can help me help them better to transfer that to much in language, lotals. You're so right. This is like a real time example of how valuable it is to get two people who think differently about things either because as you're saying that I realize, yes, you're 100% right. You have to you have to learn how to articulate things in the right language for people who don't have a computer science on the flip side of that. If you don't have technical skills, if you learn some of those technical terms, you can get an LLM to work a hell of a lot better for you. If you use the three terms, it was trained on whether you deeply understand what you're saying or not like, for instance, I said distil this conversation. So if I say summarizes conversation, I will get a lot less value of what I'm looking for than if I use the word distil, which is the industry word for what happens when you take a big model and crush all the valuable stuff down into a little model, right? So it kind of works both ways. 100%. One of the quote I like from David Duitch has one of the businesses. It's a good explanation as hard to bury, which basically means that good explanation is hard to bury because every single detail has a function or role. And if you ever add or delete any of the details, it will lose as value being a good explanation. So I think the details matter a lot and I think you can only grow the taste and judgment and sense of spotting the details as you get more experience. So we circle back to the conversation of how the graduates should not only look at the big tech companies, but also go to smaller companies and medium-sized companies so that they can grow like hands-on experience so that they can grow taste, judgment, problem, solving skills, creative skills so that it cannot be replaceable by AI. Totally. And look, just imagine this. Like let's just say I have a 50 person healthcare company, right? And we have an IT guy and I've worked in that industry. So I know your average 50 person healthcare company has an IT guy. His main job is to make sure that you don't violate HIPAA within either the third party software that you're using and that if anybody has an issue, he helps them debug it, whether it's to get back into their email or the CRM is not working properly, right? So now you come in as an undergraduate and this company is starting to explore AI, right? That IT guy who's probably close to my age, he's probably in his 40s or 50s, may or may not have very much experience with actual large language more like what you're looking at, Gintake AI. All of a sudden, you're this undergrad who's 24 years old coming in an entry level position and every time there is a leadership meeting with the C suite, what we call, you know, like the CEO, the chief operating officer, the chief finance officer, and they want to talk about whether or not it's plausible to do something with AI, guess who they're going to bring into that meeting. Yeah, absolutely. I think it's more vertical to this. It's not horizontal. You're not going to be go far from life if you're just a shallow expert in like so many like hundreds of fields. You got to learn how to like grow your roots into deeper and it might not be visible at first, but you know it's there internally. And I think that's one of the thing is, if you're technical, you got to like, really go deeper roots into like architecture and security and all this system and progress so that when you're facing specific unique, unique problem, they cannot be solved by generic AI. You have to go another layer. You now have expertise and experience to handle that because you know how to build those architectures that is fitted to the specific problem and solution. I tell you one thing your generation has, I think it's kind of cool and not very many universities are offering it yet, but red teaming as an actual like degree. Like in my day, if you were a hacker, you like went to prison like that was you didn't have a lot of choices as an actor, maybe the FBI or Department of Defense. If you were like in the top 3% of good at that kind of thing would hire you to help them with their defenses. Like now that's actually a major at some universities. I love the new doors that are opening for people in the computer science realm, whether it's evals, whether it's red teaming, what's the opposite of red teaming? There's a blue or green, but cybersecurity should be huge, right? I think it's also like playing deep things and all of that. It's like one of the companies saying like, oh, this next model is too powerful. So we're going to only release it to the white hacker, which is like the good guy, another attacker. And then we're going to build a defense first. So when the AI is the new model's release, the offense, offense people would like to have bad intention with using the AI cannot like destroy the defense. So you're gonna build a defense first. So there's always like two sides of the pancake. So, AI is just a tool and depending on your intention and the purpose, like it can go either way, it can go to the bad or good or like some kind of mediocre. But I think you just got to learn how to accept it to the specific environment. Then if you're not regulated, you got to know how to accept it first so that you know it better. So you can control it better. So instead of just rejecting certain, and paradigm, like for example, like in school, like instead of just 100% like rejecting AI, you got to like know how to absorb it and like use it in like a certain manner and certain constraint so that in the future when everyone is ahead of you, because they have already like when some school have already like, like put it into their system of learning, like their way ahead of you when you were just rejecting it, because you have to somehow accept it in the near future. So why not do it fast and learn through the iteration and mistakes so that it's the same mistake. It's not gonna happen with this rules and regulation and like certain constraints that is always needed. So I think it's always the balance between the two. All right, let me ask you one last question because I've had you on for quite a while now I want to give you your mind back, but. AI friends like do you have any do is anybody in your sphere of of students to do they anthropomorphize AI to the level of thinking that it's their buddy or their girlfriend or their do you see that. I think most people I think it's whether they do or not I think it's whether the degree of it. So let me explain this sometimes people are feeling like oh, chatchipity is like my homework body. There can be degree of that and oh like chatchipity is my, I don't know more romantic partner. I think there's some some people and some guys that internet who is into the AI romantic partnership and. I'm not really familiar to it, but I know there's people out there who does that. So I think it's like degree I think it's like more like a spectrum right some people might just use it and that say thank you. And then is that is that like treating as human maybe not in an extreme level of like having as like romantic partner and like thinking like you don't need to socialize because you got this AI thing like me personally. I like to use it as some kind of people who have to have conversation or debate and different like feedback without having to risk somebody else's time because it might be something stupid anyway. So like I might just like talk to AI because like I don't want to waste people's time talking about stupid stuff. So I talk to AI so that like I don't have to waste other people's time I can just like have fun like talking to it and have this pink ball of like cable tennis like pink ball of this conversation so that it is more. Kind of like way to validate your thinking you don't want to validate. I wouldn't say validate, but you want to test your thinking in different perspective just by talking to a. Yeah, it totally lines up with with my theory and my theory going forward is that everybody needs to have from kindergarten up age appropriate like light understanding of what's going on over the hood because I could not agree with you more like I how I laugh at Claude sometimes like I and it is like kind of like a coworker but not to the extent that I think that it is live at all to the extent that I'm delighted by the fact that it is so good at replicating natural human language that I can't really understand. I think human language that I find a humorous or or uplifting right and I think that the that the I've never talked to a computer science student who has a problem with anthropomorphism I think the danger comes when you have absolutely no idea how this architecture how a transformer architecture and large language models and predicted next word prediction like when you have no idea of any of that it can seem especially to a 13 year old a 12 year old like. Because it is so so good at replicating natural human language that there's something more inside the box than there actually is. Yeah, 100% and I think the problem also comes from that people are not really empathetic anymore in terms of in relationship so they're using in Korea we call it. Feeling trash bin which is basically referring to the person you keep like during away the words like you're just being angry at somebody for your own problem and you're using them as like a feeling trash bag and usually happens in romantic relationship when like boyfriend leads it to the girlfriend all this problem and you're just like kind of. Being the not really a therapist but like you're gonna absorb the negative feelings without like gaining anything from it other than like keeping this relationship so I think many people so like. Comes up with their personal problem that people are not going to really care outside of this a I think and they're just going to throw the problems at them and I will be like 100% and then there will be they will listen to you and I think this is the problem of like there's more and more number of increasing number of. Only people so I think we got a kind of tackle that issue if you don't want this extreme level of talking to AI more than talking to human. Wow that's that that is sage advice that I want to end there because that's you shared so much grace of tonight man I hope you'll come back on the show sometime. Yeah thank you for inviting like that's my first time in podcast so I don't know how I did but it was at the end of the day it was great talking to you and I'm glad that we have that conversation in matchbox that left to you. Yeah man I listen you did produce proud and yourself proud and your parents proud like you've got a great head on your shoulders and some really pragmatic sage advice on how to approach AI in a way that is going to benefit humanity not just entertain us for the next 10 years. Thank you so much. That's not real looks like this dude's riffing on 11 laps is this guy using 11 laps and I wonder what model he'll work with next week. Yeah right that wave.