AI Engineering 2025: Bestselling Author Reveals What’s Next

Hello everyone I am so excited to interview Chip Huen.

she's the author of this book, and she has her new book, AI Engineering just came out. I'm looking forward to receive my copy of,

her second book as well. if you don't know her, she's a computer scientist from Stanford. She's a bestselling author. She has two very amazing and very easy to understand AI and machine learning books.

The ones I just mentioned, She's AI startup founder herself, with a successful exit. and she's one of the most influential voices, shaping, how we build, deploy, and think about AI today.

So let's dive in.

Happy New Year, by the way.

Yeah, but I'm like, waiting for the little new year, I want,

me too.

want to get some candies.

yeah, all right, we can get started. uh, tip, really nice to have you on the podcast. we can start with my first question. [00:01:00] Yeah, are you currently building something new and exciting and can you tell us more about it?

Hey. Yeah. Thank you so much for being on the podcast. I really appreciate the invite and hi Angelina and hi everyone. my name is Chip. so currently what I'm working on is that I'm just gonna be doing something, but I feel like there's not much to share yet. I was looking back previous years, and I realized that in many of the previous years, I have a list of to dos.

I have to do this, I have to do that, and this year, I was like, let's just try to pursue my interests and see where it goes. This interest take me so I have a few interests in mind and let's see how they go.

Sounds good.

we're looking forward to hear like more about what you're working on But I know that you have a new book AI engineer just came out

and um, I love your first book So I'm looking forward to get your second book. So looking forward to that

Yeah, I'm really happy to be done with the book. I feel like, writing a book is a little bit of a long process, and you don't [00:02:00] really know how you want to pursue it. we're going to do a startup, right? It was like, again, launch early, and then iterate often. But a book is not quite like that, right?

You finish a book, and ooh, I'm done. I think I tried to do that, with, I tried to follow, launch early for my book, so I sent a lot of, early drafts to a lot of reviewers, and I, tried to share what I can on some of my blog posts, and just try to iterate, yeah, but I'm really happy now it's out, and I'm a bit, I'm a bit nervous about, like, how the world is going to perceive it.

Yeah. Congratulations on that one. I'm curious. I know that you build your company and at the same time, you wrote two books. and now you're done with your last project, your last company. you are also exploring thinking about what to build or what to work on next.

how do you find directions? like what to work on from one project to another, especially like after an exit. Do you have a, like a thought process that you go through?

I think deciding what we want to do, I think it's a pretty difficult question. [00:03:00] It's just I think one of the things that I thought about after college, it's just like in school, you have a very clear, structured way of going about your life. You do this course, you have a credit, and then you graduate. But then I was terrified, so after graduation, it was like, oh my god, what if I, Make really bad decisions, and there's no one to tell me I've done something wrong, right? So I have a few pipes with friends, and then we try to keep each other accountable. We try to talk with each other, like a cruise, like, I will usually make a process. And this is over the years, you get a better sense of life, what you like, and what you don't like, which also help you. Decide better. So that's one thing. first of all, one thing that I realized running a company is that building a company is a very different experience from building a product. So video company has a lot of operations, a lot of people, punishments, recruiting, talking to customers and. And I think it's like I realized that I like building a product, but I don't quite like a [00:04:00] lot of those operations. so this year, I hope that I can focus more on building products. I keep getting small. I do think that,one thing that AI is like people have been talking about is like, um, it has, you know, Being able to automate a lot of the operations aspect of a company, for example, like before, right? If you want to build a company website, it's like quite a lot of work, but now you can actually get AI to do some, small work for you, or if you want to design a logo or do some, copywriting, I think it's like, so I'm very excited for, like, uh, to see, a new generations of companies, extremely lean and be really cool products.

Yeah.cause you are an entrepreneur as well. was there any dark moments in your entrepreneurial journey? if you remember

when was the darkest moment?

You asked a really hard questions. Yeah, of course, of course. Doing a company is hard and I have a lot of respect for founders out there. I think like the experience did pay off. Give me a [00:05:00] new perspective. I think I did my company for pretty short, like a little bit over two years. so and I thought it was like really hard and I see people doing it for 10 years, 15 years. And I was like, wow, these people are amazing. So I did, there were definitely difficult moments and one thing that I talked to my mentors about a lot is that like, so one thing it's like sometimes when you go through like difficult times, right?

Like you don't know whether you're going through like, you know, I know we're going to a cave. So a tunnel is okay, you go through the dark moment, come out in the bright side, right? So you should keep pushing, but even in a cave, it shouldn't be pushing, right? Like it should turn around. So I do think it's like, there's a one moment and it was hard to determine whether it's something I should push through, or is that I'm just like, not going to get anywhere, no matter how hard I push. And I do think it's like, In, entrepreneurship, a lesson we are taught a lot is just like you keep on grinding, keep on pushing.

But I do think a very important lesson is like knowing like when to stop and when to turn around.[00:06:00]

Totally. We don't know what we don't

know, If I had, we have the crystal ball. What's the point? Who have bought Bitcoin and NVIDIA stocks, I know. Yeah, exactly. Oh, darn

I have a question about your books. So

what, drove you to write your first book?

Oh, so designing machine learning systems. Yeah,

Yeah,

yeah, So, um, always, I've always really liked writing. So, um, You know, like, see, I think that if you're upset and if you do it, it makes you feel better. So writing has always been like that for me. So it's like a kind of therapy. So I always make a lot of notes. So for the first book, I didn't really set out to write a book. It was just my writing. So I was teaching a course, a Stanford back then is, of course, it's like on a machining system design. And I was pretty conscious of my accent, so I was like, Oh, my God, what if my students don't understand me?

So I just try to, write a lot of notes so that if they didn't understand me in lectures, they could [00:07:00] reference this still. Um, and then I had notes for the course, and it's like, maybe you can put them all together to become a book. And that's how the first book came about. Yeah, it was like very accidental.

Oh, actually, like before that, I should like also publish a note. It's probably 8, 000 word note, uh, on like, just like some learnings. Um, so also like for my first job, uh, I was at NVIDIA. I learned a lot of new stuff. I was like, hmm, let's just try to like,

for me, like writing is a way to organize my thinking as well.

So just like try to make note. I didn't really think much about like how it would be perceived. Um, so is this. Posted there, and then we got a lot of good feedback. So, like, so those feedback help me, like, improve and, like, learn new things. And then the course, and so learn a ton from my students. Um, yeah, and that's how designing machining systems came about.

I see. Yeah, it's very interesting experience, but it makes sense,

did it happen in a similar way for your [00:08:00] second book?

so the second book is a bit different. so when Chachiviti came out, I'm not sure how it felt. It was a group chat with a few of my friends. We was just like on not quite surprised, but have a little bit of like existential crisis. And the reason is We weren't surprised by the progress in the capabilities of AI.

if you have been following the language modeling progress over the last, I don't feel like 12, like, since the AliceNet, right? I've seen the language models are getting better and better. So, we were surprised. Personally, I was surprised because I thought that. A small improvement in AI capabilities will lead to a small like number of new use cases, but it's actually a small improvement in the abilities of AI somehow But if it's a tipping point where a lot of new applications became available and the possibilities were like both very exciting, but it's it's a little bit like scary because we're questioning of okay, what that thing could be automated and what kind of thing going to be [00:09:00] automated, right?

we saw that a very good coding. And then we also saw that it's really good at writing. And what am I right? I feel like I'm both a writer and an engineer. And I was like, okay, so it was, I spent a lot of time trying to understand, Yeah. what kind of like engineering can automate and what kind of writing can it automate.

So, yeah, so it's so I interviewed like a ton of people and then I asked a lot of questions and then I had a lot of notes and, and then, yeah, and then a big question people ask me, I asked myself is that, It's not time to write a book because the space is like changing so fast and it's quite scary.

and yeah, so I had to figure out the questions

by the way, do you use AI to help you writing at all?

I usually do use AI quite a lot. so I use, both Chachapi and Cloud. And also tests are like Gemini and so use a bunch of I'm doing a lot of experiments. I do a lot of with open source model or trying to fight to an answer, help with a few companies just like casually,not. No, no real responsibilities, just like, how about when I [00:10:00] can,so I get a chance to like, find you in a few models and just do a bunch of evaluations, but for the book, specially, so I actually did some fun experiment. So I just exported on my chat to PD conversations, and I found I was like, during the process of writing the book, I had about, 3000 conversations.

So, so, yeah, so, so it definitely helped me a lot. And, and I, what he helped me with is like, uh, first it's like, uh, help me read papers. Um, so the process of writing, reading paper before I was like, okay, read the abstract, read the conclusion discussions. And then I, it went through, like, skim through the paper, look at the charts, like, grab the results, right? Uh, but then the transcript helped me, like, to understand the paper a lot deeper. So I still read the abstract, I still skim it. I like presenting some, like, art questions. One thing I very, I felt very useful. It's like, for me, uh, to me, to understand a new technology, they want to compare it to like, there's an older technique. So, like, if they are like a new technology technique, I was like, get the paper for that new technique and then it gets a paper for the [00:11:00] previous technique. And they actually should be like, hey, so how is that different? Like, what? And why is it change? Like, sometimes, like, Just you just feel like if I can understand more, um, yeah, it also helped me come up with, like, um, different ways to define some things like definitions, like, refine my sentence.

So that, was easier to understand. I also come up with examples. I think it's just like, yeah, it's pretty good. It's like, just brainstorming a bunch of examples. and, yeah, um.

That's true.

Thank you. Yeah,

do you think one day AI will replace human writers? Do you think you will completely rely on it?

I think it depends on what kind of writing. So I do think that's a certain type of writing that AI is pretty good at. for example, if you wanted to explain, what is a new, what is something like, defy something, right? it can come up with, a lot of different definitions, explanations, maybe, As if you were five, or as if you were a writer, or as if you were a grandma, so it's very [00:12:00] useful. So I find that, AI is generally pretty good at answering common questions.what is hard, though, is just, coming up with the right questions. like, how do you ask the right questions? So that made me think about, maybe as a writer, I should focus on, the standings of right questions

and also questions.

Yeah.yes, it's not just from engineering. definitely. Definitely. But understanding,what are things that. It's important to know.

how do you keep updated, like in this fast moving space? You mentioned that this is like

terrifying, exciting, but terrifying. How do you keep yourself updated?

so I would say So, um, the person who's up to date with, like, on the latest pipes or news, or I don't really read a lot of news because I think the definition of news is like, it's new. So it means that it get old.

Quickly. Uh, so it's something that lessons and learned. Uh, so I was working as [00:13:00] a, when I was working, um, in high school and working as a reporter. So I understand news, like new cycles, very, very, very quick. And you keep trying to stay up to date with news. It's just like a never ending. Like, it's very tiring. So I talked to a few friends and like, they were telling me, it's like, Oh, are you going to start like a newsletters? And I was like, Like to do what, like, if I don't want to be, if, like, if you commit to, like, be the person to stay up to date with the latest AI news, you're going to have a pretty intense life. Like, yeah, um, so, so, so I, I focus more on, like, a few things. So, like, first, I focus more on, like, interesting problems.

These are things that I'm interested in, and I try, like, if there's something coming out, I would try to say, okay, it doesn't help, so don't make me a problem or not. If it doesn't, then, then I'm just, like. We're free to calm down and then see if it's my stay. Right. Um, another is I rely on friends like way, way smarter than me. So it has this heuristic of like, if three friends, like, smart friends mentions the same thing, I would definitely look it up.

[00:14:00] Um, and then the another something is like Lindy's law. It's pretty interesting. It's like, it's saying that the future lifespan of something like an idea or a new technology. Can be assumed to be equal to the how long has been a wrap. so so I find it's like useful. I think actually I got that advice from a professor in college. It's about relationship, uh, because like part of it is studying for someone for like a week, right? You probably don't want to make like a plan with a person a year ahead. But if you've been with someone for four years, then maybe it's okay to make a, to make plan, like a year ahead because it's a future life of that relationship, maybe like another four years. Right. So, so I find it's very useful. And in the process of being researched for a engineering, I also found out that like, even though a lot of new applications are like new and exciting, like the funny blocks of like models and the best engineering practices have been around for a while. So, like, for example, like, language [00:15:00] modeling was introduced back in nineteen fifties, or a lot today. A lot of us talk about rack. They love. We use rack, right? Retrieval, augmented generations.

So we should be used based on retrieval technology, which is not new. So retrieval technologies have been powering. Many Internet applications, I recommended systems or search. Um, so, so, yeah, so, so I do think there's a lot of things that have been around and if we can focus on like, once it's like the core problems. Like, core solutions, court, fundamental technology, I think hopefully we will. We get less overwhelmed.

Yeah. I really like,the approach you mentioned about looking at the problems

instead of looking at the hypes of the technologies. how do you spot those problems?

a lot of it is through, um, try and error like experience. Like, if I do, for example, like, if I look at [00:16:00] writing, um, I think, like, see, like, what, what is hard about writing and see, like, try, try to use AI for different things. Um, also, like, it's based on really, like, personal interest. Like, if you ask the core problems about, like, quantum physics, I would not. No, the

core from school for the man. Yeah. Yeah. So, I feel like if you have personal interest, personal, like, I have a lot of interest in, like, uh, Um, like educations, right? So, so I was thinking because as an educator, um, I think about what is hard, like what is help with learning, um, or like, as an engineer, I think about, like, what kind of jobs, like, how, yeah, like, how to solve, like, challenging challenges and, like, making people better engineers or help us build, like, better software products, um, as a writer, or I'm so interested in gaming, game design. Um, so, so I think all of that, like whatever you're interested in, uh, I do, I do, I do really have a lot of, um, respect for domain [00:17:00] expertise. Um, and I do think it's like nowadays, especially with AI capabilities getting like really, really good. Um, which is a big challenge, building good AI product is like product experience, understanding users, understand the problem space, um, you're in.

Yeah, let's switch gear a little bit.

what excites you most about the current state of AI like AI agents? what's most exciting to you?

Hmm. So things that what really what's really exciting to me, um, is. New applications in AI, and I think that's like we in the last two years ever since, uh, I think we got pretty excited about like, wow, what I can do this and I can do that. Uh, but like just a big gap from like a demos of like what AI can do and actually making them into like good products.

Yeah, I think that it was the last few years. I think we are learning more and more. About like the framework of like building like [00:18:00] those applications, best practices, challenges around in the process and then what to do about them. So I do think we have better understanding and which allows us to be like more complex.

I can actually like useful applications. So, uh, so I do think it's like, um, so that I was like, I'm really excited about. And also, uh, another aspect is that, um, Last year, a lot of people told me that one of the biggest blocks for AI adoptions is governance and compliances. So, like, a lot of companies can't, don't want to deploy applications because they're worried about, like, a lot of, like, do you have, like, licenses of the models or, like, is the AI trained on, like, data? Copyright data or like what if the AI like responds something risky, can contain BI information? So, so I do think it's like, I'm excited to see more of the space being, uh, not quite resolved, but have better guidelines and maybe more movement from like AI law, [00:19:00] um, regulations so that build more trust Also like education.

So like the more people understand ai, the more users understand how AI works under the hood, then the more trust it will be in these applications.

I actually had that question earlier for you as well about what you're seeing as People are resisting on adopting AI in their work or their products, right? so you're saying like governance, the risk associated with AI usage is one big item,

right?

like definitely like governance compliance is one big thing. Um, I think there's other, another is like, uh, I think it's just like with every new technologies, there are people who are like, can be like scared. Um, for example, as a writer, right. I'm like, Oh, like engineer. I'm worried. I would put AI, like take away my job. So, so I do think there's some, uh, I think in the past when, when we introduced like industrial revolution, when new machines were introduced, we do see people like go and try to destroy. those [00:20:00] machines because they thought that they were taking away jobs. So I do think it's like social implications of AI can be, it's very important to think through. Another question of like safety for general. So, so nowadays, uh, I think we live in the Bay, right? So we, we are fairly exposed to like new technologies, but think people like my parents who live in like a village in Vietnam. Um, so they don't quite know like the new and what AI can do yet. Um, so, so it can be like scariness.

Imagine like, is it, it could be like good target for scams, Right.

Somebody to send them like a videos of me, like totally, I don't know, fake videos and the say something, Hey, transform and immediately, you know, uh, like, um, it can be quite scary. Like we can see a lot of, yeah, demographics could be like targets for, for scams or even like educated people.

I've seen like some, some, my group chat, people sharing news. It's like totally not true at all. Yeah. Um, so so I do think it's like, keeping people safe would be 1 aspect. [00:21:00] Um, yeah, like, that makes a lot of people hesitate to, like, adopt a.

have you seen any successful use cases? that AI or AI technologies has been adopted and use in any companies that worked really well Have you seen any?

Yeah, I mean, I have seen, I should have seen a lot. And it's the funny thing is that

like the most successful companies I've seen actually don't talk a lot about that

because I feel like if you find, if find good applications, it's printing money. Like, why do you the secret? so, so, so do think I do think a lot, a lot of it, um, going pretty well.

what kind of use case are they,

So, so definitely. Um, I think that, um, a very classical use case,

a lot we using is like, uh, coding, right? Like, I think that coding is really good. Yeah. It's pretty good at like helping people, um, write code. Um, I think like, of course it's like not all coding equal, right? Like is equal, like there's some coding tasks [00:22:00] easier.

For example, like writing documentations pretty easier like. Um, like writing, like, low level systems, compiler code. Uh, so, for example, like, um, a friend of mine was, like, looking into, um, getting, like, AI to Nigeria, like, compile code. And it's actually pretty hard. And the reason is there are a lot of, like, not a lot of low level code examples on the Internet. He used to train this model, but they fill out the aspect of, like, coding, like, a lot of simpler tasks, like this, like, like, documentations or or, like, a small, right? Small functions or, like, for example, for me, I'm not, I'm not, I'm not an expert in JavaScript, but then I want to build some a test website for my, my project right now.

I can actually do it pretty quickly. A lot faster than I did before, um, all I did. Alice is actually pretty useful. So I actually asked is like, Hey, I have this like data as this graph, like, help me come up with, like, a better way to, like, represent, like, the data. So, like, it's pretty good, like, brainstorming a bunch of my different graph.

Um, and [00:23:00] before I could still do it myself, just take me like a bit longer. But now I can do it instantly. So, coding is pretty useful.

We see a lot of like, um, chatbot use

cases and it's something on the fence about, uh, because like, in a way, I think it's customer support chatbot enterprise use case. And so it's pretty good, but I'm talking about more like personal chatbot, like, companionships, um, I think those, those are pretty popular and we see a lot of people already spending, like, way more. Talking to bots and with humans a lot of time. It can be useful. Like, uh, first of all, like, it's I think there's a lot of value in being able to interact with technology without having to type. So, like, from the text base, right? But now we can do voice. Questions like interaction, I think it's like those pretty very, very, very useful.

Um, yeah, um, I think there are a lot of others use cases, like, in gaming, like, [00:24:00] different, I think it can be huge for educations. And it doesn't have to be something big, like, can only, like, assist a lot of teachers or, like, students in learning. We already see a lot of, like, students, like, using AI to, like, help with homework. I'm saying it's a good use case, but I'm saying it's, like, the potential of it. Like, for example, instead of, like, having, using AI to do, like, instead of, like, creating exciting exercise for students that can easily be done by AI, maybe teachers will need to think through, like, what kind of exercise students should leverage AI for. And still. Learn something new. Does that make sense?

yes. so you're pro students actually using AI, right?

But how do you value do evaluation is going to, should be. Rethink.

Yes. So like how, yeah, so I feel like as, um, so school shouldn't ban ai, but I should incorporate AI into the curriculum. Uh, so like some, some pictures, some like a very interesting, um, cases I saw online, like, uh, so for example, a teacher [00:25:00] ask each student like, like, so they have a prompt for an essay. Um, so instead of asking the student to run an essay about that, which student can use AI to generate, so he's like, okay, you would use AI to generate the essay. But then you have to go through and, like, correct all the mistakes the AI makes. Uh, because a lot of hallucinations, you have to fact check them. So, if you're a good exercise, or, like, it can be a debate partner, or you can just, like, help you, or, like, it can, I think, be very cool for, like, adaptive learning. Uh, because something I thought about is just, like, we all know that everyone is different. And that's why we have personalized ads. Like, we see, like, you're going to see different ads from me, and probably see different from, like, some, some other friend. Uh, but then, like, Education is like, somehow, like, we,

we just don't have educations, like, really adapted to a student.

I think that's something I can be, like, incredibly good at, like, first of all, like, if you like, uh, elephant, then maybe you can have examples, like, using elephants as you have, you understand the concept more.

25:58] angelina-host702_1_01-10-2025_120433: I think Chinese culture [00:26:00] is the same. Like in the ancient times, Confucius says that you should teach each person differently,

that's tailored to their own talents. uh, thinking about still curious about, like the companies,if some companies are successful at integrating AI, Are you seeing, like, how are they evaluating the ROIs?

do you see, what's the trigger for people or executives to decide whether they should use it or they should build it or buy it? Are you seeing them using some evaluation?

Yeah, think, that, like, the one thing is, like, I found out, like, when, um, a lot of, like, technology executives actually really smart, uh, and, and they actually, like, thought, I think like why there are suddenly companies that like make decisions based on hype. I see this like the vast majority of like sensible business decisions are still driven by like return investment. So it would do things through like, uh, do things through like, um, what's the return [00:27:00] on this? Um, and I think it's not a coincidence. It's like a lot of the, some of the most popular AI use cases today are those that you can evaluate the return very clearly. So first of all, recommended systems. You can evaluate like how good I even have an e commerce website, right?

You can evaluate whether your recommendation system is doing well, based on like, where they increase the purchase through rate increases sale. All right, my mobile fighting, what's a looking for and purchase it or like, with, um, for detection, you can measure is be like, how many, like, fraudulent transactions that the algorithm stops with coding. Coding is like one of the. Rare use cases of AI, general AI nowadays, where you can actually evaluate beyond functional correctness, like, okay, you have a generic code and you can see whether the code compiles, whether it generates the puts that you want and whether it does so efficiently. So, so I do think a lot of the use cases are very much like evaluations or like returns, [00:28:00] reinvestment driven. Um, so that's why I still think it's like evaluation is like one of the biggest, if not the biggest, Bottleneck for adoptions, because you cannot like evaluates the outcome of a I adoptions. It's going to make it very, very hard to adopt a I

What is one question that executives are asking themselves? Are they asking, like, How much more sales I'm going to generate out of like integrating with something like this,

yeah. So, so I think that's like, um, you see like the three different angles to it, like when I talk with like executives. So, like, one is that, like, they want to, like, needs to strategize, like, how much they should care about AI. So, so I think it goes from different angle, right? Like, um, so, so the business is at, like, a huge risk. Of like, a, I could just blow them out of the water, right? And we see something with check, like, having serious homework. Um, so, so I it's a business, like, existential space, existential threat from a, I didn't need to act like, pretty [00:29:00] fast.

So, another, so the 2nd angle is, it's like, what kind of use cases should we focus on? Because there's so many different use cases out there, like so many potentials. And once you decided what use cases, and I think

Microsoft has a pretty good framework to talk about, it was going to crawl, walk, run. So that means it's like, you can start with something that's like, um, um, lower risk. Uh, maybe you can involve something in human in the loop because you don't quite trust AI yet.

So one example is like customer support. So instead of having AI to generate like the respondent sent user directly, maybe you can start by like, hey, like, incorporate, like, use, um, use, like, customer agent, human agent in the loop so that the agent human agents can maybe see a few suggestions. From a I and choose the one just like the thing is good.

So instead of having to write responses from scratch, you can I use it's like a response as a base and continue from there, which could be a lot [00:30:00] faster. So another is like the next step is like a walk. It's like. Instead of like, uh, giving AI more like, so the step, you give AI more automations. So, but instead of like showing to like external users, you can use it for internal use cases, maybe like a Slack chat bot.

So that's where like the employees, so, so like when you do it internally, it's less risk and only when you feel really confident that you can like stop like showing it to like external. It's only users. Um, so, so, like, that's the 2nd aspect, right? we talk about, uh, the 1st aspect is, like, like, how much to give AI in the 2nd, and what use cases to focus on.

And the next 1, the last 1 is like, the human aspect, like, you have the teams, you have, like, those employees, like, how do you, how do you upscale them? So does it get affectionate leverage AI? So I see like exactly if it's like, oh, some, some teams, they will like give everyone access like free subscriptions to different services.

So ideas like, no, just go play with that. I see like, try to use them as much as possible and [00:31:00] see and see, like, understand them more. I give them courses, training, or like, Giving them, like, opportunities to experiment, building demos, maybe holding some companies, like, um, hackathons. So, like, the idea is to expose your employees, um, so that they learned and better utilize AI in their work. Like to be a product.

what's your prediction of 2025? What do you think is going to happen in the field of AI?

Well, I know for sure what would happen if I first we're gonna have more dramas for open AI second NVIDIA shares in the country to fluctuate. Um, so I feel like I think it's like, I think it's a joke joke answer. Um, but I do think it's like, we will see. A lot more exciting use cases, uh, in AI in 2025. Um, I think, uh, there are certain, um, areas.

So I don't really know everything about like the wall about like AI. Like, I don't, [00:32:00] um, I tend to my focus. Like I keep focusing on A few problems that I'm really interested in, so I'm excited to see more, um, progress in like, um, more applications in those space. But in general, like, I do believe that as we get better, so, you know, how you do a setup, right? You don't, you don't start by, like, trying to take over the whole big market.

start to start with some segment and I go really wild it and then and then branch out. I do think this is the idea that you can apply to the personal interest. So instead of trying to keep up with everything, which can be extremely overwhelming and impossible, like, I tried to narrow down on a few things I really care about. And the thing is, like, a lot of those experience you learn through caring about a few things that's important to you can be applied to, like, a new areas.

Um, are you optimistic or pessimistic about AI in general? Not just [00:33:00] 2025.

I'm very optimistic, actually really, really excited. I do think that's like one thing about the availability of like so powerful model. Um, I like accessible to users. People is that like, it's really low as entry barrier for people should be a applications. Um, so, uh, before, like, if you wanted to build a applications, you had to be able to build models.

If you had data, you to do. Yeah. So, so it's like, only a few organizations could afford to do so. But nowadays, like, anyone can do it. Like really, really anyone. So, so I think it serves the energy from the company community is incredible. Um, and of course, I put any new technologies they wouldn't bring about like social changes and we do need to care about that. Um, and I think a lot of things can go wrong, but I do hope that, um, I also see a lot of thoughtful people working on on that. [00:34:00] So, I, in overall, I think to be a net positive, and I'm very excited.

Well, how about you? Are you not?

I'm optimistic about a as well. but I do feel overwhelmed and I feel a lot of anxious. I've been like, reading papers, things like Elmo paper, fancy ice outfit and then transformer and all those kinds of things.

Right. And then. it's an explosion of things. to be honest, I feel pretty overwhelmed, anxious about it. And so I need your help on what do you think is the future? Like, how do we survive? either as developers or people with any kind of roles, They could be data scientists, a finance person, or,

how do we survive this age?

I do think is that because of world changing, the individual will also have to adapt. So I think, like, a lot of jobs will change, but we also have a lot more new opportunities. So the question of, like, software engineering, I don't really like a webinar. From Stanford, um, so it was [00:35:00] from, uh, the chair of curriculum committee for the CS department.

Like, Mira, Sammy is absolutely amazing professor and then Andrew was absolutely amazing because we're talking about, like, um, how would the future of the engineering job will, um, how will, how will it change? And one thing I really like about what they said is that putting by itself is a pretty manual task, like coding, like writing code, right?

I was putting. It's just not what's the rendering is about. So engineering is about, like, solving problem, like, building, like, good software products and writing code is actually a very small part of it. So, if it can be automated, that's great. But, like, as an engineer, you still need to think through, like, What problem you want to solve, how to come up with like an architecture solution to like solve that beautifully, and then you can have like, you can leverage AI help to like write the solutions more like faster. Yeah, I hope, I hope, I hope that helps.

I agree with you I think AI makes our [00:36:00] lives easier in a sense. We are more productive so that I want to spend more time. I want to give that time back to myself. So I have the time to think. And meet people, make friends. And actually I enjoy more in person meeting these days rather than like just consume online information.

It helps me to read paper faster,but that doesn't mean that I want to read more papers. You see what I mean?

Yeah, I certainly hear you. I feel like having time to think is really underrated.

It's right?

yeah, I think it's, yeah, I also feel like the more we consume, news, actually the less time that we have to just, let things sink in, and it's yeah.

can we predict that if AI will facilitate our lives, makes, make us more productive, and if we can actually give time to ourselves, and to live our lives and think more, and enjoy life, it's gonna be more optimistic [00:37:00] in the future.

it really depends on each individual. everyone's a bit different. one thing's interesting. It's like we've been talking about technology. So it's a little bit of like cynical view. It's just like

wanting there. It's just like. That's an interesting note. I'm not sure how true it is. Somebody please help me fact check. Um, this is like, um, technology making people more productive for, like, many, many years, right? But the number of working hours in a week has not gone down. Like, we still work 40

Oh, yeah. Oh, true.

Yeah. So, so, so I think this is not just technology itself. I do think it's like, We need some like social changes or, uh, corporate culture changes.

Um, yeah.

Um, thank you so much tip. I really enjoyed chatting with you.

It's an honor to, learn from you.

Thank you. Appreciate it. You're very thoughtful. I really appreciate [00:38:00] you.

Inviting me on this podcast.

Yeah. Thank you so much.